🌏 Exclusive! Grab the NordVPN deal ➼ nordvpn.com/angethegreat and get extra subscription time. Try it risk-free now with a 30-day money-back guarantee ✌
If I remember correctly from the physics of musical instruments course I took in undergrad, the cylindrical bore closed-open configuration limits the harmonic content above the fundamental to the odd numbers partials only. A conical bore does not. (It has to do with nodes/anti node locations in the pressure wave at the closed end) this is why a saxophone and a clarinet sound so different. And I believe this is why your clarinet is a more convincing approximation than the trumpet.
Former professional tuba player here - you warned us about the intonation and the difficulty of tuning all of the lip & instrument parameters in real time, and then when the sound simulation started... it sounded *exactly* like someone struggling to get a good tone from a weirdly-built trumpet. This is IMPRESSIVE
exactly! it sounds like someone playing one of those really shitty plastic pocket trumpets! i was never pro, but i played brass and woodwind for many years. well i suppose i did get paid to play a few times so maybe i can claim to be professional lol. i've heard a lot of weird horns over the years though, and the sim sounds like a weird plastic trumpet in its mid range. in the low range it becomes more "cheap yamaha keyboard" to me.
@@Gunbudder It makes sense to say it sounds like those plastic trumpets I think, since the simulation does not take into account things like materials and how those materials vibrate through pressure fluxuations. Perhaps ATG might add something like that since there is a big reason why professinal instruments are brass. Though I have never played a instrument. The sim is really cool and I like where it is going.
@@nimits9437 yeah I don't think the material was brought into account for now. More of a big deal when talking about instruments instead of engines. Anyways reeeealy cool stuff!
In reality the geometry and material properties of the brass and it's connection's, as well the contained and surrounding air produce very complex resonance characteristics with very similar dominant harmonics but different timbre due to small deviations and imperfections.
@@wordsareuneccesary Its a society thing. Even if he doesn't care, others do. Employers do. Employers that could give more money. Its practical, especially if he's done the work and just needs to do some paperwork and formalities.
@@MrCreeper20keven discarding societal labels, PhD work is excellent place to get big sound corporation sponsorship for this amazing work. From corporate management perspective, it's helluva lot lucrative to get PhD work and university tie ups for investor confidence than let's say a youtuber. Number of patents and publication under the belt go a long way
@@MrCreeper20k Why should he care about whether employers do? I don't think this guy is making this trumpet for money either. I'm sorry mate, but following societal expectations is not the road to happiness for everyone
@@yesto9676to defend his point. Its not a road for sadness either. Nothing wrong with submitting it. He literally loses nothing(maybe time and other stuff). And the guy just gave a suggestion. And sorry mate but following the societal expectations is not inherently wrong. He can do it if he wants to. I dont know why you are so defensive.
As a physics student and as a hobbyist trumpet player I just want to add my name to the choir singing this video's praises. It's really impressive what you have managed to pull off here!
What is really good about it as opposed to a convolution model is that you can generate NEW instruments with this! and then make a convolution model out of them *at least*, discounting other possibilities of this running real-time
@@Beatsbasteln well as he says, the cpu / gpu demands are insane - plus an audio plugin has a lot of different demands. You’ve really got to be in that field to pull off that stuff well. But the technology and theory is very exciting. It would just take a lot of work to move it from a purely simulation based goal to a creative goal
@@NightMind0 exactly - it’s a step beyond the delay line based physical modelling that has come before. Full on fluid dynamic simulation would be an insane game changer from a technological point of view - but such an idea would live and die by its implementation and workflow.
I think the reason you need lots of air to get good sound is for a very simple reason, the bell will enhance acoustic coupling and also resonate, making the instrument louder and sending that resonant acoustic energy back into the trumpet.
Think also because the actual embouchure opening is much smaller than the instrument bore. Probably only a few millimeters compared to the trumpet bore at like .468”. Sure it takes air to play, but it’s more about compressing the air with your lips.
@@prkassel Yes I agree, its not like your lips are flopping around in the mouthpiece like what the simulation depicted. As well, when you play the trumpet, youre trying to pressurize the air from the bottom of your lungs to your lips using your diaphram. Creating a large area of high pressure before a tight embouchure. Hope it helps Ange, fantastic video.
The mouth isn't a brick wall either. Some amount of backpressure will cause some amount of air to slosh across the lips. I think this video shows the lips being modeled as a perfect pressure source. Making it somewhat softer might lead to a mouth resonance.
physical emulation of music instruments has been developed since a very long time in the pro audio world, and contrary to this "experiment" they are ressource efficient, and sound same as actual instruments (which is the purpose). A lot of engineers have worked on the topic since a very long time with actual results used everyday by audio engineers, producers and musicians. You just ignore everything of this world, but it doesnt mean it doesnt exist, and it doesnt mean what you discovered today is any kind of achievement or "revolution", far from that.
I know this is really not the goal of this project, but as a music producer, I have to say, that this would be the most awesome vst plugin for physical sound modeling… especially with the possibility to create completely new instruments and tweak real world properties, this would be an entirely new way to sound design!!! Sooo cool
The swam trumpet from audio modelling does that. I believe their approach to pgysical modelling synthesis is a even more sound oriented and it is super efficient and realistic
Trumpet player here. Your simulation is uncanny. The tension in the lips being higher reduces air volume required and increases air Velocity. The resonance of the horn provides orders of magnitude more amplification than you would expect. The density of the metal plays a role too. Lots of sympathetic resonance in real instruments. Simulation inception would be necessary to capture all of these effects. The fluid simulation is but one part of it. It's probably the most important though which is why your results are so close. Love it man. You should share this tech with music companies like Roland, Yamaha, Nord, etc. They could really benefit from your achievement, and their effects and signal manipulation, user control, and expression technology could make this sound 100% better. Like a quantum leap in woodwind and brass synthesis. Patent it please. Open source it. Share it some how. We need this in synthesizers!
Share? Hell no, he spent a lot of time, effort and I bet fixing all the bugs decreased his lifespan by at least a few years. I looked up roland - 3k employees in 2013 and you say to share it with them? He will get 0 profit, credit or anything from that
@@shauas4224 if you open source it you can open source it a way such that it cannot be used to generate profit, there are licenses for that. Or he could just BSD type beat licensing and say fuck it. patenting it is admittedly pretty cringe, especially for home gamer stuff.
@@killingtimeitself yeah good luck finding out that company won't care a single bit about license and then good luck finding out that they used it since I highly doubt any of companies will open source their software as you say author should do. Don't get me wrong, I m also curious to dig in the code and see how this works but downside to that - companies will abuse and steal his creation
Trumpet player here. I’m surprised how good the trumpet in your model sounds. The characteristic sound definitely comes through. It sounds a lot better than many Synthesizers I heard.
Not _most_ synthesizers? If it _wasn't_ better than most wave-based synths, that would be disappointing. Is it better than the _good_ ones is the question, and comparing it to Samplemodeling and even Audio Modeling gives you the answer to said question.
As an engineer I would 100% pay for classes taught by you, very clear and to the point. Demonstrating a high level of dominion of material and amazing pedagogy skills. Keep it up! 🚀
Unfortunately my virtual trumpet skills are just as bad as my real life trumpet skills. Having a semi-realistic simulation didn't help matters but kind of made things worse.
In the overblown demo, I was amazed it recreated the "double buzz" effect, where a lower frequency is also heard. This usually happens when a tired player can no longer control their embouchure (lip stiffness and geometry) and unintended vibration modes are produced.
Also, I wonder (1) what would happen if you're fully simulating the valves and what happens when you only depress the valves half-way. Half-valving produces a very distinct tone and diminishes the volume. And (2) if you can slide the stiffness of the mouth to recreate a lip slur between harmonics.
As a musician by training, I can tell you that this trumpet simulation sounds precisely like a trumpet -- and before adding the reverb, it sounds precisely like a trumpet in a well-dampened practice room. Awesome job mate!!!
@@gownerjones There will be some reflections from the ground, but if you have a soft ground, like grass, it will be minimum. To make an echo you will need a hard surface for the sound to bounce off (off). Where I live it is much easier to find a place with no echo that one with. But if you live in a mountainous area it might be harder. But you can try the top of the highest mountain or the peak with the larges distance to the nearest mountain.😀 Also you need a low wind speed to reduce the bagground noise and new fallen snow can also help you reduce any reverbe or echo. glhf.
I've been playing the trumpet for about 8/9 years now, it's absolutely amazing how close this sounds to a beginner trumpet player. The coolest part to me was when you did the model of increasing airspeed, the way that the pitch increases, then decreases and distorts is EXACTLY how an actual trumpet plays when you do the same thing.
the meaning of "great" isnt so definitive, so maybe it kinda isnt? a baby finger painting for the first time is great, but so is winning the lottery. obligatory 🤓 here
@@genralty ackshually a baby's first finger painting isn't anything great, as it's not an art and even among baby's firsts it's not that important as walking or a handprint 🤓
as a first chair trumpeter, i can say you may be getting the weird intonation from lack of resistance based on the valve position. no 1 valve generates sufficient resistance but there are a few combinations that do generate resistance at certain octaves due to the resonance. to someone without experience playing it can be hard to feel it, its subtle but very significant. all in all, very cool to see this done.
I think the biggest reason there's not much resistance is that the model doesn't have a mouthpiece yet. We've basically got a really narrow didjeridu so far. (I think the model is doing a great job giving us what it would actually sound like to play a tube that's shaped like that!) Once we get a mouthpiece that's the right shape, it's going to give us *way* more realistic resistance.
@@thomasscheevel5779 i feel really stupid and am kicking myself in the ass now. i totally neglected the fact that different mouthpieces have an effect as well. good spot thomas
@@puddlejumper6999 tbf, i probably wouldn't have recognized the sound if i wasn't the kind of person who's tried to play brass (a tuba, for me) with no mouthpiece before. so when he went in the video about how surprised he was about how much pressure it took to get a sound, it just clicked - yep, that's how it feels and sounds to play a brass instrument with just your face right on the leadpipe
Valve horn player here. Great job on trying to capture this complex procedure! Couple things I know from playing a brass instrument: - The most important part is the lips that distincts between a good player and beginner. I can make my instrument sound like your simulation, just by adjusting my lips. - Air Pressure and Lipstiffness increases with higher notes. Playing loud and deep is as hard as playing quiet and high. But one can compensate this by having a more accurate and more stiff/loose lip - The way I let my instrument sound different with the same pitch is by increasing or decreasing the volume of my mouth. Bigger Volume results in a more round tone that is often nice to hear. But sometimes in certain musical genres you want this duck like quacking and therefore you decrease this volume (by adjusting the tounge mostly). This also is dependent on the tone - higher tones are generally played with less volume while lower tones require more volume for the same quality of sound. Do with that information what you will, just some insights from my end.
Thanks for sharing your knowledge! It would be cool to refine the lip model using information from real trumpet players since most of the parameters are purely physical/scientific
As a trombone player, I agree with this comment. I think to really improve the sound you would need to simulate the mouth cavity (up to the whole body) as it acts as a resonance box. It would be fun to play around with a trombone sim as it is simpler than a trumpet (sim-wise, there are no valves) and it would illustrate how the pitch changes when the length of the tube changes. Awesome work, I really enjoy what you're doing !!
@@AngeTheGreat You may try to simulate buzzing just on the mouthpiece. Some consider it a bad practice because the embouchure for buzzing is not exactly the same as for playing the instrument but without the brass tube's resonance you can freely change the pitch so might be able to see better what lip configuration is in tune with each tone.
@@AngeTheGreatDr. David Wilken did his dissertation on brass embouchure, and he has some interesting videos showing brass players playing into transparent mouthpieces. One of the more interesting things I learned from that is that not only are the lips not centered vertically on the mouthpiece, but a brass player will adjust the mouthpiece position depending on range.
One reason I love these simulations is because they push the limits of computation. With web development dominating mainstream, it’s refreshing to see non-trivial C++ projects.
I used to be a trumpet player. I am very impressed with this. Especially the over-blow bit. The simulation was able to pretty accurately predict what actually happens when you blat. The mouth simulation would need a lot of work to be accurate. There is much more to embouchure that would need to be considered. You'd probably have to model a whole head. Its all very impressive!
Agreed to at least the whole mouth, not sure about the whole head. Like, how essential is your skull when buzzing, y'know? I think simulating all the way to the back of the throat would make the most sense, because of how important the tongue is in shaping your tone. Still super impressive!
@@angelicbeast7793 i think will's point was once you have put in the effort to model the mouth, you might as well just simulate the synapses in the brain that decide which note to play :DDD
Im a trumpet player and not an engineer. The thing i like the most in this video is that, although the trumpet isn't tooned, the fist demonstration of the trumpet's notes is as accurate as it gets. Although im not a luthier, i can definitly say the chromatic scale is exactly as its portrayed: 000, 010, 100, 110, 011, 101, 111. Descending, not ascending.
I'm a trumpet player too, and loved that this video is a defiantly on point for how a trumpet produces sound. I think some of the parameters that are missing to be able to enhance the sound go something like this..... 1. lips, tension. There's several ways to achieve this, basically we tighten up the cheek muscles to make the lips more tense. In addition, we can change the pressure of the mouthpiece on the lips to achieve the same. 2. Instrument construction. The tone is effected by the mixtures of metals that make up the instrument. I'm talking about he zinc/copper ratio on the brass, which effects the "brightness" of the sound, also whether the instrument is silver plated, or lacquered. Again this finish can have an effect on the tone. 3. And here's a whole new can of worms, what's happening behind the lips. The tongue is important as it's placement will dramatically change the sound, as well how we open our airways in the throat whilst playing. Puffing out the cheeks. Also whether we are breathing with the upper part of our lungs or (as is better) the lower part using our diaphragm to apply the pressure that creates the volume. I believe that we tend to use higher pressure than you'd think to produce the vibration at the lips, but the opening is very small, so that volume of air passing is relatively low. It's common for a trumpet player to be able to sustain a note for about 60 seconds with one breath. This should be able to give you the idea of the amount of air that can be delivered with normal lung capacity.
@@Goliath83It’s counting in binary, except the digits are shuffled, because the pipes attached to the valve aren’t in ascending order of length. Instead of being worth 4, 2, and 1 semitones, they’re worth 2, 1, and 4 in order from closest to farthest.
As an audio engineer who sometimes dabbles in recreative game development, loves physics and used to play horn, this couldn't be more up my alley. I absolutely adore your recent projects, Ange. Much love and massive props for all the progress you have made in such a short time frame. You rock my dude :)
Its hard for me to describe how captivating it is, to listen to a truly passionate and intelligent person explain something they've made. Its just SO nice.
I mean, if someone says they do computational fluid dynamics, "engineering, acoustics, and coding" is basically what it says on the tin. CFD is vital to all sorts of engineering work, and coding is its own sort of engineering. Acoustics basically falls out of the equations that you use to model the behavior of air. And any field that has "computational" in the name is going to involve coding.
Brasswind player, mechanic, and "armchair engineer" here. A huge thing about trumpet playing is that the lips form a seal on the mouthpiece. I couldn't tell if what was shown is what was being simulated, but it appears the lip model wasn't fully pressed against the mouthpiece. This would make a lot of sense why it required so much more air to create a sound, as much of the air pressure can escape around the sides. Playing trumpet well includes having a lot of air pressure behind the lips, as well as carefully measuring out the correct frequency for the lips to move to match the note. Air pressure is achieved by tightening the diaphragm to force the air out as hard as possible. Face muscles control the tightness of the lips. The process of playing is an intricate dance of maintaining air pressure, matching lip frequency to the desired note, and using the correct valves. Also, the bowl inside the mouthpiece affects tone as it changes the shape of the lips and how far apart they can move. When the lips have a shorter distance to travel, they can trill faster, creating a higher frequency. Novice trumpet players can reach higher notes by pushing the trumpet into their lips, forcing the lips further into the mouthpiece to make it easier. Skilled trumpet players have developed their facial muscles enough to tighten their lips without having to do this. Ideally, you don't want to strain your neck by putting any more pressure than is necessary. Some trumpet players make vibrato by repeatedly pressing the instrument into their lips to vary the tone and pitch at the same time. Also, different mouthpieces have deeper or shallower bowls to make it either have a more "full" sound or easier to play higher. Hope this helps. I've enjoyed your engine sim adventure since the beginning. Seeing you make something from it that touches upon more than one of my interests makes me very happy.
Another tidbit, intonation is also controlled by the first and third valve slides. You would have to look up which notes need it, as I've forgotten at this point. I play the baritone mostly and haven't seriously touched a trumpet in over a decade.
It’s so crazy to me that like 8 or so years ago I had an idea that modeling instruments using physics would either happen or I would have to learn to do it myself. I would absolutely love to see a version of this project in vst format later down the road. As a producer this excites me so much!!! Keep up the great work!
Well, physical modelling is all the rage now, look at Plasmonic or SWAM. Swam is especially impressive at modelling the wind instruments, i wish it would be cheaper 😢
@@snesmocha That's what Samplemodeling is for. Listen to some good demos of it, i.e. Sam Paryss's Star Wars mockup using it or The Trumpet v.3's demo medley.
With like two more layers of simulation inception this could get scary close. Aside from the bell, material type also plays a large factor in sound resonance. This video is truly awesome, you did a great job.
Experienced brass musician here. It sounds like an early beginner trumpet player who is trying to find out the best way to make a sound! I'm mostly impressed at how accurately the valve movements distort the tones between presses, which sound *very* accurate. From the sounds of it overall, the tone of the trumpet could mostly be fixed by making adjustments to the player's embouchure (the mouth and its positioning on the mouthpiece). But I can say that I've heard these kinds of tones from the beginning trumpet player.
This is amazing - as a physics graduate working on particle simulations, and a trumpet player of 10 years, its amazing to see those two worlds collide. The fact that you can even get a roughly-correct chromatic scale and multiple harmonics working is extremely impressive. To me, it sounds exactly like a trumpet playing into a cushion, or a really closed mute. I think simulating the bell flare will definitely help (I've experienced many dented flares). For even better tuning, the third valve tube would need to be extended slightly in the lower harmonics.
Use the system to simulate the call of the parasaurolophus. It has a large resonating chamber in its crest that drastically changed the sound they make.
As a former trumpet and horn player and current software developer, BRAVO sir. The way the note breaks apart when the embouchure is too flat for the instrument on that note is EXACTLY how it sounds in real life! And the way the pitch transitions between pitches on input changes was spot on too
I've been a trumpet player for about 10 years, and indeed, the lip parameters are extremely important to hit the right tone. Trumpet players even anticipate their lip setting and sometimes also pull the instrument tighter to their lips to prepare for the next note, at least to minimise the air gap between each note. If not well prepared, it even occurs to players that only air comes out, completely missing the note. As found in the simulation, the amount of pressure lips have endure on a Trumpet is immense, professional players need multiple years of lip training and lip hardening to cover an acceptable tone range. It happened to me to have shattered lips close to bleeding after a concert. The simulation exactly sounds like someone with no experience in lip setting, which is the case here. Lip tension is even changing along the duration of the note, call it expressions, phrasing of the notes or modulation, like vibrato, staccato, pitch bend and more... Adding randomness in lip tension is an easy way to improve realism, but simulating expression with lip tention needs musical feeling, which is a learning process that probably only AI would be able to implement.
Came here to say something similar about the lips. The tweaking of the params to get the 'clarinet' sound definitely made that pop out. I Imagine more work on the flare will help, didn't hear any talk about the "material", i.e. certainly a wooden trumpet would sound much different than a brass one--I don't know how that's accounted for here... maybe covered in a different video, but yeah, so much is in the lips and I wonder if the original model is just inadequate for the needs here. Buzzing with even just your lips, you can quickly start to understand how much that contributes to the sound. I imagine he could also simplify things and just focus on perfecting buzzing with just a mouthpiece and then plug in the rest of the horn to a massive improvement. Hope he decides to take at least one trumpet lesson as I feel like that will be an epiphany on how important that part is.
I've played the trumpet for well over 10 years throughout middle and high school. I'm also self taught in C/C++ and 3D Graphics / Game Engine programming using DirectX, OpenGL and Vulkan. I'm also adept with mathematics and physics. To see a video of a physics simulation that uses C++ and Vulkan to project the fluid dynamics of a trumpet is really cool. This video is very informative, creative and quite awesome. Job well done!
I am a trumpet maker, and this is really an incredible simulation. There is only one other publicly available simulation software available for brass instrument makers. This could be a hugely valuable resource for people wanting to fine tune the shape of the air column before making tooling, or to correct things in R&D.
first thought on the first sound was "hm, straight cut trumpet sounds like Sax with a very, very wet mouthpiece", i guess it's more about the lip params and I wish someone will take it upon themselves to jump through the hoops making this the most versatile physical modeling VST-Instrument ever. Ange, you are breaking new ground in software development, with a perseverance and consistency that is nothing short of legendary.
This is so cool! It goes quickly from "huh, that's a neat simulation" to "wow, that visualization looks incredible!". I also appreciate the non-fearmongering VPN ad.
Played trumpet in junior high and high school, and I can say with confidence, trumpet is among the easier instruments to learn, and if you up and bought one, I'd highly encourage trying to learn it.
I love that it takes multiple research papers to understand the effects of your lips on sound, and musicians are just like "it's right when it sounds good"
I really like the particle visualization for fluid flow! It reminds me of… an infomercial for a vacuum cleaner. I think including the ability to create extruded 2D shapes in engine simulator would be a good feature, because that would allow simulating different muffler designs.
Trumpet player here, been playing for about 13 years and just finished my degree in games development. WOW!! Christmas has come early! I find it super interesting that despite modelling it off a real Bb 'key' trumpet (i.e, when you play a 'C', it sounds like a 'concert pitch' Bb), it sounds like it's tuned to Ab. It was a super minor section in the demo, but when you demonstrated the crescendo (going from low volume to high volume), I was blown away by how it accurately simulates the sound 'opening up' - something that trumpet manufacturers can play with to change the character of each instrument. This is awesome!
This is absolutely incredible from start to finish. The data visualization is a master class in visual representation, and the project itself is just mind boggling.
do you realize how much of a market there is for tools like this? sample libraries of physical instruments are costly to make and often inflexible to use. with a ui designed for musicipans, the ability to physically model wind instruments accurately and in real time would be a gamechanger!
Incredibly fascinating demo and also a very clear visualization of what exactly is going on physically in a wind instrument and why! I think one of the hardest things about the learning curve of a wind instrument in general is understanding and controlling all those various parameters in real time while only being able to abstractly grasp what the physics behind the instrument actually are doing, so this is amazing to see! Also I definitely laughed out loud when you started playing the Jurassic Park theme haha
make a realistic pipe organ, replete with all the stops etc. That would make not only a great steam game but also fill a market void of physically based music synthesis for musicians
Oh my god, this is an amazing project. As a passive fluis sim enthusiast, you have just reignited my passion for this field thank you! 1 tip, i think 1 of the reasons why you needed so much pressure to make a noise is the mouth piece. Not sure if the modeled mouthpiece is the same as the simulated one, but the concave section is too small. From what I remember from playing a trumpet in middle school is that the domed section in the mouthpiece was much deeper, and the walls where thinner. Hope that makes sense haha, can't wait to see the next update of this!
The resonance of brass as a metal is vital to the reduced air flow necessary to produce sound because every inch of the tube is added shaking of the instrument which aids in the increased efficiency of kinetic energy’s transformation into sound.
This video devoured my internet bandwidth like a snack... What truly blows me away about you, and this resonates with me (haha, resonates), is that you give your 110% in everything you do. In this instance, not only is the video content top-notch, but the production quality is off the charts. I mean, even your compression settings seem like you've mastered them to perfection just to deliver the most incredible results. Simply mind-blowing.
I find this project so inspiring, and I can't help but feel so much pride watching this! You've done so well! Edit: like seriously, as someone hugely interested in audio, music, physics, engineering, computer science, statistics, and many forms of visual art: this project meshes my biggest interests together in a way I'd never have expected. I cannot wait to see what's to come!
As a trumpet player, and a CSE student in a computer graphics class, this is amazing! Idk if it would help, but if an IRL trumpet student is overblowing, it often means they have to tighten up their lips a little. When you have air blown way too hard in the simulation, you can hear the pitch change as the pressure changes the forces on the lips. As a trumpet player, my best sound is made by balancing the tension in my lips with the air pressure. A way to measure if you hit a reasonable balance is to study the amplitude of the sound. When you have a good tension for a given pressure, the horn will resonate better and become louder. This is not necessarily the sound all musicians aim for, but it is a discrete and measurable goal that's good for beginners. I noticed that the simulated trumpet's tuning slides are all pushed in all the way. Pulling out the main tuning slide and the 3rd valve slide (at the same time as the valve is pressed) should improve the overall sound and make the intervals between notes more even. Another thing to consider would be to add tonguing, where the tongue is used as a valve to quickly start each new note. This would eliminate some of the out-of-tune movement of pitch when the valves and slides move. I really love this project you've made! If you have any questions about real trumpets, feel free to hit me up.
Would you consider doing a similar simulation for the human vocal tract? That would be a great educational tool. There are no truly good online resources that explain how and where the different formants of speech are created, the explanations are quite rudimentary. E.g. visualize how the same formant (in the output spectrum) may originate from two alternative configurations of the tract, due to the tounge position.
I don't think this type of simulation is capable of coming close to simulating human voices. It's useful for systems that consist of connected cylinders of air, not something with complex geometry. I imagine it would be very hard to physically simulate a realistic / convincing sounding voice.
if this could be possible, then it might be possible to generate voices from it, I imagine it would be impossible to generate the voices that is the pressure waves and other mouth expressions manually to output voice, we may train a neural network to generate the needed air pressure and toung and mouth movement to generate a voice, This could be the most realistic TEXT to SPEECH generator of all time. have goosebumps just thinking about it.
This project is insane to me in the best way possible. As a former awful trumpet player, the off putting part about the first demo was how clean the sound was with a stable air supply and no reverb. Once some reverb was added, it just sounds like a trumpet with some tuning problems. Can't wait for this to release and see what people can do with it.
The detail in the attacks of those notes is insane. If someone works out a comfortable mapping for pitch+velocity, this would be an incredibly useful VST synthesizer on its own :D Keep it up!
@@YCbCr This has got to be one of the most generalized physically modelled synthesizers right? Is there even anything we can download right now that would be able to simulate arbitrary horns and intonation at this level?
Truly unbelievable the fidelity you can get with a simulation, I cant beleive how you can get those note transitions without nuances of the valve shapes, and the sound quality with that simple lip simulation, amazing.
Oh my god. Oh my god. This is AWESOME! First of all, this video showed up in my feed while i was in a practice room practicing trumpet. Second, when this is finished, PLEASE PLEASE PLEASE PLEASE PLEASE make a VST3, AU, and AAX plugin for this if you can. Also! As a trumpet player and an audio engineering student I have a few observations, questions, and suggestions which you may or may not have already taken into consideration: It sounds like a french horn, or other low brass instrument. Making it sound more like a trumpet (i.e. generally brighter) is going to come down to the resonation of the metal in the bell. I couldn't tell if you were simulating the actual metal itself resonating, but the temperature, thickness, material, and plating of the metal is going to be very important to the sound. Heres a few parameters that I think could improve the accuracy of the simulation (and are also interesting to tweak): metal materials (i.e brass, copper), plating (gold, silver, none), presence of lacquer, Bore size, Bell size, Bell thickness, Different mouthpieces(??), just to name a few. I noticed that your tuning slide wasn't adjusted at all. Unless it is VERY cold in your room, you are going to need to put it out a little to get proper intonation. Also, when playing middle D (Valves 1&3) the third slide will need to be extended slightly, and fully extended when playing middle C# (123) for proper intonation otherwise they will be very sharp. On the topic of the bore and the bell, these things have a significant effect on the brightness and tone. I play a Bach Stradivarius Model 25L, meaning it has a relatively small bell (brightens the tone and increases projection) and a large bore increases the richness but requires significantly more air. I've played quite a few other trumpets and the difference in intonation and playability is massive. Mouthpiece - not sure if you did research into this. Theres a bunch of different parameters for this (cup size, depth, shape, rim size, thickness, etc.) and it affects both intonation and playability. Each mouthpiece has its pros and cons in terms of range, tone, and playability. You can buy a screamer mouthpiece for playing lead and screech all day long but it wont be very rich and warm. Im not sure if any of that applies tho, as theoretically the simulated lips aren't subject to pain or fatigue. Again, if you aren't already simulating resonating metal id say do that first, as it has an enormous impact on tone. I know thats a lot of things to simulate, but I think it would make a significant improvement. Also, I think the most impressive thing here is what you've done already. It sounds like a brass instrument, and it does tuba pretty convincingly. You just need a few finishing touches (and high end). It would be super cool to see this released to the public, but regardless I'm dropping a sub on patreon right after I finish typing this out so I can start playing around. The ramifications for this in the production community are insane. Hyperrealistic brass instruments right in your daw???? Theoretically, this simulation could be used to recreate room reverbs, without the need for an Impulse Response and convolution reverb. I wonder if there is enough performance optimization available to make a sim that large work effectively in real time. Anyways, this is super awesome, and bravo for making something real that I've only ever dreamed about. I can't wait to see how you improve this, and whatever other projects you have in store.
You're absolutely correct in describing how a simulation actually does a calculation, but there's a much simpler way of describing it (FEA incoming!). Some terminology: - Node: a single point representing a boundary - Element: a line connecting two nodes together which is represented by a characteristic equation. - Characteristic Equation: the equation that models what we expect to happen based some input (i.e. F = -k•x, ∆Q/∆t = k•A•∆T) Finite Element Analysis is the method by which all simulations solve complex problems that either have dynamic states, or shapes which are hard to represent mathematically. Depending on the problem you want to solve you can either solve a problem in 1D (a line), 2D (a plane), or 3D ( a volume). Typically, most problems are homogeneous (meaning the characteristic equation is the same throughout the entire volume) and we can get away with modeling things in 1D. In such cases, we can break up the problem into some number of nodes (n), and elements connecting the nodes (n-1). Each node has input and output variables and serve as a direct connection between elements, meaning that we can take the reaction from one element to accurately represente the reaction of the following element. An easy way to visualize what this looks like is by thinking about lines at a checkout. When the person at the front of the line is done ordering they move out of the way and the rest of the people in line move forward. Each person represents a node containing the current state, which in this example is position in line from the cashier. What you mentioned about elements only interacting with adjacent elements is a constraint on all FEA models in which we need the output from the characteristic equation to serve as the input for the next element in line. An easier way of thinking of this is by imagining imagining a highway with 4 lanes (2 in each direction). If 2 cars are in 2 separate lanes then they don't interact with each other. However, if 1 car is in the same lane as another car, then said car is limited in how fast it can go by the car in front. Hopefully this explanation helped. FEA is a difficult subject to understand for anyone who isn't in the STEM field, and throwing in fluid mechanics takes the complexity to a whole new level (source: I have my bachelor's in engineering with a focus on heat transfer and fluid mechanics).
Former trumpet player, now engineer who had to give up dreams of programming interesting side projects. I'm floored by what you've done here! Absolutely amazing! I was just recommended your video today, but will now follow with interest!
having heard a good deal of trumpet-including music, both live and online, i have got to say that this sounds pretty damn good for a fluid simulation. there were some small parts where the audio became noticably "rough" in a way that i only really know from computer-generated audio, but during some other parts it sounded like an authentic trumpet being played by someone just getting used to playing the instrument. i've got to say, i am very, very impressed by your work and i'm certain that the quality of your other computer simulations can only go up from here.
Dude I swear the entire time I was sat here expecting this to be a long drawn out VQ engine joke, but alas, you really did deliver! This is seriously impressive stuff, you can definitely hear the "life like" tone in the sound, proving the simulation is astonishingly accurate.
During your research did you use the book Computational Fluid Dynamics with Applications by John D. Anderson? It uses a finite difference approach which can be easier to program. The other very good book that maybe you used is Introduction to Computational Fluid Dynamics - The Finite Volume Method by Versteeg. This book really gets into the details of the solvers, PISO and SIMPLE.
First of all, this is amazing and thank you for showing us your work. I understand this is a giant technical feat and worthy of praise and I agree with everything you said. Second of all, the moment the trumpet started playing, I burst out cry laughing and haven't stopped since
I want to say that its impressive how you send the information to the viewer. I dont know most of the things you say but somehow I understand everything.
I was thinking about how you could use a fluid simulation to simulate a subwoofer or smaller speakers in a range of different enclosures with different sizes and such, and how different length ports and transmission lines change the sound and group delay and such, it would be a cool video, I don't know how hardware intensive it would be but it would be cool.
You could make it as a VST so it can be played with midi inside a DAW or other programs. There aren't many pysical trumpet simulation so it could be interesting
Because of the embouchure needing to be emulated, this isn't so easy to replicate or achieve using a MIDI keyboard, but I do agree that this would be great. Solo trumpet VSTs and sound fonts tend to sound less than ideal for a quality recording, so something like this would be incredible.
The sound breaking up at 18:48 is spot on for the state the simulation is in, I had that happen quite a few times when learning a decade ago and to hear a simulator replicate it almost perfectly is amazing
Dude you’re such a role model for me. I’ve always had ambitions to write a ton of simulations myself, but you’re actually doing it, fast and in amazing detail. Do you take apprentices? 😁
It'd be really cool to hook this up to a reinforcement learning agent to have it learn how to play or tune an instrument. Level of detail is amazing. Really tickling my physical simulation buttons.
The mathematics reminds me very much on my recently finished university project about finite difference methods for electrodynamics (Since they are also hyperbolic equations). You said something about further sources about the math behind your simulation is available in the video description. But I can’t find any link. Edit: Links are now available. Thanks!
I forgot to add them before, but they should be there now. I have some additional links that I'll post later as well, I'm just trying to collect them all lol
@@Erin-ks4jp I didn't mean as-is, I meant with more work ofc! :D Considering this was made mostly for a fun quick video, and the video mentions many things that need to be fixed to improve it, I think the potential is there though, if it was focused on.
I played the trumpet for almost 10 years and I can say that this is actually amazing to me. Despite its flaws it is the most accurate sounding audio I've heard a trumpet make artificially. Everything I heard sounds very familiar to what I'm used to hearing. I could almost feel the different pitches as if I was playing it, definitely a wierd feeling 😂 if there are any other players here, you might know what I mean. Now that i think of it, the sounds the simulation makes reminds me of what it feels like to play with a earbud in one of your ears. You can feel those same vibrations through the audio of the video.
It's amazing how you manage to get that running on a normal PC. Every time I try to simulate something myself from scratch I use way too much memory lmao
@@calvinkohl6220Learning the math is one thing, but to optimize it to make the code run faster is a different can of worm, especially if aiming for accuracy.
I was about to say, people would pay an arm and a leg for a VST of this. I think VSTs can't be multi-threaded though, so you'd need to run it as a seperate program and record and export the audio.
This could definitely be great for music production purposes as its honestly the most realistic trumpet simulation I've heard. Idk if it'd be possible but integration for use with DAWs would take this to another level.
This project is so cool! From what I’ve read about mouth pieces, they actually function as Heimholtz resonators, with another pressure boundary (or whatever its called, like the bell) at the base of the mouthpiece cup. This junction is the interior point at which the vibrating column of air reflects. In real trumpets, the size and taper of the backbore - as well as whatever gap there is between the end of the mouthpiece and the beginning of the leadpipe - has a large effect on tone and slotting (how easily the notes pop out when you play). And the size and shape of the cup also affects whats going on. I know thats a lot of extra stuff to model but would you consider it? Other fun aspects of brass instruments which I’ve wanted to see modeled for a very long time are curvature and bracing. The curves in the pipes of brass instruments seem to function as longer pipes than the straight sections. Brace placement/rigid points along the tube also affects slotting and intonation. I know you’re trying to keep things simple and only doing a straight tube right now but these are some things I would absolutely love to see. Amazing job!
I can't escape those Euler equations. I am in compressible aerodynamics and you jumpscared me with that partial differential equation. Awesome video! I love everything that you do. It is also nice to see your assumptions within this project. Cool engineering.
Trumpet player here, the tone sounded better than most middle schoolers and some on high school. This could be a great interactive teaching tool! Another important variable to consider happens before the lips, to get a nice wide tone, we open our mouths while keeping our tongues low. The smaller the mouth opening, the more we squish the sound.
To echo what a lot of other trumpet players are saying here, it really speaks to the accuracy and sophistication of this model to say that you have accurately recreated a beginner trumpet player. You said yourself that the embouchure is difficult to model. Well, a beginner trumpet player has little concept of embouchure and therefore sounds almost exactly like your trumpet model. You have perfectly modeled the trumpet, but not the player (which is beyond the scope of this project). Anyone who has learned the trumpet or been in a middle/high school band can vouch for the accuracy of this model!
I would give anything for university lectures for every course, or even just 1 of my courses, with this kind of clarity & production quality. Crazy part is, watching this & learning is $0, whereas university cost me $31k/yr. Awesome video Ange, Thank you for your massive contributions to the community!
I have undiagnosed epilepsy and this entire video was triggering me really bad. 😅 (I don't seem to have normal photosensitive epilepsy and it seems to be a time dialation thing, slowing down and speeding up things) I had to keep looking away for most of the video, but I really enjoyed listening to this. Your videos are super easy to understand and really helps give insight on how fluid dynamics work in the motor. I have been doing a lot of research while building some headers for my engine, and watching your videos have helped a lot to visualize and tie everything I am learning together to understand how everything works and affects things. Thanks for the heads up in the title.
Do you have any ideas on how I could fix this? I don't really know too much about this subject and it's hard for me to judge what the problem is and how it can be resolved
@angethegreat i mean if you want to correctly visualize sound waves, i dont think there is any fixing it. For me, its that constant speeding up and slowing down that gets me, which is just going to happen as you increase frequency visually. A possible fix for this video might be not using a black background, but visually, the black background looks great. I have not had any triggers whyle playing with your engine simulator though for what it is worth
🌏 Exclusive! Grab the NordVPN deal ➼ nordvpn.com/angethegreat and get extra subscription time. Try it risk-free now with a 30-day money-back guarantee ✌
Ange the great synth-engine simulator pending :-) as a musician & auto enthusiast, you sir, are my hero.
I feel like i've seen that "exclusive" and sometimes "limited time" deal on every other channel on youtube over the past 4-5 years.
If I remember correctly from the physics of musical instruments course I took in undergrad, the cylindrical bore closed-open configuration limits the harmonic content above the fundamental to the odd numbers partials only. A conical bore does not. (It has to do with nodes/anti node locations in the pressure wave at the closed end)
this is why a saxophone and a clarinet sound so different. And I believe this is why your clarinet is a more convincing approximation than the trumpet.
Exclusive! Grab the fuck off blatant advertiser.
Get that bag king
Former professional tuba player here - you warned us about the intonation and the difficulty of tuning all of the lip & instrument parameters in real time, and then when the sound simulation started... it sounded *exactly* like someone struggling to get a good tone from a weirdly-built trumpet. This is IMPRESSIVE
exactly! it sounds like someone playing one of those really shitty plastic pocket trumpets! i was never pro, but i played brass and woodwind for many years. well i suppose i did get paid to play a few times so maybe i can claim to be professional lol. i've heard a lot of weird horns over the years though, and the sim sounds like a weird plastic trumpet in its mid range. in the low range it becomes more "cheap yamaha keyboard" to me.
Too good to be not real.
@@Gunbudder It makes sense to say it sounds like those plastic trumpets I think, since the simulation does not take into account things like materials and how those materials vibrate through pressure fluxuations. Perhaps ATG might add something like that since there is a big reason why professinal instruments are brass. Though I have never played a instrument.
The sim is really cool and I like where it is going.
@@nimits9437 yeah I don't think the material was brought into account for now. More of a big deal when talking about instruments instead of engines. Anyways reeeealy cool stuff!
In reality the geometry and material properties of the brass and it's connection's, as well the contained and surrounding air produce very complex resonance characteristics with very similar dominant harmonics but different timbre due to small deviations and imperfections.
You honestly should submit all of this work as credit towards a PhD. It is mighty impressive!
Why would he care when he can just add real value. that’s another title.
@@wordsareuneccesary Its a society thing. Even if he doesn't care, others do. Employers do. Employers that could give more money. Its practical, especially if he's done the work and just needs to do some paperwork and formalities.
@@MrCreeper20keven discarding societal labels, PhD work is excellent place to get big sound corporation sponsorship for this amazing work. From corporate management perspective, it's helluva lot lucrative to get PhD work and university tie ups for investor confidence than let's say a youtuber. Number of patents and publication under the belt go a long way
@@MrCreeper20k Why should he care about whether employers do? I don't think this guy is making this trumpet for money either. I'm sorry mate, but following societal expectations is not the road to happiness for everyone
@@yesto9676to defend his point. Its not a road for sadness either. Nothing wrong with submitting it. He literally loses nothing(maybe time and other stuff). And the guy just gave a suggestion. And sorry mate but following the societal expectations is not inherently wrong. He can do it if he wants to. I dont know why you are so defensive.
This project is so refreshing! You rarely see a single person project that has so consistent improvements, it's a joy to witness!
O
O
Randy 😂
Hätte es nicht besser sagen können
Remember to donate to him so he is able to keep doing it!!!
Many community projects die because lack of funding!!!
As a physics student and as a hobbyist trumpet player I just want to add my name to the choir singing this video's praises. It's really impressive what you have managed to pull off here!
Yep. This fluid dynamics tech could be used in next gen synthesizers, such as a hypothetical future model of an electronic wind instrument. :)
I don’t think you realise how important this project will be to the audio synthesis world. Physical modelling taken to the extreme
he already didn't get it when he was making engine sim. if this was a vst plugin it would be winning prices for years
What is really good about it as opposed to a convolution model is that you can generate NEW instruments with this! and then make a convolution model out of them *at least*, discounting other possibilities of this running real-time
@@Beatsbasteln well as he says, the cpu / gpu demands are insane - plus an audio plugin has a lot of different demands. You’ve really got to be in that field to pull off that stuff well. But the technology and theory is very exciting. It would just take a lot of work to move it from a purely simulation based goal to a creative goal
@@NightMind0 exactly - it’s a step beyond the delay line based physical modelling that has come before. Full on fluid dynamic simulation would be an insane game changer from a technological point of view - but such an idea would live and die by its implementation and workflow.
@@NightMind0 convolution doesn't allow for the same flexibility
I think the reason you need lots of air to get good sound is for a very simple reason, the bell will enhance acoustic coupling and also resonate, making the instrument louder and sending that resonant acoustic energy back into the trumpet.
Think also because the actual embouchure opening is much smaller than the instrument bore. Probably only a few millimeters compared to the trumpet bore at like .468”. Sure it takes air to play, but it’s more about compressing the air with your lips.
@@prkassel Yes I agree, its not like your lips are flopping around in the mouthpiece like what the simulation depicted. As well, when you play the trumpet, youre trying to pressurize the air from the bottom of your lungs to your lips using your diaphram. Creating a large area of high pressure before a tight embouchure. Hope it helps Ange, fantastic video.
The mouth isn't a brick wall either. Some amount of backpressure will cause some amount of air to slosh across the lips. I think this video shows the lips being modeled as a perfect pressure source. Making it somewhat softer might lead to a mouth resonance.
Ange out here literally revolutionizing the gaming/simulation industry just to make engine go brrrr. He's a godamn hero.
And Music industry... This could become a crazy realistic Trumpet VST
@@triplezgames3882Yes dude I'm a music producer and this waa literally my first though
Ok correction, apparently this is already a big thing in VST industry. Look up Audio Modelling SWAM... Those sound crazy
I know! This guy is cutting edge. Was not expecting something this complicated to come next. This is amazing!
physical emulation of music instruments has been developed since a very long time in the pro audio world, and contrary to this "experiment" they are ressource efficient, and sound same as actual instruments (which is the purpose). A lot of engineers have worked on the topic since a very long time with actual results used everyday by audio engineers, producers and musicians. You just ignore everything of this world, but it doesnt mean it doesnt exist, and it doesnt mean what you discovered today is any kind of achievement or "revolution", far from that.
I know this is really not the goal of this project, but as a music producer, I have to say, that this would be the most awesome vst plugin for physical sound modeling… especially with the possibility to create completely new instruments and tweak real world properties, this would be an entirely new way to sound design!!! Sooo cool
I'm not even a producer and I thought that as well. This would take a virtual orchestra to a whole new level.
That's already been a thing. Arturia Piano does that.
@@bengsynthmusic I know, but it’s kinda limited in its sound
@@franciscoferraz6788 A slightly enhanced version of "Animusic" ?
ua-cam.com/video/hyCIpKAIFyo/v-deo.html
The swam trumpet from audio modelling does that. I believe their approach to pgysical modelling synthesis is a even more sound oriented and it is super efficient and realistic
Trumpet player here. Your simulation is uncanny. The tension in the lips being higher reduces air volume required and increases air Velocity. The resonance of the horn provides orders of magnitude more amplification than you would expect. The density of the metal plays a role too. Lots of sympathetic resonance in real instruments.
Simulation inception would be necessary to capture all of these effects. The fluid simulation is but one part of it. It's probably the most important though which is why your results are so close.
Love it man. You should share this tech with music companies like Roland, Yamaha, Nord, etc. They could really benefit from your achievement, and their effects and signal manipulation, user control, and expression technology could make this sound 100% better. Like a quantum leap in woodwind and brass synthesis.
Patent it please. Open source it. Share it some how. We need this in synthesizers!
Share? Hell no, he spent a lot of time, effort and I bet fixing all the bugs decreased his lifespan by at least a few years. I looked up roland - 3k employees in 2013 and you say to share it with them? He will get 0 profit, credit or anything from that
He should certainly not share it with any company. Selling the rights for them to use it, maybe.
@@shauas4224 if you open source it you can open source it a way such that it cannot be used to generate profit, there are licenses for that. Or he could just BSD type beat licensing and say fuck it.
patenting it is admittedly pretty cringe, especially for home gamer stuff.
LGPL?
@@killingtimeitself yeah good luck finding out that company won't care a single bit about license and then good luck finding out that they used it since I highly doubt any of companies will open source their software as you say author should do. Don't get me wrong, I m also curious to dig in the code and see how this works but downside to that - companies will abuse and steal his creation
Trumpet player here. I’m surprised how good the trumpet in your model sounds. The characteristic sound definitely comes through. It sounds a lot better than many Synthesizers I heard.
Not _most_ synthesizers? If it _wasn't_ better than most wave-based synths, that would be disappointing. Is it better than the _good_ ones is the question, and comparing it to Samplemodeling and even Audio Modeling gives you the answer to said question.
I gotta say, as a kid who grew up playing the trumpet, this wasn't too bad! The sound, I mean. The engineering and creativity are off the damn scale.
As an engineer I would 100% pay for classes taught by you, very clear and to the point. Demonstrating a high level of dominion of material and amazing pedagogy skills. Keep it up! 🚀
Unfortunately my virtual trumpet skills are just as bad as my real life trumpet skills. Having a semi-realistic simulation didn't help matters but kind of made things worse.
Now that you bought a trumpet, you have the perfect excuse to start practicing a new hobby.
How did you comment on your video 5h before it released? Love your work btw
It's relatively convincing of a trumpet with a mute on the end. Any idea what might be causing that phenomenon?
I suspect that it might be the lack of the flaring in the bore, but this is an uneducated guess
Your (accidental?) interpretation of Jurassic Park theme at the end was just fine. No practice needed. New music channel has been born. 😅
In the overblown demo, I was amazed it recreated the "double buzz" effect, where a lower frequency is also heard. This usually happens when a tired player can no longer control their embouchure (lip stiffness and geometry) and unintended vibration modes are produced.
Also, I wonder (1) what would happen if you're fully simulating the valves and what happens when you only depress the valves half-way. Half-valving produces a very distinct tone and diminishes the volume. And (2) if you can slide the stiffness of the mouth to recreate a lip slur between harmonics.
As a musician by training, I can tell you that this trumpet simulation sounds precisely like a trumpet -- and before adding the reverb, it sounds precisely like a trumpet in a well-dampened practice room.
Awesome job mate!!!
Or outside on an open field or on a top of a soft hill top.😊
@@Petch85 In those scenarios you'd be hard pressed to find a location that won't create an echo, no?
@@gownerjones There will be some reflections from the ground, but if you have a soft ground, like grass, it will be minimum. To make an echo you will need a hard surface for the sound to bounce off (off). Where I live it is much easier to find a place with no echo that one with. But if you live in a mountainous area it might be harder. But you can try the top of the highest mountain or the peak with the larges distance to the nearest mountain.😀
Also you need a low wind speed to reduce the bagground noise and new fallen snow can also help you reduce any reverbe or echo. glhf.
@@Petch85 It sounds like you have a lot of experience doing this. Do you conduct an outdoor orchestra or something?
@@gownerjones hehe no. I just love physics, simulations, music (making, listening, editing, simulating) and hiking.
I've been playing the trumpet for about 8/9 years now, it's absolutely amazing how close this sounds to a beginner trumpet player. The coolest part to me was when you did the model of increasing airspeed, the way that the pitch increases, then decreases and distorts is EXACTLY how an actual trumpet plays when you do the same thing.
when this game releases, saying, "this has great attention to detail!" is going to be a huge understatement
the meaning of "great" isnt so definitive, so maybe it kinda isnt? a baby finger painting for the first time is great, but so is winning the lottery.
obligatory 🤓 here
@@genralty ackshually a baby's first finger painting isn't anything great, as it's not an art and even among baby's firsts it's not that important as walking or a handprint 🤓
as a first chair trumpeter, i can say you may be getting the weird intonation from lack of resistance based on the valve position. no 1 valve generates sufficient resistance but there are a few combinations that do generate resistance at certain octaves due to the resonance. to someone without experience playing it can be hard to feel it, its subtle but very significant. all in all, very cool to see this done.
I think the biggest reason there's not much resistance is that the model doesn't have a mouthpiece yet. We've basically got a really narrow didjeridu so far. (I think the model is doing a great job giving us what it would actually sound like to play a tube that's shaped like that!) Once we get a mouthpiece that's the right shape, it's going to give us *way* more realistic resistance.
@@thomasscheevel5779 i feel really stupid and am kicking myself in the ass now. i totally neglected the fact that different mouthpieces have an effect as well. good spot thomas
@@puddlejumper6999 tbf, i probably wouldn't have recognized the sound if i wasn't the kind of person who's tried to play brass (a tuba, for me) with no mouthpiece before. so when he went in the video about how surprised he was about how much pressure it took to get a sound, it just clicked - yep, that's how it feels and sounds to play a brass instrument with just your face right on the leadpipe
@@thomasscheevel5779 same here XD and such facts oh such good times
Valve horn player here. Great job on trying to capture this complex procedure!
Couple things I know from playing a brass instrument:
- The most important part is the lips that distincts between a good player and beginner. I can make my instrument sound like your simulation, just by adjusting my lips.
- Air Pressure and Lipstiffness increases with higher notes. Playing loud and deep is as hard as playing quiet and high. But one can compensate this by having a more accurate and more stiff/loose lip
- The way I let my instrument sound different with the same pitch is by increasing or decreasing the volume of my mouth. Bigger Volume results in a more round tone that is often nice to hear. But sometimes in certain musical genres you want this duck like quacking and therefore you decrease this volume (by adjusting the tounge mostly). This also is dependent on the tone - higher tones are generally played with less volume while lower tones require more volume for the same quality of sound.
Do with that information what you will, just some insights from my end.
Thanks for sharing your knowledge! It would be cool to refine the lip model using information from real trumpet players since most of the parameters are purely physical/scientific
As a trombone player, I agree with this comment. I think to really improve the sound you would need to simulate the mouth cavity (up to the whole body) as it acts as a resonance box.
It would be fun to play around with a trombone sim as it is simpler than a trumpet (sim-wise, there are no valves) and it would illustrate how the pitch changes when the length of the tube changes.
Awesome work, I really enjoy what you're doing !!
@@AngeTheGreat You may try to simulate buzzing just on the mouthpiece. Some consider it a bad practice because the embouchure for buzzing is not exactly the same as for playing the instrument but without the brass tube's resonance you can freely change the pitch so might be able to see better what lip configuration is in tune with each tone.
@@AngeTheGreat time to build the worlds first lip simulator
@@AngeTheGreatDr. David Wilken did his dissertation on brass embouchure, and he has some interesting videos showing brass players playing into transparent mouthpieces.
One of the more interesting things I learned from that is that not only are the lips not centered vertically on the mouthpiece, but a brass player will adjust the mouthpiece position depending on range.
One reason I love these simulations is because they push the limits of computation. With web development dominating mainstream, it’s refreshing to see non-trivial C++ projects.
I used to be a trumpet player. I am very impressed with this. Especially the over-blow bit. The simulation was able to pretty accurately predict what actually happens when you blat.
The mouth simulation would need a lot of work to be accurate. There is much more to embouchure that would need to be considered. You'd probably have to model a whole head. Its all very impressive!
Agreed to at least the whole mouth, not sure about the whole head. Like, how essential is your skull when buzzing, y'know? I think simulating all the way to the back of the throat would make the most sense, because of how important the tongue is in shaping your tone.
Still super impressive!
@@angelicbeast7793 i think will's point was once you have put in the effort to model the mouth, you might as well just simulate the synapses in the brain that decide which note to play :DDD
@@angelicbeast7793need the nose for that circular breathing haha
You'll need to model the lips, mouth including tongue and teeth, the larynx, lungs, and the diaphragm... Good luck!
This dude is slowly building the most realistic virtual world ever
Im a trumpet player and not an engineer. The thing i like the most in this video is that, although the trumpet isn't tooned, the fist demonstration of the trumpet's notes is as accurate as it gets. Although im not a luthier, i can definitly say the chromatic scale is exactly as its portrayed: 000, 010, 100, 110, 011, 101, 111.
Descending, not ascending.
Luthier's make sting instruments... you are thinking of a brass instrument maker.
at first i thought it'd be like counting in binary; 000, 001, 010, 011, 100, 101, 110, 111.
I'm a trumpet player too, and loved that this video is a defiantly on point for how a trumpet produces sound. I think some of the parameters that are missing to be able to enhance the sound go something like this..... 1. lips, tension. There's several ways to achieve this, basically we tighten up the cheek muscles to make the lips more tense. In addition, we can change the pressure of the mouthpiece on the lips to achieve the same. 2. Instrument construction. The tone is effected by the mixtures of metals that make up the instrument. I'm talking about he zinc/copper ratio on the brass, which effects the "brightness" of the sound, also whether the instrument is silver plated, or lacquered. Again this finish can have an effect on the tone. 3. And here's a whole new can of worms, what's happening behind the lips. The tongue is important as it's placement will dramatically change the sound, as well how we open our airways in the throat whilst playing. Puffing out the cheeks. Also whether we are breathing with the upper part of our lungs or (as is better) the lower part using our diaphragm to apply the pressure that creates the volume. I believe that we tend to use higher pressure than you'd think to produce the vibration at the lips, but the opening is very small, so that volume of air passing is relatively low. It's common for a trumpet player to be able to sustain a note for about 60 seconds with one breath. This should be able to give you the idea of the amount of air that can be delivered with normal lung capacity.
@@Goliath83that would make a terrible ergonomic. Imagine having to use your ring finger so much.
@@Goliath83It’s counting in binary, except the digits are shuffled, because the pipes attached to the valve aren’t in ascending order of length. Instead of being worth 4, 2, and 1 semitones, they’re worth 2, 1, and 4 in order from closest to farthest.
As an audio engineer who sometimes dabbles in recreative game development, loves physics and used to play horn, this couldn't be more up my alley. I absolutely adore your recent projects, Ange. Much love and massive props for all the progress you have made in such a short time frame. You rock my dude :)
I love how versatile the fluid simulation turned out to be! Have you tried simulating the sound of a turbine? (or compressor?)
Or a death whistle :)
@@revimfadli4666that would actually be really cool
Interesting idea
Steam engine simulator?
It's nicely versatile & good tbqh
Its hard for me to describe how captivating it is, to listen to a truly passionate and intelligent person explain something they've made. Its just SO nice.
I'm always so blown away at how one person can have such a good combined understanding of engineering, acoustics and coding 😮
And video creation! That's not easy either
@@timo1294 And 3D modelling which is a career itself.
Totally, this is what brilliance looks like
Not to mention 3D modeling! Is there anything this man can't do?
I mean, if someone says they do computational fluid dynamics, "engineering, acoustics, and coding" is basically what it says on the tin. CFD is vital to all sorts of engineering work, and coding is its own sort of engineering. Acoustics basically falls out of the equations that you use to model the behavior of air. And any field that has "computational" in the name is going to involve coding.
Brasswind player, mechanic, and "armchair engineer" here.
A huge thing about trumpet playing is that the lips form a seal on the mouthpiece. I couldn't tell if what was shown is what was being simulated, but it appears the lip model wasn't fully pressed against the mouthpiece. This would make a lot of sense why it required so much more air to create a sound, as much of the air pressure can escape around the sides.
Playing trumpet well includes having a lot of air pressure behind the lips, as well as carefully measuring out the correct frequency for the lips to move to match the note. Air pressure is achieved by tightening the diaphragm to force the air out as hard as possible. Face muscles control the tightness of the lips. The process of playing is an intricate dance of maintaining air pressure, matching lip frequency to the desired note, and using the correct valves.
Also, the bowl inside the mouthpiece affects tone as it changes the shape of the lips and how far apart they can move. When the lips have a shorter distance to travel, they can trill faster, creating a higher frequency. Novice trumpet players can reach higher notes by pushing the trumpet into their lips, forcing the lips further into the mouthpiece to make it easier. Skilled trumpet players have developed their facial muscles enough to tighten their lips without having to do this. Ideally, you don't want to strain your neck by putting any more pressure than is necessary. Some trumpet players make vibrato by repeatedly pressing the instrument into their lips to vary the tone and pitch at the same time. Also, different mouthpieces have deeper or shallower bowls to make it either have a more "full" sound or easier to play higher.
Hope this helps. I've enjoyed your engine sim adventure since the beginning. Seeing you make something from it that touches upon more than one of my interests makes me very happy.
Another tidbit, intonation is also controlled by the first and third valve slides. You would have to look up which notes need it, as I've forgotten at this point. I play the baritone mostly and haven't seriously touched a trumpet in over a decade.
It’s so crazy to me that like 8 or so years ago I had an idea that modeling instruments using physics would either happen or I would have to learn to do it myself. I would absolutely love to see a version of this project in vst format later down the road. As a producer this excites me so much!!! Keep up the great work!
Well, physical modelling is all the rage now, look at Plasmonic or SWAM. Swam is especially impressive at modelling the wind instruments, i wish it would be cheaper 😢
@@4CloudySkylaughs in their trumpets and brass
@@snesmocha That's what Samplemodeling is for. Listen to some good demos of it, i.e. Sam Paryss's Star Wars mockup using it or The Trumpet v.3's demo medley.
With like two more layers of simulation inception this could get scary close. Aside from the bell, material type also plays a large factor in sound resonance. This video is truly awesome, you did a great job.
Experienced brass musician here. It sounds like an early beginner trumpet player who is trying to find out the best way to make a sound! I'm mostly impressed at how accurately the valve movements distort the tones between presses, which sound *very* accurate. From the sounds of it overall, the tone of the trumpet could mostly be fixed by making adjustments to the player's embouchure (the mouth and its positioning on the mouthpiece). But I can say that I've heard these kinds of tones from the beginning trumpet player.
This is amazing - as a physics graduate working on particle simulations, and a trumpet player of 10 years, its amazing to see those two worlds collide. The fact that you can even get a roughly-correct chromatic scale and multiple harmonics working is extremely impressive. To me, it sounds exactly like a trumpet playing into a cushion, or a really closed mute. I think simulating the bell flare will definitely help (I've experienced many dented flares). For even better tuning, the third valve tube would need to be extended slightly in the lower harmonics.
Use the system to simulate the call of the parasaurolophus. It has a large resonating chamber in its crest that drastically changed the sound they make.
As a former trumpet and horn player and current software developer, BRAVO sir. The way the note breaks apart when the embouchure is too flat for the instrument on that note is EXACTLY how it sounds in real life! And the way the pitch transitions between pitches on input changes was spot on too
Exciting progress! Quite neat how flexible the ES3D fluid simulation is
My theory was that if it could simulate a trumpet sound, it could probably simulate an engine sound. I guess we'll find out soon enough
@@AngeTheGreatFingers crossed! I'll be looking forward to seeing it in action :-)
@@AngeTheGreatI’ve got my 22 bucks ready and waiting for ES3D. Keep up the good work!
@@tescoshortageis it gonna be paid?
@@mr.mirror1213why not? Its an extremely well-put-together project
I've been a trumpet player for about 10 years, and indeed, the lip parameters are extremely important to hit the right tone. Trumpet players even anticipate their lip setting and sometimes also pull the instrument tighter to their lips to prepare for the next note, at least to minimise the air gap between each note. If not well prepared, it even occurs to players that only air comes out, completely missing the note. As found in the simulation, the amount of pressure lips have endure on a Trumpet is immense, professional players need multiple years of lip training and lip hardening to cover an acceptable tone range. It happened to me to have shattered lips close to bleeding after a concert. The simulation exactly sounds like someone with no experience in lip setting, which is the case here. Lip tension is even changing along the duration of the note, call it expressions, phrasing of the notes or modulation, like vibrato, staccato, pitch bend and more... Adding randomness in lip tension is an easy way to improve realism, but simulating expression with lip tention needs musical feeling, which is a learning process that probably only AI would be able to implement.
Came here to say something similar about the lips. The tweaking of the params to get the 'clarinet' sound definitely made that pop out.
I Imagine more work on the flare will help, didn't hear any talk about the "material", i.e. certainly a wooden trumpet would sound much different than a brass one--I don't know how that's accounted for here... maybe covered in a different video, but yeah, so much is in the lips and I wonder if the original model is just inadequate for the needs here.
Buzzing with even just your lips, you can quickly start to understand how much that contributes to the sound.
I imagine he could also simplify things and just focus on perfecting buzzing with just a mouthpiece and then plug in the rest of the horn to a massive improvement.
Hope he decides to take at least one trumpet lesson as I feel like that will be an epiphany on how important that part is.
I've played the trumpet for well over 10 years throughout middle and high school. I'm also self taught in C/C++ and 3D Graphics / Game Engine programming using DirectX, OpenGL and Vulkan. I'm also adept with mathematics and physics. To see a video of a physics simulation that uses C++ and Vulkan to project the fluid dynamics of a trumpet is really cool. This video is very informative, creative and quite awesome. Job well done!
this video seems to have brought out all the C++ trumpet players and I'm all for it
I was already thinking that BeamNG devs were the best devs out there, but the fact that they're helping you on your projects is just awesome
Can't wait to test a 27-cylinder engine with real headers soon!
I am a trumpet maker, and this is really an incredible simulation.
There is only one other publicly available simulation software available for brass instrument makers.
This could be a hugely valuable resource for people wanting to fine tune the shape of the air column before making tooling, or to correct things in R&D.
first thought on the first sound was "hm, straight cut trumpet sounds like Sax with a very, very wet mouthpiece", i guess it's more about the lip params and I wish someone will take it upon themselves to jump through the hoops making this the most versatile physical modeling VST-Instrument ever.
Ange, you are breaking new ground in software development, with a perseverance and consistency that is nothing short of legendary.
Former french horn player here. The sound character is amazingly close to the real thing. This is going to be huge once out. Good luck man
this audio is no where near what modern horn libraries are currently. It is one aspect of a useful addition to what we have.
@@morbidmanmusicI think physical modelling will surpass sample libraries in the future
This is so cool! It goes quickly from "huh, that's a neat simulation" to "wow, that visualization looks incredible!".
I also appreciate the non-fearmongering VPN ad.
Played trumpet in junior high and high school, and I can say with confidence, trumpet is among the easier instruments to learn, and if you up and bought one, I'd highly encourage trying to learn it.
Fart simulation when?
I don't know if the world is ready for such technology
How does this still have 2 likes
I love that it takes multiple research papers to understand the effects of your lips on sound, and musicians are just like "it's right when it sounds good"
I really like the particle visualization for fluid flow! It reminds me of… an infomercial for a vacuum cleaner.
I think including the ability to create extruded 2D shapes in engine simulator would be a good feature, because that would allow simulating different muffler designs.
Trumpet player here, been playing for about 13 years and just finished my degree in games development. WOW!! Christmas has come early! I find it super interesting that despite modelling it off a real Bb 'key' trumpet (i.e, when you play a 'C', it sounds like a 'concert pitch' Bb), it sounds like it's tuned to Ab. It was a super minor section in the demo, but when you demonstrated the crescendo (going from low volume to high volume), I was blown away by how it accurately simulates the sound 'opening up' - something that trumpet manufacturers can play with to change the character of each instrument. This is awesome!
This is absolutely incredible from start to finish. The data visualization is a master class in visual representation, and the project itself is just mind boggling.
The one person who deserves to call themselves great on youtube x)
do you realize how much of a market there is for tools like this? sample libraries of physical instruments are costly to make and often inflexible to use. with a ui designed for musicipans, the ability to physically model wind instruments accurately and in real time would be a gamechanger!
100%! I would love to see what MuseScore hooked up to this as simulation could create!
Incredibly fascinating demo and also a very clear visualization of what exactly is going on physically in a wind instrument and why! I think one of the hardest things about the learning curve of a wind instrument in general is understanding and controlling all those various parameters in real time while only being able to abstractly grasp what the physics behind the instrument actually are doing, so this is amazing to see! Also I definitely laughed out loud when you started playing the Jurassic Park theme haha
make a realistic pipe organ, replete with all the stops etc. That would make not only a great steam game but also fill a market void of physically based music synthesis for musicians
Oh my god, this is an amazing project. As a passive fluis sim enthusiast, you have just reignited my passion for this field thank you! 1 tip, i think 1 of the reasons why you needed so much pressure to make a noise is the mouth piece. Not sure if the modeled mouthpiece is the same as the simulated one, but the concave section is too small. From what I remember from playing a trumpet in middle school is that the domed section in the mouthpiece was much deeper, and the walls where thinner. Hope that makes sense haha, can't wait to see the next update of this!
You're one of a kind man, keep up the insane work!
The resonance of brass as a metal is vital to the reduced air flow necessary to produce sound because every inch of the tube is added shaking of the instrument which aids in the increased efficiency of kinetic energy’s transformation into sound.
This will also aid in your high harmonics necessary to form timbre
you could totally charge money to have a sim like this as a very accurate VST for music producers
18:42 thank you for making this accurate representation of a 350z flyby
This video devoured my internet bandwidth like a snack... What truly blows me away about you, and this resonates with me (haha, resonates), is that you give your 110% in everything you do. In this instance, not only is the video content top-notch, but the production quality is off the charts. I mean, even your compression settings seem like you've mastered them to perfection just to deliver the most incredible results. Simply mind-blowing.
This is clearly a big step for mankind in reproducing VQ exhaust noises as accurately as possible
I find this project so inspiring, and I can't help but feel so much pride watching this! You've done so well!
Edit:
like seriously, as someone hugely interested in audio, music, physics, engineering, computer science, statistics, and many forms of visual art: this project meshes my biggest interests together in a way I'd never have expected. I cannot wait to see what's to come!
As a trumpet player, and a CSE student in a computer graphics class, this is amazing! Idk if it would help, but if an IRL trumpet student is overblowing, it often means they have to tighten up their lips a little. When you have air blown way too hard in the simulation, you can hear the pitch change as the pressure changes the forces on the lips. As a trumpet player, my best sound is made by balancing the tension in my lips with the air pressure. A way to measure if you hit a reasonable balance is to study the amplitude of the sound. When you have a good tension for a given pressure, the horn will resonate better and become louder. This is not necessarily the sound all musicians aim for, but it is a discrete and measurable goal that's good for beginners.
I noticed that the simulated trumpet's tuning slides are all pushed in all the way. Pulling out the main tuning slide and the 3rd valve slide (at the same time as the valve is pressed) should improve the overall sound and make the intervals between notes more even. Another thing to consider would be to add tonguing, where the tongue is used as a valve to quickly start each new note. This would eliminate some of the out-of-tune movement of pitch when the valves and slides move.
I really love this project you've made! If you have any questions about real trumpets, feel free to hit me up.
Would you consider doing a similar simulation for the human vocal tract? That would be a great educational tool. There are no truly good online resources that explain how and where the different formants of speech are created, the explanations are quite rudimentary. E.g. visualize how the same formant (in the output spectrum) may originate from two alternative configurations of the tract, due to the tounge position.
That's a good idea
I don't think this type of simulation is capable of coming close to simulating human voices. It's useful for systems that consist of connected cylinders of air, not something with complex geometry. I imagine it would be very hard to physically simulate a realistic / convincing sounding voice.
@@aspzxIt indeed can, albeit in a simplified form: check out “Pink trombone”, an interactive vocal tract simulator that runs in the browser.
I'm not sure how it compares to what Ange is working on; but you might wanna take a look at Neil Thapen's Pink Trombone
if this could be possible, then it might be possible to generate voices from it, I imagine it would be impossible to generate the voices that is the pressure waves and other mouth expressions manually to output voice, we may train a neural network to generate the needed air pressure and toung and mouth movement to generate a voice, This could be the most realistic TEXT to SPEECH generator of all time. have goosebumps just thinking about it.
This project is insane to me in the best way possible. As a former awful trumpet player, the off putting part about the first demo was how clean the sound was with a stable air supply and no reverb. Once some reverb was added, it just sounds like a trumpet with some tuning problems. Can't wait for this to release and see what people can do with it.
The detail in the attacks of those notes is insane. If someone works out a comfortable mapping for pitch+velocity, this would be an incredibly useful VST synthesizer on its own :D
Keep it up!
Envelope generators and such buzz through the mind. :) MIDI control?
@@YCbCr This has got to be one of the most generalized physically modelled synthesizers right? Is there even anything we can download right now that would be able to simulate arbitrary horns and intonation at this level?
@@DuckPerc I don't know, is there? The excitement is VERY real anyway!
Truly unbelievable the fidelity you can get with a simulation, I cant beleive how you can get those note transitions without nuances of the valve shapes, and the sound quality with that simple lip simulation, amazing.
Oh my god. Oh my god. This is AWESOME! First of all, this video showed up in my feed while i was in a practice room practicing trumpet. Second, when this is finished, PLEASE PLEASE PLEASE PLEASE PLEASE make a VST3, AU, and AAX plugin for this if you can.
Also! As a trumpet player and an audio engineering student I have a few observations, questions, and suggestions which you may or may not have already taken into consideration:
It sounds like a french horn, or other low brass instrument. Making it sound more like a trumpet (i.e. generally brighter) is going to come down to the resonation of the metal in the bell. I couldn't tell if you were simulating the actual metal itself resonating, but the temperature, thickness, material, and plating of the metal is going to be very important to the sound. Heres a few parameters that I think could improve the accuracy of the simulation (and are also interesting to tweak): metal materials (i.e brass, copper), plating (gold, silver, none), presence of lacquer, Bore size, Bell size, Bell thickness, Different mouthpieces(??), just to name a few.
I noticed that your tuning slide wasn't adjusted at all. Unless it is VERY cold in your room, you are going to need to put it out a little to get proper intonation. Also, when playing middle D (Valves 1&3) the third slide will need to be extended slightly, and fully extended when playing middle C# (123) for proper intonation otherwise they will be very sharp.
On the topic of the bore and the bell, these things have a significant effect on the brightness and tone. I play a Bach Stradivarius Model 25L, meaning it has a relatively small bell (brightens the tone and increases projection) and a large bore increases the richness but requires significantly more air. I've played quite a few other trumpets and the difference in intonation and playability is massive.
Mouthpiece - not sure if you did research into this. Theres a bunch of different parameters for this (cup size, depth, shape, rim size, thickness, etc.) and it affects both intonation and playability. Each mouthpiece has its pros and cons in terms of range, tone, and playability. You can buy a screamer mouthpiece for playing lead and screech all day long but it wont be very rich and warm. Im not sure if any of that applies tho, as theoretically the simulated lips aren't subject to pain or fatigue.
Again, if you aren't already simulating resonating metal id say do that first, as it has an enormous impact on tone.
I know thats a lot of things to simulate, but I think it would make a significant improvement. Also, I think the most impressive thing here is what you've done already. It sounds like a brass instrument, and it does tuba pretty convincingly. You just need a few finishing touches (and high end).
It would be super cool to see this released to the public, but regardless I'm dropping a sub on patreon right after I finish typing this out so I can start playing around.
The ramifications for this in the production community are insane. Hyperrealistic brass instruments right in your daw???? Theoretically, this simulation could be used to recreate room reverbs, without the need for an Impulse Response and convolution reverb. I wonder if there is enough performance optimization available to make a sim that large work effectively in real time.
Anyways, this is super awesome, and bravo for making something real that I've only ever dreamed about. I can't wait to see how you improve this, and whatever other projects you have in store.
You're absolutely correct in describing how a simulation actually does a calculation, but there's a much simpler way of describing it (FEA incoming!).
Some terminology:
- Node: a single point representing a boundary
- Element: a line connecting two nodes together which is represented by a characteristic equation.
- Characteristic Equation: the equation that models what we expect to happen based some input (i.e. F = -k•x, ∆Q/∆t = k•A•∆T)
Finite Element Analysis is the method by which all simulations solve complex problems that either have dynamic states, or shapes which are hard to represent mathematically. Depending on the problem you want to solve you can either solve a problem in 1D (a line), 2D (a plane), or 3D ( a volume).
Typically, most problems are homogeneous (meaning the characteristic equation is the same throughout the entire volume) and we can get away with modeling things in 1D. In such cases, we can break up the problem into some number of nodes (n), and elements connecting the nodes (n-1). Each node has input and output variables and serve as a direct connection between elements, meaning that we can take the reaction from one element to accurately represente the reaction of the following element.
An easy way to visualize what this looks like is by thinking about lines at a checkout. When the person at the front of the line is done ordering they move out of the way and the rest of the people in line move forward. Each person represents a node containing the current state, which in this example is position in line from the cashier.
What you mentioned about elements only interacting with adjacent elements is a constraint on all FEA models in which we need the output from the characteristic equation to serve as the input for the next element in line.
An easier way of thinking of this is by imagining imagining a highway with 4 lanes (2 in each direction). If 2 cars are in 2 separate lanes then they don't interact with each other. However, if 1 car is in the same lane as another car, then said car is limited in how fast it can go by the car in front.
Hopefully this explanation helped. FEA is a difficult subject to understand for anyone who isn't in the STEM field, and throwing in fluid mechanics takes the complexity to a whole new level (source: I have my bachelor's in engineering with a focus on heat transfer and fluid mechanics).
Fantastic video as always, keep up the great work. I can’t wait to see what’s to come!
I will and thanks for the support 🙏
Former trumpet player, now engineer who had to give up dreams of programming interesting side projects. I'm floored by what you've done here! Absolutely amazing! I was just recommended your video today, but will now follow with interest!
having heard a good deal of trumpet-including music, both live and online, i have got to say that this sounds pretty damn good for a fluid simulation. there were some small parts where the audio became noticably "rough" in a way that i only really know from computer-generated audio, but during some other parts it sounded like an authentic trumpet being played by someone just getting used to playing the instrument. i've got to say, i am very, very impressed by your work and i'm certain that the quality of your other computer simulations can only go up from here.
Dude I swear the entire time I was sat here expecting this to be a long drawn out VQ engine joke, but alas, you really did deliver!
This is seriously impressive stuff, you can definitely hear the "life like" tone in the sound, proving the simulation is astonishingly accurate.
During your research did you use the book Computational Fluid Dynamics with Applications by John D. Anderson? It uses a finite difference approach which can be easier to program. The other very good book that maybe you used is Introduction to Computational Fluid Dynamics - The Finite Volume Method by Versteeg. This book really gets into the details of the solvers, PISO and SIMPLE.
The way it blows sharp and gets a splatty intonation when you up the pressure is incredibly accurate. Great work
First of all, this is amazing and thank you for showing us your work. I understand this is a giant technical feat and worthy of praise and I agree with everything you said. Second of all, the moment the trumpet started playing, I burst out cry laughing and haven't stopped since
You get taught in most science classes to show your work, this is just a side effect. ;p
I want to say that its impressive how you send the information to the viewer. I dont know most of the things you say but somehow I understand everything.
I was thinking about how you could use a fluid simulation to simulate a subwoofer or smaller speakers in a range of different enclosures with different sizes and such, and how different length ports and transmission lines change the sound and group delay and such, it would be a cool video, I don't know how hardware intensive it would be but it would be cool.
Finally someone who actually explained what VPS can't do. A lot of UA-camrs just say it keeps you safe from threats on the internet
You could make it as a VST so it can be played with midi inside a DAW or other programs. There aren't many pysical trumpet simulation so it could be interesting
Because of the embouchure needing to be emulated, this isn't so easy to replicate or achieve using a MIDI keyboard, but I do agree that this would be great. Solo trumpet VSTs and sound fonts tend to sound less than ideal for a quality recording, so something like this would be incredible.
An EWI would be the way to go. I don't think they do anything for embouchure transcription but it's still closer than a keyboard or a mouse. 😊
@@JXter_ DAW would let you adjust embouchure with things like, automation, without having to manually play it on a midi keyboard
The sound breaking up at 18:48 is spot on for the state the simulation is in, I had that happen quite a few times when learning a decade ago and to hear a simulator replicate it almost perfectly is amazing
Dude you’re such a role model for me. I’ve always had ambitions to write a ton of simulations myself, but you’re actually doing it, fast and in amazing detail. Do you take apprentices? 😁
It'd be really cool to hook this up to a reinforcement learning agent to have it learn how to play or tune an instrument. Level of detail is amazing. Really tickling my physical simulation buttons.
The mathematics reminds me very much on my recently finished university project about finite difference methods for electrodynamics (Since they are also hyperbolic equations).
You said something about further sources about the math behind your simulation is available in the video description. But I can’t find any link.
Edit: Links are now available. Thanks!
I forgot to add them before, but they should be there now. I have some additional links that I'll post later as well, I'm just trying to collect them all lol
18:40 The way the tone spatters like a junior high early learner is absolutely hilarious!!! Really LOL!!!
That trumpet sounds great! Have you considered turning this into a VST instrument to sell to people using music software / digital audio workstations?
I’d pay for it
The quality and level of control isn't nearly good enough for this to be a VST worth buying. Well... not yet anyway.
@@Erin-ks4jp I didn't mean as-is, I meant with more work ofc! :D Considering this was made mostly for a fun quick video, and the video mentions many things that need to be fixed to improve it, I think the potential is there though, if it was focused on.
I played the trumpet for almost 10 years and I can say that this is actually amazing to me. Despite its flaws it is the most accurate sounding audio I've heard a trumpet make artificially. Everything I heard sounds very familiar to what I'm used to hearing. I could almost feel the different pitches as if I was playing it, definitely a wierd feeling 😂 if there are any other players here, you might know what I mean. Now that i think of it, the sounds the simulation makes reminds me of what it feels like to play with a earbud in one of your ears. You can feel those same vibrations through the audio of the video.
It's amazing how you manage to get that running on a normal PC. Every time I try to simulate something myself from scratch I use way too much memory lmao
have you tried learning more maths? like.. a LOT more?
@@calvinkohl6220Learning the math is one thing, but to optimize it to make the code run faster is a different can of worm, especially if aiming for accuracy.
Using C++ is much more efficient the memory side of things especially compared to something like Python or Java, without even mentioning the cpu side
@@calvinkohl6220would more maths translate into better CPU and/or cache optimisation?
Or he could be using Rust to have an easier time
crazy how the simulation gets the sound of overblowing a trumpet perfectly
I see someone creating VSTs based on this.
I was about to say, people would pay an arm and a leg for a VST of this. I think VSTs can't be multi-threaded though, so you'd need to run it as a seperate program and record and export the audio.
This could definitely be great for music production purposes as its honestly the most realistic trumpet simulation I've heard. Idk if it'd be possible but integration for use with DAWs would take this to another level.
This project is so cool! From what I’ve read about mouth pieces, they actually function as Heimholtz resonators, with another pressure boundary (or whatever its called, like the bell) at the base of the mouthpiece cup. This junction is the interior point at which the vibrating column of air reflects. In real trumpets, the size and taper of the backbore - as well as whatever gap there is between the end of the mouthpiece and the beginning of the leadpipe - has a large effect on tone and slotting (how easily the notes pop out when you play). And the size and shape of the cup also affects whats going on. I know thats a lot of extra stuff to model but would you consider it?
Other fun aspects of brass instruments which I’ve wanted to see modeled for a very long time are curvature and bracing. The curves in the pipes of brass instruments seem to function as longer pipes than the straight sections. Brace placement/rigid points along the tube also affects slotting and intonation.
I know you’re trying to keep things simple and only doing a straight tube right now but these are some things I would absolutely love to see. Amazing job!
As a clarinet player (and hobbyist musician) i found this video very pleasing. It sounds REALLY good.
I can't escape those Euler equations. I am in compressible aerodynamics and you jumpscared me with that partial differential equation. Awesome video! I love everything that you do. It is also nice to see your assumptions within this project. Cool engineering.
Trumpet player here, the tone sounded better than most middle schoolers and some on high school. This could be a great interactive teaching tool!
Another important variable to consider happens before the lips, to get a nice wide tone, we open our mouths while keeping our tongues low. The smaller the mouth opening, the more we squish the sound.
As a trumpet player, every time you rested the trumpet on the second valve slide, I flinched
Good thing you didn't see what my hardened steel calipers did to it
@@AngeTheGreat I didn't watch the full vid yet but yikes
Don't worry, it's not in the video lol
@@AngeTheGreat ok then lol
Y'all those scammers really need to stop
This is the best sounding CAD-simulation of a Trumpet I have ever heard.
To echo what a lot of other trumpet players are saying here, it really speaks to the accuracy and sophistication of this model to say that you have accurately recreated a beginner trumpet player. You said yourself that the embouchure is difficult to model. Well, a beginner trumpet player has little concept of embouchure and therefore sounds almost exactly like your trumpet model. You have perfectly modeled the trumpet, but not the player (which is beyond the scope of this project). Anyone who has learned the trumpet or been in a middle/high school band can vouch for the accuracy of this model!
I would give anything for university lectures for every course, or even just 1 of my courses, with this kind of clarity & production quality. Crazy part is, watching this & learning is $0, whereas university cost me $31k/yr.
Awesome video Ange, Thank you for your massive contributions to the community!
I have undiagnosed epilepsy and this entire video was triggering me really bad. 😅 (I don't seem to have normal photosensitive epilepsy and it seems to be a time dialation thing, slowing down and speeding up things) I had to keep looking away for most of the video, but I really enjoyed listening to this. Your videos are super easy to understand and really helps give insight on how fluid dynamics work in the motor. I have been doing a lot of research while building some headers for my engine, and watching your videos have helped a lot to visualize and tie everything I am learning together to understand how everything works and affects things. Thanks for the heads up in the title.
Do you have any ideas on how I could fix this? I don't really know too much about this subject and it's hard for me to judge what the problem is and how it can be resolved
@angethegreat i mean if you want to correctly visualize sound waves, i dont think there is any fixing it. For me, its that constant speeding up and slowing down that gets me, which is just going to happen as you increase frequency visually. A possible fix for this video might be not using a black background, but visually, the black background looks great. I have not had any triggers whyle playing with your engine simulator though for what it is worth
@angethegreat from the papers ive read about photo-epilepsy, it seems high contrast-lowfrequency(