Chat3D is a cool idea. I've seen comments on other videos showing AI created 3D printable objects with criticism of the process, preferring to use CAD rather than going back and forth trying to describe to an AI what you want it to do, and I can see how that would be frustrating to someone with good CAD skills but it misses the point. It was easier and safer to walk than fly in the Wright Flyer but if everyone walked we'd never have aviation. I look forward to seeing where Chat3D is going. I played with OpenSCAD years ago and was impressed with how quickly object complexity was created and how it facilitated creativity, particularly iterating geometric designs. That's a good way to create visually complex vases, flower pots, lamp shades, abstract sculptures. etc. It was so fun that I'm surprised someone hasn't created a website that's an OpenSCAD front end that allows people to create their own lamps or vases that they can then have 3D printed and shipped to them. In a world of mass produced products, I think people would appreciate having a truly unique item they designed themselves.
This is fire dude, great job, this works like a final product, if this is implemented into consumer cheap printer that would lower the barrier to entry to insane levels
I know people love downloading models and 3D printing them but that was never my cup of tea. We're starting to see AI tools to create meshed object models. "Make a pig with a straw hat, sitting on a rock." Kinda cool but still not my cup-o-tea. An AI that allows functional parts to be created from a description would make 3D printers useful for regular people who want to make the same type of functional parts I make, but who don't want to learn CAD. This is a much larger market.
Very cool! Features I would like and would pay for: uploading an STL (technical parts) or step and getting a parametric model - since I do a lot of reverse modeling.
- Thx for sharing. - First, wishing you success in your endeavor! You have clearly put a lot of effort into things. - BTW, I'm a retired 3D guy (started in 3D 30+ years ago). Have explored many aspects of 3D, including writing custom code professionally. Also, I have a significant background in mathematics, plus a formal degree in engineering. Oh and, I have been keeping up with the many recent developments in AI (including the math/algorithms). - In addition to a personal project for your learning/growth, what is the goal for the end use? What are the intended use cases? Who are the target users? What are the benefits of this "system" of modeling versus others in its class? - Now, to give proper feedback, and constructive criticism, the intent of the project must be known, which at this time I do not (just a hint, but not enough details). Anyhow... - As for CSG, it's good at what it does. It has its capabilities, but also its limitations. - My current focus is 3D CAD - specifically, for design, modeling, manufacturing (subtractive, and additive). --- I have used OpenSCAD, and Tinkercad - both CSG-based, but they are generally not well-suited for proper 3D CAD. --- As for OpenSCAD... ----- There are "enhanced" forks (perhaps you are already aware). Does/could Claude support such? One might offer some advantage. Hm? ----- As for export formats, a mesh is fine for 3D printing - but as you noted, a .csg can be exported, then imported into FreeCAD (FC), then exported as '3D solid' (STEP). I will note though (and you may already be aware), I've had limited success with this: FC does not parse all trees correctly. - OK, hopefully there's been some value in my comments. But as with any feedback, you can take it or leave it, in part or in total. - And again, wishing you success; sounds like an interesting/challenging project...
thanks for your thoughts! I am aware of the OpenScad limitations and for sure I am looking into alternatives for comparison like directly using freecad scripting, replicad or cadquery. They all seem interesting and I will do some research which works with different llms. On the matter of the purpose: I think the target group will be people with little design skills that want something quickly designed that either works as it comes out of chat3d or as a starting point in their cad tool. I don’t think that professional cad designers will be a good fit but I can imagine that non designers could use it to describe something they like to get designed from a professional and use this tool as a prototyping tool to better be able to show the pro what they want. Finally I had someone from the gaming industry comment that they could absolutely imagine to use the results as placeholder assets for 3d game prototyping. The tool could be instructed to follow a certain style guide to create multiple game assets in an automated or manually supervised fashion. Let’s stay in touch if you’re ok. Why don’t you ping me on discord so I have your contact?
Feedback - Have the program label everything for an example make a box it would label the top of the box the bottom of the box and all sides of the boxes so if you say change the top of the box to Red it would know where the top of the box is and it would change only that to Red same thing for a cat have the model or the program label all the parts of the cat like the teeth label the eyes label the nose label the lips label the neck label the feet label the fingernails label for everything so then you can say change the fingernails make them longer it knows what the fingernails are and it would only change the fingernails and make them longer, and maybe you can feed the models the 3D models back into the program and have it imagine what it would be so feed it the model of for an example your elephant and then have stable diffusion imagine using the geometry to create new elephants
To be honest it is interesting very impressive, but needs a massive update to be really useful. Don't get me wrong this is something that could solve problems for some. It probably finds a niche in e.g. the storage solution sector, where you have a fixed mount and need to design something specific in combination or so. I personally often design replacement parts based on the original part in FreeCAD and start with a photo to model on or to create an SVG shape from it to use as a sketch. These parts cannot be described in text without any hassle. OpenSCAD needs everything broken down in basic shapes and bringing these shapes and text input together is not going to work. For creating a vase, a storage box, decorative items or other non real world random designs text to 3d may work, but I don't see a way to describe more complex models than that to get a specific result. Other than photogrammetrie with a thousand pictures and a point cloud I think the best way to combine AI and 3D is using just a few photos. Give the AI five to ten photos of an object and it can figure out what to design.
Even with the early prototype capabilities. I can see Chat3D creating simple but useful functional prints similar to the wall mounted DVD shelves I made, the angle bracket supports I made so I can stack 20 Plano 3700 organizer boxes between two studs in my unfinished basement to organize my fasteners and hardware, etc. Even now, in its fledgling form, it already looks useful for simple functional parts, and it'll get better very quickly.
@ that’s basically what I said. Now take a tool from your work desk and let the AI create a mount for it. This will not work, because you need to describe the shape of the object you want to create. It is more complex than a box with a mounting solution or a handle.
Exactly what I said in the video. You can do all that manually step by step. The point was not that but that it’s easier and at some point going to be way better due to the increasing amounts of training data over time
Where can I switch of this stupid UA-cam AI audio translation of this video? Looks like the AI is to stupid to think about people being able to listen to multiple languages...
What do you think about this app? Let me know! If you like to be one of the first testers, hit me up on discord!
Chat3D is a cool idea. I've seen comments on other videos showing AI created 3D printable objects with criticism of the process, preferring to use CAD rather than going back and forth trying to describe to an AI what you want it to do, and I can see how that would be frustrating to someone with good CAD skills but it misses the point. It was easier and safer to walk than fly in the Wright Flyer but if everyone walked we'd never have aviation. I look forward to seeing where Chat3D is going.
I played with OpenSCAD years ago and was impressed with how quickly object complexity was created and how it facilitated creativity, particularly iterating geometric designs. That's a good way to create visually complex vases, flower pots, lamp shades, abstract sculptures. etc. It was so fun that I'm surprised someone hasn't created a website that's an OpenSCAD front end that allows people to create their own lamps or vases that they can then have 3D printed and shipped to them. In a world of mass produced products, I think people would appreciate having a truly unique item they designed themselves.
Haces muy bien en traducir tus videos al espapañol con IA , esto te abrira tu contenido a un publico de habla hispana muy amplio 😊
Suerte
This is fire dude, great job, this works like a final product, if this is implemented into consumer cheap printer that would lower the barrier to entry to insane levels
I know people love downloading models and 3D printing them but that was never my cup of tea. We're starting to see AI tools to create meshed object models. "Make a pig with a straw hat, sitting on a rock." Kinda cool but still not my cup-o-tea. An AI that allows functional parts to be created from a description would make 3D printers useful for regular people who want to make the same type of functional parts I make, but who don't want to learn CAD. This is a much larger market.
@@Liberty4Ever Absolutely
Very cool! Features I would like and would pay for: uploading an STL (technical parts) or step and getting a parametric model - since I do a lot of reverse modeling.
you can open source it, and let us use with our own api key. if you don't someone else would replicate it.
- Thx for sharing.
- First, wishing you success in your endeavor! You have clearly put a lot of effort into things.
- BTW, I'm a retired 3D guy (started in 3D 30+ years ago). Have explored many aspects of 3D, including writing custom code professionally. Also, I have a significant background in mathematics, plus a formal degree in engineering. Oh and, I have been keeping up with the many recent developments in AI (including the math/algorithms).
- In addition to a personal project for your learning/growth, what is the goal for the end use? What are the intended use cases? Who are the target users? What are the benefits of this "system" of modeling versus others in its class?
- Now, to give proper feedback, and constructive criticism, the intent of the project must be known, which at this time I do not (just a hint, but not enough details). Anyhow...
- As for CSG, it's good at what it does. It has its capabilities, but also its limitations.
- My current focus is 3D CAD - specifically, for design, modeling, manufacturing (subtractive, and additive).
--- I have used OpenSCAD, and Tinkercad - both CSG-based, but they are generally not well-suited for proper 3D CAD.
--- As for OpenSCAD...
----- There are "enhanced" forks (perhaps you are already aware). Does/could Claude support such? One might offer some advantage. Hm?
----- As for export formats, a mesh is fine for 3D printing - but as you noted, a .csg can be exported, then imported into FreeCAD (FC), then exported as '3D solid' (STEP). I will note though (and you may already be aware), I've had limited success with this: FC does not parse all trees correctly.
- OK, hopefully there's been some value in my comments. But as with any feedback, you can take it or leave it, in part or in total.
- And again, wishing you success; sounds like an interesting/challenging project...
thanks for your thoughts! I am aware of the OpenScad limitations and for sure I am looking into alternatives for comparison like directly using freecad scripting, replicad or cadquery. They all seem interesting and I will do some research which works with different llms. On the matter of the purpose: I think the target group will be people with little design skills that want something quickly designed that either works as it comes out of chat3d or as a starting point in their cad tool. I don’t think that professional cad designers will be a good fit but I can imagine that non designers could use it to describe something they like to get designed from a professional and use this tool as a prototyping tool to better be able to show the pro what they want. Finally I had someone from the gaming industry comment that they could absolutely imagine to use the results as placeholder assets for 3d game prototyping. The tool could be instructed to follow a certain style guide to create multiple game assets in an automated or manually supervised fashion. Let’s stay in touch if you’re ok. Why don’t you ping me on discord so I have your contact?
I wanna try it 🤔
Feedback - Have the program label everything for an example make a box it would label the top of the box the bottom of the box and all sides of the boxes so if you say change the top of the box to Red it would know where the top of the box is and it would change only that to Red same thing for a cat have the model or the program label all the parts of the cat like the teeth label the eyes label the nose label the lips label the neck label the feet label the fingernails label for everything so then you can say change the fingernails make them longer it knows what the fingernails are and it would only change the fingernails and make them longer, and maybe you can feed the models the 3D models back into the program and have it imagine what it would be so feed it the model of for an example your elephant and then have stable diffusion imagine using the geometry to create new elephants
This produces 3d object litter which is already floating all over the internet. Same applies also to AI generated music.
I am interested!
To be honest it is interesting very impressive, but needs a massive update to be really useful. Don't get me wrong this is something that could solve problems for some. It probably finds a niche in e.g. the storage solution sector, where you have a fixed mount and need to design something specific in combination or so. I personally often design replacement parts based on the original part in FreeCAD and start with a photo to model on or to create an SVG shape from it to use as a sketch. These parts cannot be described in text without any hassle. OpenSCAD needs everything broken down in basic shapes and bringing these shapes and text input together is not going to work.
For creating a vase, a storage box, decorative items or other non real world random designs text to 3d may work, but I don't see a way to describe more complex models than that to get a specific result. Other than photogrammetrie with a thousand pictures and a point cloud I think the best way to combine AI and 3D is using just a few photos. Give the AI five to ten photos of an object and it can figure out what to design.
Even with the early prototype capabilities. I can see Chat3D creating simple but useful functional prints similar to the wall mounted DVD shelves I made, the angle bracket supports I made so I can stack 20 Plano 3700 organizer boxes between two studs in my unfinished basement to organize my fasteners and hardware, etc. Even now, in its fledgling form, it already looks useful for simple functional parts, and it'll get better very quickly.
@ that’s basically what I said. Now take a tool from your work desk and let the AI create a mount for it. This will not work, because you need to describe the shape of the object you want to create. It is more complex than a box with a mounting solution or a handle.
You can already have chatgpt create an openscad script, I'm not seeing anything new here. Am I missing something?
Exactly what I said in the video. You can do all that manually step by step. The point was not that but that it’s easier and at some point going to be way better due to the increasing amounts of training data over time
Where can I switch of this stupid UA-cam AI audio translation of this video? Looks like the AI is to stupid to think about people being able to listen to multiple languages...
It's an option in the player (the gear icon), there is an audio track selector
French audio translation is so boring …. IA is useless
Oh did this auto translate? It is then the first time ever that it happened. 😂