I think a big factor going forward with diffusion is going to be the user base learning the language of this AI software. once we know how to prompt it better we might start to get more desirable results. being able to pinpoint certain elements of a photo would be super helpful aswell may be a new feature once it leaves labs
I agree. You have to be able to label parts and regions of the basic sketch to indicate how you want the AI to render them (eg set colours and textures) rather than list a bunch of words in a separate box and hope the AI can figure it out.
The main problem with AI Rendering is that it takes a lot of renders and put a little bit of each of them into your render. There are many more problems, these are just a few: 1. You cannot render from multiple perspectives because it will render differently each time. 2. He changes the design of the object too much and adds some unrealistic things (e.g. replace the wardrobe in place of the TV). 3. He himself invents the environment and roads and breaks them. 4. Textures are placed by him and you have no control over adjusting that texture. 5. Cannot create animation. 6. What if a client asks to change the rendering (for example, to increase the window size), how will we do the same rendering again? We can't do it at all. There is also one good thing, which is called concept design, to give us an idea. I support AI for that. In my opinion, for now it is best to stick to classic 3D modeling and rendering.
Thanks for the rundown! Always wanted a quick and dirty way to make SU images OK for viewing. All the Dennis methods I've done take too long, Visualizer came close but needed ability to adjust. What I want is lighting enhancement, entourage and background. But there is indeed a problem if the software is going to change materials etc. The client would immediately question the design. Addition of shaded lighting and reflections brings life to the model immediately.
Well exactly - there's no real place for design iteration in this (or most) AI workflows. It's fine for conceptual idea generation, but as soon as someone wants to change a flooring material and keep everything else the same it's out the window. There were some interesting new features added into D5 render in their newest feature that I want to check out. They have an AI environment generator that changes the environment but not the geometry, and a tool that auto generates PBR maps for rendering, both of which I think are more in the direction of what useful "workflow" AI might look like
Ideas generated are random still. Even thought it's promising and placing image helps with editing the geometry behind, not being able to change minor things or parts is a big problem. I recommend using online AI tools and photoshop AI for that. But amongst all of these, the coolest for us is the Ipad version. That's very useful.
First ytb to bring up the issues with AI for all those who actually use 3d professionally. You just don't have enough control and fine tuning to make it usable commercially.
Not with these Tool.. If you want that, you need to Localy install tools like Comfy UI, Invoke Ai ore Automatic1111 where you can also inpaint, use Image transfer tools, controlnets, and where you can create your own node workflows to get specialiced results.. Thats allready by far better but of course also less easy.. But the biggest Problem is, that the moddels are not trained for big architecture yet.. for smal houses, livingrooms, or kitchens the dataset involves allready enough images.. But you can be sure.. Big Arealviews in What ever Ai you use are still a Blind spot..
I got my master degree in architectural computing at 1990 & my thesis was about Expert System of AI. At that time pattern recognition was something I supposed to be something really futuristic thing. Now I am presented with the fruitful application of AI. Diffusion is able to 'read' the image and translate 'the context' of our prompt. But unfortunately (I hope this only about time) many options can happen & may not suit our intention. Hope more control will be provided for the designer
Hey Justin, thank you for your content is awesome. I've being following for years now. I do think this new tool is going to help tons, specially considering the other rendering tools where all paid and Blender learning curve so slow. I will def use it a lot for "quick and dirty" visualization on my projects. On a separate note, what happened to the 3D warehouse? I used to enjoy so much seeing how many times and when each of my models got downloaded, etc. Now I don't seem to get any info, just a shitty analytics graph... I will no longer upload my models because of it. Feels like they just want to use us the users and not give anything back. Cheers!
I am trying to see how AI can be used in my work process, and so far, I can only see it used as an image generator for inspirational images to work from. I am using Skecthup and Layout for me work using Nick Sonders' methods. Even the renderings, as you have pointed out here, can not be controlled for accuracy. There are other AI software out there that can basically design the whole thing and go to CD's but they are for multi-family residential work, and not single-family custom projects.
Trimble: We have this amazing wowie tool you're omg gonna love so hard. You're welcome!! Users: Well that was fun for a few renders. I'll save this in my "could be useful eventually" folder.
diffusion kicks asss!!!!!!! sketchup is way ahead in the game now you can do 2D 3D and Renders all in one program either on your pc or your ipad . 5 stars from me all the way
From what I've used with AI generative imaging, it would all be based on how deep the AI model was trained. It seems there isn't a linear correlation to descriptor usage. Meaning, in reality it'd probably only pick up the "cut oak" descriptor in your sentence and of course, "cut oak" can be wide open to interpretation. I'm sure the deeper these models are trained, the better they'll become.
I'm intrigued by this. I still run a permanent licence for 2020. I'd like to know if you can get a generated image to match the viewport size? I like doing images with linework overlays and it's a pita trying to match a vray safe view to the sketchup window view. Something like this my tempt me to jump on the sub bandwagon. The other reason I probably will is trimble have stopped supporting 202 format which is a bit mean imho as they know people have licenses.
hello. how can I use Sketchup diffusion in web version. I have not purchased Sketshup so is it might be reason? or does anyone have any idea how to use this AI for completely free?
You are great, you are my go to for all things SketchUP, but this product I am really not finding helpful. Very basic and inaccurate so far, however I look forward to the improvements as I feel it definitely has its purpose for quick visualizations.
Justin Thank you so much for all your amazing videos, I have a quick question. Usually, Vray and Enscape are very heavy for my computer, so they made my computer shut down several times, so I wonder if does this AI also require an enough capacity on my computer. It seems it doesn't, but I want to make sure 100% ! Thank you
AI is processed on the servers of the AI company that provide the service to elaborate and produce those images, so your computer is not really involved in the process of the creation of the image, like instead happens when you use any other rendering image that use your own processor.
I'm pretty sure adding a integration to a web service and COMPLETELY changing the engine under the hood of a 3D modeling program are not the same thing....And honestly, I'm not sure that's a change I'd want, as much as I do like tools like Fusion 360 and Solidworks
@@Thesketchupessentials Plasticity. I'm learning that every day now after 10 years with Sketchup. Sketchup hasn't been upgraded with anything that genuinely increases productivity since Trimble bought it. Yes a few of the 3D party plugins have been very innovative; but SketchUp itself? Very little.
D5 Render is better than Enscape. Diffusion simply takes your Sketchup design and makes a "randomized" single image. Final rendering software such as D5 Render or Enscape offers far superior control without the randomness.They also offer video making abilities.
No ... not yet.. it lacks very much control, even with far superior tools like InvokeAI or ComfyUI... at least theses tools.. Invoke for example are the dooom for Graphic Artists.. but not yet for architecture, because you have to be much much much much much more precice here as if you just created some nice girls or advertising images..
Are you sure? It says in the extension Warehouse that it's compatible back through 2021, and it's showing up in my web version as well. Now if your SketchUp is older than 2021, yeah I wouldn't think it would work
sketchup should do better..many of the problem shown in this video can be solve using segment anything and inpainting. in which the segment anything will seperate each object based on colors code (material) and inpaint it to replace the object with the one we need... even photoshop generative fill is great for removing object and introduce new thing
Interesting - I think I see what you're getting at, but I'm not sure if SketchUp currently has a tool that does the masking that would need to be fed to this tool - that would probably be the time consuming thing about this I would think
I hate Ai. I used to be the most passionate guy out there, and it’s destroyed my passion. Everything I care about and that moves me is completely trivialized by generative Ai bs (that’s built upon nothing but stealing from all of humanity’s great artists.)
Weeelllll it not yet there.. But what else are architects doing when creating architecture? They start with a moodboard, remember all the images they have seen before, alll what their Wetware has learned so far.. its realy not that different, exept, that the Artificial intelligences are not yet! as good as humans at the moment.. But this might change.. Of course its dangerous.. for jobs, for news, for fakenews, and furthere more it uses up gigantic energy, so beeing bad for the Planet, and... some more realy big! dangers.. But that won t stop humankind using it. because it can make at least in the future many companys much more efficient (just look at microsoft copilot) so everyone will try to be first with this, to not get eaten by competition.
It's the fault of where A.I. is at currently. I can imagine using it at early stages of design and wondering what a few versions would look like such as more windows, less doors, more vertical lines, etc. It might give you some ideas to explore later. For anything approaching a final presentation, it's absolutely useless.
I think it’s actually pretty helpful with very early visualisation. Especially in understanding the feel of a building in space. Before I would need to spend a whole day dressing a model (plants, furniture, lighting, reflections, materials) for a single render to get a feel of the building to see if it’s working. Not to mention setting weather and environmental conditions. Now I only need basic geometry and a prompt to generate hundreds of variations. It doesn’t replace traditional rendering but it enhances early development significantly.
Well if you realy want more control, you cant go with thes very reduced Plugins.. You rather would go with standalone installatons using the frontends... Automatic1111, ComfyUI or the best for architects.. InvokeAI. You got here by far more control, you can inpaint, outpaint, use IP-Adapters (transfer a style of an image to another one..etc etc). But at the moment all of the Ai´s even with INvoke ore Comfy UI are bad for big architecture, because that hasn´t been in the trainings dataset enough at the moment. But thats just a Question of furthere trained better Ai Moddels..
Hi everyone! Let me know what you think about SketchUp Diffusion in the comments below! :)
its Free oR paid ???
Diffusion is a subscription-only feature. To get access, start a trial or choose a subscription plan
how to fix that, i can't open
its free for now. still in development. they will charge for it in the future.@@usmansadiq6006
is it for free?
I think a big factor going forward with diffusion is going to be the user base learning the language of this AI software. once we know how to prompt it better we might start to get more desirable results. being able to pinpoint certain elements of a photo would be super helpful aswell may be a new feature once it leaves labs
I agree. You have to be able to label parts and regions of the basic sketch to indicate how you want the AI to render them (eg set colours and textures) rather than list a bunch of words in a separate box and hope the AI can figure it out.
The main problem with AI Rendering is that it takes a lot of renders and put a little bit of each of them into your render.
There are many more problems, these are just a few:
1. You cannot render from multiple perspectives because it will render differently each time.
2. He changes the design of the object too much and adds some unrealistic things (e.g. replace the wardrobe in place of the TV).
3. He himself invents the environment and roads and breaks them.
4. Textures are placed by him and you have no control over adjusting that texture.
5. Cannot create animation.
6. What if a client asks to change the rendering (for example, to increase the window size), how will we do the same rendering again? We can't do it at all.
There is also one good thing, which is called concept design, to give us an idea. I support AI for that.
In my opinion, for now it is best to stick to classic 3D modeling and rendering.
Thanks for the rundown! Always wanted a quick and dirty way to make SU images OK for viewing. All the Dennis methods I've done take too long, Visualizer came close but needed ability to adjust. What I want is lighting enhancement, entourage and background. But there is indeed a problem if the software is going to change materials etc. The client would immediately question the design. Addition of shaded lighting and reflections brings life to the model immediately.
Well exactly - there's no real place for design iteration in this (or most) AI workflows. It's fine for conceptual idea generation, but as soon as someone wants to change a flooring material and keep everything else the same it's out the window. There were some interesting new features added into D5 render in their newest feature that I want to check out. They have an AI environment generator that changes the environment but not the geometry, and a tool that auto generates PBR maps for rendering, both of which I think are more in the direction of what useful "workflow" AI might look like
D5--Thanks for the reminder. I haven't needed too much rendering this year so I've lost track. Things are changing fast.@@Thesketchupessentials
Good analysis of a developmental tool, 'Things can only get better'!
Ideas generated are random still. Even thought it's promising and placing image helps with editing the geometry behind, not being able to change minor things or parts is a big problem. I recommend using online AI tools and photoshop AI for that. But amongst all of these, the coolest for us is the Ipad version. That's very useful.
First ytb to bring up the issues with AI for all those who actually use 3d professionally.
You just don't have enough control and fine tuning to make it usable commercially.
Not with these Tool.. If you want that, you need to Localy install tools like Comfy UI, Invoke Ai ore Automatic1111 where you can also inpaint, use Image transfer tools, controlnets, and where you can create your own node workflows to get specialiced results..
Thats allready by far better but of course also less easy.. But the biggest Problem is, that the moddels are not trained for big architecture yet.. for smal houses, livingrooms, or kitchens the dataset involves allready enough images.. But you can be sure.. Big Arealviews in What ever Ai you use are still a Blind spot..
Cool tool. I will be using this to help customers visualize projects better. The AI has a bit of randomness but that can always inspire an idea.
I got my master degree in architectural computing at 1990 & my thesis was about Expert System of AI. At that time pattern recognition was something I supposed to be something really futuristic thing. Now I am presented with the fruitful application of AI. Diffusion is able to 'read' the image and translate 'the context' of our prompt. But unfortunately (I hope this only about time) many options can happen & may not suit our intention. Hope more control will be provided for the designer
I believe this will also change as we as the user base learns how to "speak the language" of diffusion
Hey Justin, thank you for your content is awesome. I've being following for years now. I do think this new tool is going to help tons, specially considering the other rendering tools where all paid and Blender learning curve so slow. I will def use it a lot for "quick and dirty" visualization on my projects. On a separate note, what happened to the 3D warehouse? I used to enjoy so much seeing how many times and when each of my models got downloaded, etc. Now I don't seem to get any info, just a shitty analytics graph... I will no longer upload my models because of it. Feels like they just want to use us the users and not give anything back. Cheers!
excatly , why the frick they ruined the whole warehouse system, like who asked
I am trying to see how AI can be used in my work process, and so far, I can only see it used as an image generator for inspirational images to work from. I am using Skecthup and Layout for me work using Nick Sonders' methods. Even the renderings, as you have pointed out here, can not be controlled for accuracy. There are other AI software out there that can basically design the whole thing and go to CD's but they are for multi-family residential work, and not single-family custom projects.
What programs are you seeing that can design a complete project through CD's? I've not seen anything close to that level of sophistication yet
Trimble: We have this amazing wowie tool you're omg gonna love so hard. You're welcome!!
Users: Well that was fun for a few renders. I'll save this in my "could be useful eventually" folder.
Really interesting view point. It’s early days but as you say a great tool like many others
One of my class midterms is making me make an outdoor space but I have no idea how to use sketchup
Hi, is SketchUp Go also have option of Diffusion in ipad for Rendering. Thanks
diffusion kicks asss!!!!!!! sketchup is way ahead in the game now you can do 2D 3D and Renders all in one program either on your pc or your ipad . 5 stars from me all the way
Is it possible to use the same render style in different views?
I think that's the biggest issue for me and no one talks about when it comes to AI in architecture industry
Hi will you consider doing a video about how to get a render with a sketchup model into Coohom
You'll need to export the model first
Or save as skp file..then drag it as skp file in Coohom
Free extension?
Could the way you name materials in a model i.e. “rift cut white oak” v “wood” improve the material rendering?
From what I've used with AI generative imaging, it would all be based on how deep the AI model was trained. It seems there isn't a linear correlation to descriptor usage. Meaning, in reality it'd probably only pick up the "cut oak" descriptor in your sentence and of course, "cut oak" can be wide open to interpretation. I'm sure the deeper these models are trained, the better they'll become.
@@schlenbea Thanks 🙏@schlenbea 🙏
I'm intrigued by this. I still run a permanent licence for 2020.
I'd like to know if you can get a generated image to match the viewport size? I like doing images with linework overlays and it's a pita trying to match a vray safe view to the sketchup window view.
Something like this my tempt me to jump on the sub bandwagon. The other reason I probably will is trimble have stopped supporting 202 format which is a bit mean imho as they know people have licenses.
That’s high value content! Thanks!
Glad it was helpful!
hello. how can I use Sketchup diffusion in web version. I have not purchased Sketshup so is it might be reason? or does anyone have any idea how to use this AI for completely free?
You are great, you are my go to for all things SketchUP, but this product I am really not finding helpful. Very basic and inaccurate so far, however I look forward to the improvements as I feel it definitely has its purpose for quick visualizations.
Fair comment :)
Please help : my sketchup diffusion Generate no image. Many thanks
Justin Thank you so much for all your amazing videos, I have a quick question. Usually, Vray and Enscape are very heavy for my computer, so they made my computer shut down several times, so I wonder if does this AI also require an enough capacity on my computer. It seems it doesn't, but I want to make sure 100% ! Thank you
AI is processed on the servers of the AI company that provide the service to elaborate and produce those images, so your computer is not really involved in the process of the creation of the image, like instead happens when you use any other rendering image that use your own processor.
@@robertoercolani9978 exactly. It's using cloud computing so there would be no additional effect on your CPU/GPU.
Is there a way to do parametric design just grasshopper inside SKP?
Thank you Justin, it's always good to have AI. Diffusion is lots of fun and a great tool at the very beginning of the design stage!^^
Converting it to the Parasolid kernel would acually be useful..not just bolting on more gimmiks.
I'm pretty sure adding a integration to a web service and COMPLETELY changing the engine under the hood of a 3D modeling program are not the same thing....And honestly, I'm not sure that's a change I'd want, as much as I do like tools like Fusion 360 and Solidworks
@@Thesketchupessentials Plasticity. I'm learning that every day now after 10 years with Sketchup. Sketchup hasn't been upgraded with anything that genuinely increases productivity since Trimble bought it. Yes a few of the 3D party plugins have been very innovative; but SketchUp itself? Very little.
How can I define material? Or hidden lights for example
I don't know what this means - you basically have to describe what you want
@@Thesketchupessentials useless tools, just for fun not for professionals
im glad than this is not strong like i imaginate ai in the begnng, good tool for inspiration in a volumetric order
Thanks for posting
It would be cool if I could place my model in a car trunk.
that was nice, but AI was NOOOOT dealing well with straight lines in my renders. perhaps I need to raise the "respect original geometry" bar
How to uninstall Diffusion? Thank you!
Also is this better or enscape?
D5 Render is better than Enscape. Diffusion simply takes your Sketchup design and makes a "randomized" single image. Final rendering software such as D5 Render or Enscape offers far superior control without the randomness.They also offer video making abilities.
No ... not yet.. it lacks very much control, even with far superior tools like InvokeAI or ComfyUI... at least theses tools.. Invoke for example are the dooom for Graphic Artists.. but not yet for architecture, because you have to be much much much much much more precice here as if you just created some nice girls or advertising images..
It renders interiors as well?
Do some Googling. There are stack of apps that do it now. This is just bandwagon jumping.
who wants to love and learn sketchup should give u subscribed
thanks
that's cool
I always wondered if some day you would use one of my models in your videos… very cool!
Hey Mike! I think I've used your models in several of my videos - very well done! Thanks for making them available! :)
i cant open it,
This will kill all other rendering softwares lol
th-cam.com/video/zAiKI4gHvvY/w-d-xo.html
funtastic
d5 ai is the best for architecture
I've not tested that one yet, but it's definitely on my list :)
Doesn't work as intended so pretty much unusable. Maybe in the next few iterations of the software it can be used professionally.
your license has to be subscription, does not work on classic licenses
Are you sure? It says in the extension Warehouse that it's compatible back through 2021, and it's showing up in my web version as well. Now if your SketchUp is older than 2021, yeah I wouldn't think it would work
@@Thesketchupessentials yes, i have a perpetual license, it keeps saying i need to have subscription based license. May be it’s a bug or smth? 🤷🏻♂️
@@Thesketchupessentials I use 2023 latest version
same@@1369legend
@@1369legend It's the same for me. Does not work with a permanent license.
Found the download process to be frustrating. Their directions less than helpful.
Downloading what?
Wow😊
sketchup should do better..many of the problem shown in this video can be solve using segment anything and inpainting. in which the segment anything will seperate each object based on colors code (material) and inpaint it to replace the object with the one we need... even photoshop generative fill is great for removing object and introduce new thing
Interesting - I think I see what you're getting at, but I'm not sure if SketchUp currently has a tool that does the masking that would need to be fed to this tool - that would probably be the time consuming thing about this I would think
The plural of software is software just as the plural of sheep is sheep.
k
@@Thesketchupessentials I love you.
crack works ;)
hey, how did you do it with cracked version?
I hate Ai. I used to be the most passionate guy out there, and it’s destroyed my passion. Everything I care about and that moves me is completely trivialized by generative Ai bs (that’s built upon nothing but stealing from all of humanity’s great artists.)
Weeelllll it not yet there.. But what else are architects doing when creating architecture? They start with a moodboard, remember all the images they have seen before, alll what their Wetware has learned so far..
its realy not that different, exept, that the Artificial intelligences are not yet! as good as humans at the moment.. But this might change.. Of course its dangerous.. for jobs, for news, for fakenews, and furthere more it uses up gigantic energy, so beeing bad for the Planet, and... some more realy big! dangers..
But that won t stop humankind using it. because it can make at least in the future many companys much more efficient (just look at microsoft copilot) so everyone will try to be first with this, to not get eaten by competition.
What happened to CREATIVITY. God have mercy on our souls
If it could work with textures... would be useful. But, it doesn't... so it's not.
i am going to loss my job soon
To AI? What do you do?
@@Thesketchupessentials I am a freelance Architect but mostly I get 3D visualizing jobs.
Almost useless plugin IMHO
It's the fault of where A.I. is at currently. I can imagine using it at early stages of design and wondering what a few versions would look like such as more windows, less doors, more vertical lines, etc. It might give you some ideas to explore later. For anything approaching a final presentation, it's absolutely useless.
I think it’s actually pretty helpful with very early visualisation. Especially in understanding the feel of a building in space. Before I would need to spend a whole day dressing a model (plants, furniture, lighting, reflections, materials) for a single render to get a feel of the building to see if it’s working. Not to mention setting weather and environmental conditions. Now I only need basic geometry and a prompt to generate hundreds of variations. It doesn’t replace traditional rendering but it enhances early development significantly.
Well if you realy want more control, you cant go with thes very reduced Plugins..
You rather would go with standalone installatons using the frontends... Automatic1111, ComfyUI or the best for architects.. InvokeAI.
You got here by far more control, you can inpaint, outpaint, use IP-Adapters (transfer a style of an image to another one..etc etc). But at the moment all of the Ai´s even with INvoke ore Comfy UI are bad for big architecture, because that hasn´t been in the trainings dataset enough at the moment.
But thats just a Question of furthere trained better Ai Moddels..
@@xaviervascoexactly my thoughts
🎉😊🎉🎉
if you don´t want the tv appears, just hide it in sketchup