Definitely not the end, but I believe it will cause some disruption / transform the way we work a bit, just like how 'real time renderers' like Enscape + Lumion made visuals easier for 'non visualisers'. I made a new video yesterday on this topic if you want to check it out.
dude this is crazy i knew this was coming but my god its like 90% there. in a year or two renders made from scratch will be done for. also I always sucked at rendering, this will do 90 percent of the work.
I hope I can keep this info from my boss!! as a commercial interior designer for 17 years, this is really game changer for "concept" generation . Imagine I can finish work in a day rather than weeks...
@@varvara.g i am not on the client side who doesnt understand much and any image they see is their building. look again, its architecture, construction permit. do we need random alterations to our work? some of us yes, the other part no, thank you, i can make 2-15 proposals for a building without ai's help. anyway idiocracy is near, more than 40% of architecture means random facades, no propotions, no idea, no form, no nothing
pretty soon they will be able to do full on plans and details drawings and next thing you know boom , the field of architecture will be wiped out over night just like what happened to graphic designers pretty recently with DALE-2
nothing happened to graphic designer... its more to graphic design than just make beautiful pictures... and as an architect you need to know alot more than just make beautiful renders
I certainly see it happening. But 1000 thousand iteratives designs are of no use if you cant decide which one is better or better suits the needs of the project and most importantly, build it. We will have to adapt but we will be fine... for a while. When the AI learns how to deal with angry clients that dont want to overspend, then we are done.
i've said it before and i'll say it again, the gimmick of veras is that you can't get the same image twice nor can you get one image...move the camera and render out exact same image. Or better yet, add an additional 5 cameras around the space and get the same render + materials, lighting, etc.
How do they make the end result so neat? The samples look to have texture misalignment and the roof lines are not straight. Not like the website shows... Is this really meant to be a textural collage dreamer?
I think the challenges is that a lot of people won’t have much of an understanding of how stable diffusion and AI image creation work, and will expect this to function in a manner that results in interesting renders of a perfect model. But, at this point, the AI generated image2image when using one image is very janky. If you understand the limitations, then it’s great for iterative ideation, but not good as a final product generation tool. Also, all this program is doing is taking a Stable Diffusion iteration (probably 1.5) and having detailed prompt engineering (both positive and negative prompts) to create images. You could do this yourself and have far more flexibility. You could also generate 100 images of the same building and then train a new model specific for that project, and then generate MUCH more accurate renderings.
Can you explain a little further what you mean in you second paragraph? It sounds interesting but not sure I fully understand. When you say 'you could do this yourself and have far more flexibility' - How?
@@GeoPar Sure. The plugin has “stock” prompts that have been developed through trial/error and (from what I can see) don’t allow you to lock a seed and then it’s likely just appending what you enter to the much more developed prompts. The reason you can have very detailed prompt that generate very different iterations is because the seed changes. If you can lock a seed to something you like, then start tweaking the prompt, you’ll get much more similar results. Additionally, this program is essentially doing an image2image generation, which uses a seed and a provided image to generate a new image. With Stable Diffusion you can train a new model from 100’s (or 10, to 1000, or whatever you want) of a building from different angles and even including precedents images. With so much more data and a locked seed, you could produce renderings that are much more accurate to the actual design than using image2image. Don’t get me wrong, this is great for iterative ideation based on an image of a building you’ve designed, but at this point I’d never want to use this as a presentation image of a project.
I made a new video talking about why I think A.I will improve at a drastic rate in the future in terms of it's accuracy - to all the people saying it's not quite there yet (which is fair). ua-cam.com/video/u-0HsWUK4bU/v-deo.html Also made a Discord community for anyone who's interested in Revit / Tech in Architecture: discord.gg/9ynxXrMkSY
As an architect, this is a pretty cool plug-in, but it seems to lack precision and takes a lot of liberty with the design. Seems like it has potential, though.
There are many small perspective and shadow/lighting errors. It's apparent that the AI have no proper understandig of perspective (yet). But still it is damn pretty! Changing seasons in an image used to take me 40 hours of mixed 3d/2d work.
thats pretty cool never knew something like this existed, does it take into account the materials that are already assigned would be cool if it could use the metadata from that to help with its prompts
No, it just applies something close to the shown material in the realistic mode, not the exact one but close to it, and sometimes something off. IT can't recognize the trees or the cars and could show a car as a bush. I gave the software very clear projects but couldn't handle them well, In the video shows it is very ((intelligent)) but it's not.
For sure they need to work on recognizing the details - for example - the possibility to generate different floors and let everything else stay the same
OH BUDDY... but if you're a bit late on the work and have the meeting ahead.. you can effin show off "different concepts" to "explore the idea" and WHAAAM a happy client... but you cant rely or depend on that tho.. but great fire extinguisher.
I feel like this is the reverse of the how the workflow should work. Why spend hours making a meticulous model only to have AI show multiple options that would then have to be incorporated into the model? Would it not be better to have AI make inspirational images that then get transferred into Revit?
In the video 'creativity strength' was quite was fairly high. If i lowered it and increase style strength it'd have a more accurate representation. Done a video on this comparing A.I vs Render if you want to check it out (basically Veras vs Enscape on same scene).
This is all fine, but it pre-supposes that the designer has no actual design in their head that they're aiming at, and are just flailing around without direction. If you have no vision, the results are going to be mediocre, no matter how fast you can spit out results.
well, that's definitely a cool feat, but sadly nothing that's really useful just yet. Too inaccurate, I see those renders as something that stays between AI and architect, as, you know, a kind of dialogue in the searching phase. Still, it's a very interesting thing to know, let's stay in touch with the progress, folks )
in the future, you could just render a model with what you envision in your head or your clients, if they have good taste. . . ai steadily showing how it could wipe out careers, industries, and lines of work, just how long until it's my job. will AI ever replace the Architect? architects do specialize but most have to know a little about everything and its processes, AI will know everything about everything.
cool video, but what you said about how the AI imaging process works is not correct. The AI does not "search" images in the web and also does not combine them to create something new. The AI has been exposed to millions of images from the web together with keywords, tags and description of these images, and has now "learned" what certain words and phrases mean. It now understands what those prompts mean and can then reverse- engineer an image that will fit those terms.
it is cool but to tell you the truth I don't like it , it take away the person idea and creativity it take out the journey of the trail and error it take away the learning process how to make a render, yes its the future maybe or maybe bot be not, but for me its a no
This is the equivalent of saying I want continue travelling by horse because cars take out the fun in the journey. If the market is headed in that direction, that's where all the clients will go you will be left behind. However, this still needs a whole lot of improvements before it becomes a thing. Its a very good starting point especially if you are in the conceptual stage.
I agree, but in terms of trying lots of different concepts and the time it takes when you know not all are chosen, this can be useful then. But I'd definitely like to work on a 'final render' myself.
The beginning of the end for the Architectural profession.
its not... you really have to know alot more than just make beautiful renders...
Definitely not the end, but I believe it will cause some disruption / transform the way we work a bit, just like how 'real time renderers' like Enscape + Lumion made visuals easier for 'non visualisers'. I made a new video yesterday on this topic if you want to check it out.
dude this is crazy i knew this was coming but my god its like 90% there. in a year or two renders made from scratch will be done for. also I always sucked at rendering, this will do 90 percent of the work.
I hope I can keep this info from my boss!! as a commercial interior designer for 17 years, this is really game changer for "concept" generation . Imagine I can finish work in a day rather than weeks...
exactly
all here as a reminder of the higher definitions in design.
insane
i am dead
literally shivering (knowing the fact how much time i used to spend on lumion for renders, future is here!
without you , no one will need your services. as it is right now its unusable , time will tell
@@markflv I think you need to use these tools. Of course, somebody who uses it will be a few steps ahead of somebody who works in old ways
@@varvara.g i am not on the client side who doesnt understand much and any image they see is their building. look again, its architecture, construction permit. do we need random alterations to our work? some of us yes, the other part no, thank you, i can make 2-15 proposals for a building without ai's help. anyway idiocracy is near, more than 40% of architecture means random facades, no propotions, no idea, no form, no nothing
Thank u for spending your free trials to show us this feature
Mind blowing plugin. Keep making such videos like this. It is very helpful.
“Prompt engineer” is the new data scientist
You're getting some cool results! Thanks for sharing!
It's great to see you making these tutorials - youve always been slick with the software. Hopefully I can catch you again soon Dev
pretty soon they will be able to do full on plans and details drawings and next thing you know boom , the field of architecture will be wiped out over night just like what happened to graphic designers pretty recently with DALE-2
nothing happened to graphic designer... its more to graphic design than just make beautiful pictures... and as an architect you need to know alot more than just make beautiful renders
Revit's Generative Design is on track for those exact functions...
I certainly see it happening. But 1000 thousand iteratives designs are of no use if you cant decide which one is better or better suits the needs of the project and most importantly, build it. We will have to adapt but we will be fine... for a while. When the AI learns how to deal with angry clients that dont want to overspend, then we are done.
That is wild🔥🔥🔥🔥🔥🔥🔥🔥🔥
GAME.CHANGER.
i've said it before and i'll say it again, the gimmick of veras is that you can't get the same image twice nor can you get one image...move the camera and render out exact same image. Or better yet, add an additional 5 cameras around the space and get the same render + materials, lighting, etc.
With Neural Networks like these, the more detailed the prompt, the better. More keywords mean AI knows to look through more references
Thank you so much for the video! I wanted to know is how can I get AI renders for revit and download it?
it works for default house and this angle only, tried to render my thatre project this way, no chance, unfortunately(
cool video 🤙
How do they make the end result so neat? The samples look to have texture misalignment and the roof lines are not straight. Not like the website shows... Is this really meant to be a textural collage dreamer?
Could be usefull to be able to change the viewport, but keep the design / materials. I’ll sure have to test this. Thanks for sharing.
Could also be useful to be able to save a “style / atmosphere”, and update the model, without everything changes.
At this point im convinced 95% of software claiming to be Ai is actually machine learning.
Well it is
Machine learning is subset of AI...
What are you even talking about man 🤣
Good evening sir. Must you be online to start using Veras.?
I think the challenges is that a lot of people won’t have much of an understanding of how stable diffusion and AI image creation work, and will expect this to function in a manner that results in interesting renders of a perfect model. But, at this point, the AI generated image2image when using one image is very janky. If you understand the limitations, then it’s great for iterative ideation, but not good as a final product generation tool.
Also, all this program is doing is taking a Stable Diffusion iteration (probably 1.5) and having detailed prompt engineering (both positive and negative prompts) to create images. You could do this yourself and have far more flexibility. You could also generate 100 images of the same building and then train a new model specific for that project, and then generate MUCH more accurate renderings.
Can you explain a little further what you mean in you second paragraph? It sounds interesting but not sure I fully understand. When you say 'you could do this yourself and have far more flexibility' - How?
@@GeoPar Sure. The plugin has “stock” prompts that have been developed through trial/error and (from what I can see) don’t allow you to lock a seed and then it’s likely just appending what you enter to the much more developed prompts. The reason you can have very detailed prompt that generate very different iterations is because the seed changes. If you can lock a seed to something you like, then start tweaking the prompt, you’ll get much more similar results. Additionally, this program is essentially doing an image2image generation, which uses a seed and a provided image to generate a new image. With Stable Diffusion you can train a new model from 100’s (or 10, to 1000, or whatever you want) of a building from different angles and even including precedents images. With so much more data and a locked seed, you could produce renderings that are much more accurate to the actual design than using image2image.
Don’t get me wrong, this is great for iterative ideation based on an image of a building you’ve designed, but at this point I’d never want to use this as a presentation image of a project.
I have a question please , can I use the same rendering effect for multiple angles of the building ?
Can i install it on a windows PC?
Absolutely, that's what I used.
I made a new video talking about why I think A.I will improve at a drastic rate in the future in terms of it's accuracy - to all the people saying it's not quite there yet (which is fair).
ua-cam.com/video/u-0HsWUK4bU/v-deo.html
Also made a Discord community for anyone who's interested in Revit / Tech in Architecture: discord.gg/9ynxXrMkSY
As an architect, this is a pretty cool plug-in, but it seems to lack precision and takes a lot of liberty with the design. Seems like it has potential, though.
Made a new video on my channel where I explained why I think it'll get only get better and my prediction of 'how' if you want to check it out. :)
We want more revit content ❤
There are many small perspective and shadow/lighting errors. It's apparent that the AI have no proper understandig of perspective (yet).
But still it is damn pretty!
Changing seasons in an image used to take me 40 hours of mixed 3d/2d work.
Awesome
the death of achviz; when architectural assistants can just click a button
thats pretty cool never knew something like this existed, does it take into account the materials that are already assigned would be cool if it could use the metadata from that to help with its prompts
No, it just applies something close to the shown material in the realistic mode, not the exact one but close to it, and sometimes something off.
IT can't recognize the trees or the cars and could show a car as a bush.
I gave the software very clear projects but couldn't handle them well, In the video shows it is very ((intelligent)) but it's not.
visible in revit lite?
This is great
very cool AI software i will try this
For sure they need to work on recognizing the details - for example - the possibility to generate different floors and let everything else stay the same
good video
mind blown
Exterior #2 the creativity strength slider is up so high it is changing it too much.
WOOOW WHAT ;;;;;;;;;
Not sure how useful it is when you design a building and get a rendering with a randomly generated design instead of your design intent..
OH BUDDY... but if you're a bit late on the work and have the meeting ahead.. you can effin show off "different concepts" to "explore the idea" and WHAAAM a happy client... but you cant rely or depend on that tho.. but great fire extinguisher.
I feel like this is the reverse of the how the workflow should work. Why spend hours making a meticulous model only to have AI show multiple options that would then have to be incorporated into the model? Would it not be better to have AI make inspirational images that then get transferred into Revit?
RIP 3D illustrators
yeah great, it is my dreamjob, now i probably have left 2, 3 years in my profession...
@@METTI1986LA I’m working as one and have mixed feelings about the future.
@@BossToastMaker me too man, the only thing That is still our advantage is that we can make 3d Animations...
@@METTI1986LA dont right that in chats some AI might be collecting data
And ll work on it too!
@@BossToastMaker nope, if you watched the video, the renderings are not accurate to the 3d model.
Unbelievable! Imagine that is connected with Neuralink! Inception in real life.
Unfortunately, trial version does not offer full function.
It's too limited
It too early to comment anything without testing full option
ai for sketchup yet?
Are you mad? Sketchup doesnt even have a sheet metal tool
Wow I see in just two weeks the pricing has changed to 34/month (yearly) and 49/month (monthly), and only 30 free renderings in the trial..
I just noticed this too, I guess it's cause the plugin got quite popular it ended up demanding too much 'free' resources :(
How can this be useful if it changes the design/model? It is not rendering what is in the model.
In the video 'creativity strength' was quite was fairly high. If i lowered it and increase style strength it'd have a more accurate representation. Done a video on this comparing A.I vs Render if you want to check it out (basically Veras vs Enscape on same scene).
This is sweet but detrimental to peoples jobs
This is all fine, but it pre-supposes that the designer has no actual design in their head that they're aiming at, and are just flailing around without direction. If you have no vision, the results are going to be mediocre, no matter how fast you can spit out results.
It did very bad renders for my projects, and whatever i tried it didn't give me any good rendering not even something close to what i wanted it to be.
It definitely takes some trial and error, not all my images were good/usable either, had to tweak with settings or even the shot.
well, that's definitely a cool feat, but sadly nothing that's really useful just yet. Too inaccurate, I see those renders as something that stays between AI and architect, as, you know, a kind of dialogue in the searching phase. Still, it's a very interesting thing to know, let's stay in touch with the progress, folks )
The end of freelance render services
in the future, you could just render a model with what you envision in your head or your clients, if they have good taste. . . ai steadily showing how it could wipe out careers, industries, and lines of work, just how long until it's my job.
will AI ever replace the Architect? architects do specialize but most have to know a little about everything and its processes, AI will know everything about everything.
interesting
cool video, but what you said about how the AI imaging process works is not correct. The AI does not "search" images in the web and also does not combine them to create something new. The AI has been exposed to millions of images from the web together with keywords, tags and description of these images, and has now "learned" what certain words and phrases mean. It now understands what those prompts mean and can then reverse- engineer an image that will fit those terms.
Thanks for explaining this!
Without obtaining rights and paying licences to those original creators, mind you.
This is scary but it is not there yet to replace human beings. AI can't design. They can copy other works though.
it is cool but to tell you the truth I don't like it , it take away the person idea and creativity it take out the journey of the trail and error it take away the learning process how to make a render, yes its the future maybe or maybe bot be not, but for me its a no
And it takes out the boring stuff gaining us time to invest in something else.
those ideas and creativity have long been gone when digital rendering came existed. There's no difference for me.
This is the equivalent of saying I want continue travelling by horse because cars take out the fun in the journey. If the market is headed in that direction, that's where all the clients will go you will be left behind. However, this still needs a whole lot of improvements before it becomes a thing. Its a very good starting point especially if you are in the conceptual stage.
@@abdallahdahir2834 I'd take horse and carriages over cars any day
30$/mo for 768x1334 resolution? what a joke.
Everything creative is dead
Rendering is the most enjoyable part of the job, why give it to a robot lol 👎
I agree, but in terms of trying lots of different concepts and the time it takes when you know not all are chosen, this can be useful then. But I'd definitely like to work on a 'final render' myself.
That's so helpful, got jury tomorrow, whats your insta handle btw :'