Generate Images Faster with Stable Diffusion and RTX
Вставка
- Опубліковано 28 вер 2024
- ComfyUI offers a streamlined interface for Stable Diffusion, accelerated on NVIDIA RTX GPUs with NVIDIA TensorRT. Fast image generation combined with one-step RTX-enhanced ControlNets unlocks unprecedented workflow control.
RTX GPUs not only speed up image generation but offer access to the newest and broadest range of AI features.
Learn more about AI on RTX 👉 www.nvidia.com...
Subscribe to @NVIDIAGeForce today!
GeForce Twitch: / nvidia
GeForce Instagram: / nvidiageforce
GeForce TikTok: / nvidiageforce
GeForce X: / nvidiageforce
GeForce Facebook: / nvidiageforce
GeForce Community Portal: www.nvidia.com...
#StableDiffusion
#NVIDIATensorRT
#RTXOn
where is the comfy workflow ?
Will Smith ate it. :(
its funny to think that all of this amazing technology that demonstrate the advancement of the human race is shown to "lets make a custom sticker packm oh yeah i wanna be a super hero"
incredible, bravissimo, again, more
Es normal, los usuarios somos personas comunes con necesidades comunes, el hecho de que una tecnología sea maravillosa no significa que solo deba ser usada para hacer cosas maravillosas.
The people that force ai down the throats of users can't came up with anything really useful to show off as an example. Lucca Rossi from Lenovo at Snapdragon Computex 2024 Keynote shows how to generate pictures of penguins...
everybody with half a brain sees that AI will improve considering how fresh the tech is and its already nuts
AI in the year 3516, random space company travelling the universe "3d print a medium sized planet at these xyz coordinates, procedural generated vegetation in steampunk style, cartoon clouds, rainbow water" 🌏🌈 Printing Prompt ... Please Wait ...
Now waiting for the "Video" version...
Okie apple fanboy 🤓
Agreed, I waiting video version, but Nvidia AI not available for upscale image from lower resolution to 4k or 8K. Or like Krea AI and Bing AI are awesome
bro just stitch the images together
Does the software do what I tell it, or is it censored into the ground because of excessive safetyism? Most of the AI tools being released today are more likely to deliver a schoolmarm lecture than the user's requested result, even on utterly innocuous requests.
It does what you tell it, it's fully local and open source.
So no people draw comics anymore? AI is very convenient.
does whatever you want (depending on which model you use)
@@DORAEMON-bw8jk Considering how much stuff in comics today is traced over existing art to save time and push work over to interns, this isn't a huge departure.
It's not the software that's limiting you, its the model that you're using. Models produce what they are trained to produce. If you feed them NSFW, you can get NSFW out of them, simple as that. There are dozens of new models released every week, and a good portion of the can do whatever you throw at them.
since amd release amuse 2.0 fo sd , what is the app for nvidia part ?
How to do that???? Looks the same as a year ago.
I have RTX 3080 and my generation are not as fast like in this video, on what hardware was the samples generated on?
Is there a new RTX plugin in CumfyUI that I'm not aware of?
Well they are probably using rtx 4000 series to showcase it and those have faster and way more AI cores than the 3000 series
There is tensorrt since like today
40 series
@@jackie2-g8l not exactly, I found new node called ComfyUI_TensorRT and its amazing, its like reduced the image generation to 5-8 seconds on 1024 resolution on my 3080
Regretting paying for an education on Graphic Design now.
Don't regret that, trust me. This is only the beginning of the AI clusterfuck. There's still a demand for human made media. There will be for years. And the government will have to control who gets to use AI and even illegalize it due to how it's being misused now anyway.
Don't be. Your skill would still be superior to others.
I just graduated with a graphic design degree. I'm just tired of hearing about image generation so much.
I think most corporations would rather hire a graphic designer with hand-drawn / hand-made work than cheap out and use ai to stain their reputation or be called lazy.
can't you use AI in conjunction with your handrawn stuff?
Wow, the end product is a creepy and disturbing superhero, very cool.
So is this a node or what I don't get it. How do I make comfyui use tensor cores?
Awesome it show cases SD using comfy UI and not A1111
Is there much of a learning curve if I switch to comfy? I've been using A1111 for a few months.
@@yousuff1 Yeah, It's a stiffish learning curve, but well worth it. A1111 is knowing the check engine light. Comfy is fine tuning the hot rod motor under the hood. Easy to get started, but there is a lot to mastery.
Last time you did this for SD it quickly became abonened. It was useless anyway and caused nothing but trouble with speed improvements that have been superseded by alternatives by now.
I'd liked to generate a photo of the most beautiful pictures of real people.
that's insane.
What about slow rtx A6000 ADA slow as hell Expensive has shit you keep reaping people Nvidia most expensive consumer gpu is way slower than rtx 4090 how can high end content be made in SD with Nvidia
*smacks with shoe* bad nvidia this is not the ai we need
Sería bueno para implementarlo en Fooocus....
Lol is trx giving some updates for stable diffusion ans comfy
how can this be started even by ordinary people???
is it available in Linux?
yes
fun
Increasing prices of EVERYTHING, rent, food, healthcare, and now giving people's jobs to robots, what can go wrong? 😂
Replace human artists 7x faster with trash images that nobody other than Facebook grandmas wants
I don't care. Nobody I know cares. It's not even actual AI, it's just some nonsense... Besides, AI should be cleaning my house, and doing my shopping while I focus on writing or making art, not the other way around! This is so dumb...
Plz make ai to do simple chores like accounting rather than replacing human creativity which is inreplaceable
This is soulless and uncanny.
Awesome!
WHEN WUTHERING WAVES AND THE TOUCH CONTROLS OF THE HONKAI STAR RAIL?
RIP Graphics Designers 😢
Naaaaaaaaaaaaaah, have you tried any of them, they take a lot of time - hell lot of time ( done it myself ) - but if any big enterprise is looking for graphic services then there are companies who has lot of heavy cards but again power hungry and what not, that case is different also the money they will charge for that service is also going to be much more (obviously they invested in heavy machines so they want the returns as fast as possible ) and for small business owners or small work that would be a stupid idea if they use them soooooooo graphics designers aren't going anywhere. it will take little more years for AI to be more fast and less power hungry and with easy ui to get consistent results, consistency is also a big issue in this one too. read all the comments you will understand everything
My god, we live in a dystopia
Huh?
Cringe
Pfp checks out
amd fanboy detected
cause you don't know how to use comfyui? perhaps have zero understanding of how transformative discussion models have been or should I say, you have nill understanding of tech. Couldn't care less of nvidia or AMD (diffusion on AMD sucks), what they offering is amazing.
@@InfoWiser Glad to know you like replacing the talents of people in favor of soulless trash.
Thanks Nvidia, it's not bad enough artists all over are trying to fight AI from stealing their art and jobs, and now you just wanna make it EASIER to do so? Glad your dumb pockets are getting filled while real artists are struggling to find work as it was already.
You proved to have no knowledge regarding this topic and only repeat media propaganda.
@@BlankFX Oh wow, 20 minutes for a comment like this to show up, that took longer than i thought. I love when people like you talk about "no knowledge" with this topic. Not sure if it's comical or sad how ya'll think this isn't a problem, and just play down those of us who are actually being heavily effected by this. Hell, you didn't even bother to proof read your own comment, good job.
To say that the creativity of artists over the years, to me, is just a mixed pot mixed in different ways, in the whole history of a new thing created by others in art, it is rare. , you can admit that you draw a picture that you created, but isn't it still based on the ideas of predecessors with different variations? To me, what AI does is not much different from what artists do. What artist do today, there is only one difference: AI does not need a salary and you need a salary to live. Social development also leads to some occupations being replaced or changed, and that still happens. in the flow of human history.
@@chien.nguyen. Art is based on observation yes, but also emotions. Artists create to express themselves, and you can see it in their creations. AI cannot do this. If you type in a prompt, the person typing isn't creating anything. They aren't drawing the lines themselves. In fact, i would say that someone writing a prompt is more like someone writing what they request an artist to draw for them as part of a commission. The difference is, what they get is a soulless mess of various other artists, poorly crapped together and mangled by a machine. Not carefully crafted by a human. Humans learn from each other, and in turn we teach the next generation. AI doesn't start from a circle, and slowly build up the structure of a piece. It simply looks through it's database of artists who's pieces it stole, mash em together, and then does a bit of cleaning. Try to compare these all you want, but in the end they are two different actions, and without the emotional elements of the creator, they are just hollow. Also, don't bring up the topic of "this will help those with disabilities create art" when we have artists with no arms or legs, and are still able to create vibrant, emotional art by simply putting the stylus in their mouths and drawing that way. AI is not a living, breathing, emotional creature. This is not where it should be used, and instead of wasting AI on something it simply has no right to be a part of, it should be used for far better aspects of life like medicine and research to help with calculations and sequence models which benefit greatly from AI.
@@EricFaries Naaaah dont worry my guy dont worry, Naaaaaaaaaaaaaah, have you tried any of them, they take a lot of time - hell lot of time ( done it myself ) - but if any big enterprise is looking for graphic services then there are companies who has lot of heavy cards but again power hungry and what not, that case is different also the money they will charge for that service is also going to be much more (obviously they invested in heavy machines so they want the returns as fast as possible ) and for small business owners or small work that would be a stupid idea if they use them soooooooo graphics designers aren't going anywhere. it will take little more years for AI to be more fast and less power hungry and with easy ui to get consistent results, consistency is also a big issue in this one too. read all the comments you will understand everything
Pakistani design 😂
This sucks
Disgusting, this makes me wanna throw my nvidia card into the garbage :)
Do it, you wont
Get stuffed Nvidia
interesting
Gross.
Not sure why you'd want to generate images faster? AI has a place for those with disabilities, but certainly not an able bodied person. AI isn't ready. Leave it at that.
I'm so fucking tired of hearing about image generation. Bring on the AI that actually helps solve world problems.
So, this is where the tech companies have fallen into? We're so doomed as a society! 😔
How do we go about getting a hold of this, is this stuff going to be bundled into Chat with RTX or is it something separate. Also is this available to the general public or is it only behind closed doors?
And where is the workflow?
as far as i can see, this comfyUI workflow is using at least the following code: (maybe no links can be posted)
SDXL Evolved V1.0 = sdxl-evolved-photorealistic
Control Lora Depth Rank 128 = control-lora-depth-rank128.safetensors
Zoe Depth Anything = comfyui_controlnet_aux
InstantID = ComfyUI_InstantID
we can already run this with CUDA on pytorch, but TensorRT is much faster at running all of these things.
everything here is public code and the video shows how to hook it up.
There are instructions around to install with TensorRT. Give it a spin!
so where is the tensorrt for comfyui?
I'd love to know where I can find it. please tell me if you found a link
God those are so bad
Does it have to be an RTX card? Can a 1080 Ti using ComfyUI and SDXL run it, but slower?
yes, just slower.
Yes but it's a LOT slower (like taking minutes to do what a 4000 can do in seconds), and you need a decent amount of VRAM as well so probably anything less than 6GB is out. It can be done, though.
which app? anyone??
stable diffusion XL in comfy UI 0:39
@@2bnator so it's already out? I can test too?
@@itachi7976 already out! its been 3+ years and as far as i know its craze is over for many (including me). it was fun and now no more coz every site is charging money. you need models and training your own models takes a hell lot of time and if your settings are not optimal than time wasted. different plugins setups for different things like for consistent results and much more plus lot of space. also everything depends on the card too.
@@itachi7976 Idk.I used stable diffusion with web UI. I think all they used in the video open source?
Did you manage to get this going?
I amerely waiting for the day when AI will explain this as well.
Are you shure that These Video is Not ai generated?
SD has always been a big of a big jump to get into, but this looks relatively easier than it was a year ago!
please link comfyui workflow
Are tensorcores automatically enabled or what
Tensor cores are ONLY used for ai related computing. so yes and no. Essentially they are just sitting there idling until you actually give them a workorder
Yea, when the program is making calls on them. Games that use RTX and DLSS use tensor cores
you need to install the TensorRT Node for ComfyUI in the custom nodes
@@StefanSchmidtRegensburg Hi can you please tell me where can I find and access it? is it hosted on github?
@@thedarkfuture01101i see comfyui in Pinokio, a launcher for github apps codes, it's open source so user interface quite scuffed but fully automatic. I haven't tried yet comfyui because for now busy with Stable cascade there.
ForgeUI is faster than Comfy though.
Please never show this again
ok i'm seriously beginning to believe the marketing team for nvidia is high on something
ai hardware is such a rip off, make a gfx card with higher clock speeds or better instead of dlss software
I thought the thumbnail was Green Lantern.
I want to make Livvy Dunne into Supergirl!
AI image generation is honest to god the most useless AI tech ever. And it's sadly the most promoted.
Coming from the red team, all I can say is I'm jealous.
This is two year old tech and software essentially, but glad to see new interest
@@EhurtAfy AMD is weak in AI and 3D software rendering, which is why I'm jealous. Maybe I'll switch to Nvidia next year...
@@Crayoness Yeah I get that, I was using a GTX 1080 for AI in 2022, it took about a minute per image. I got so excited I ran out and bought an RTX 3090, it was about 10 seconds per image. Nvidia is king for AI right now and especially Stable Diffusion like in this video. Hope we get more open standards and open models so generation works on a variety of hardware and not just Nvidia 👍
workflow pls
Cringe
Listen, GeForce needs to be split off into it's own thing that focuses on gaming. We really don't care about AI, like really don't care.
Yes we do
Let's just fork off DLSS to the AI products. Wait...
Ai will be a big Think in the Future of Gaming.
Gaming is like the last thing they care about profitability wise.
@@5371W That's what I'm saying. All Nvidia cares about now is AI, and it's affecting the gaming brand.
Maybe we don't want more dorky AI art
Bye Bye Comic artists 🥲
The comic companies have been using overlay/trace methods to produce art faster for a long time now, often using copyrighted works as the basis for their stuff. This isn't really much different than that. Instead of unpaid interns doing tracing, they will have unpaid interns doing this instead.
@@jaturnley So Sad to hear that 🥲
What the sht is this
nvidia reminding us that we dont need humans anymore lmao
IMPRESSIVE... TEAM GREEN FOR LIFE
what is this ai doodoo, nobody asked for this