This Software Can Generate 3D Models and Scenes

what if I told you that all of these assets are not modeled but why they're created using AI That's Right everything you see here from the table to the TV couch books dish plant and other assets were generated using a text prompt and I'm going to show you how you can do that too so the way you can use this AI is by providing an image input you can do so in one or two ways first and the most simple one is to upload one from your computer you can actually upload multiple images if you want to diffuse different objects together and the second method is to use the integrated image generator to generate an image from a text prompt and then use that as an input to generate your model one thing that caught my eye when using this tool is that it uses quads when generating geometry which really sets it apart from the competition the name of this service is Roden Ai and today we're going to talk about it and see what you can do with it I know that Ai and AI tools and services are leaving a sour taste in every artist's mouth right now mine included but trust me it will be a long time before AI can replace artists even though the tool we are talking about today is decent I don't think it will be replacing modeling or texturing anytime soon usually whatever you here online or in the news is blown up for clicks but realistically this is a far cry from a production ready tool that's not to say it won't be helpful for you as a 3D artist anyways do you remember hyper human we talked about their service called chat Avatar a while back which is an awesome tool to generate believable faces using only a picture generally speaking the generated faces using this tool are highly detailed textured and ready to be animated but it generates faces only which is kind of a limitation Rodin on the other hand is their latest AI tool similar to Triple AI CSM and many others and this one is a generative model for creating high quality 3D assets with a decent quality because you're going to see in a bit if this tool actually manages to produce good results it is also important to note that this is still in beta testing so things may change in the future the first thing is what most people are probably used to is image generation like mid Journey stable diffusion and d e so in an nutshell you provide a PR describing what you want and then the AI through various methods will try to render the result to the best of its ability the results will vary even on the same AI tool some of them generate stylized art better others deal with realistic subjects and lighting better and some are better with cartoony stuff and so on and so forth and I think this is common knowledge by now another AI type that has been creeping up lately is the 2D to 3D generative AI what this one does is that it takes a 2D image as an input and turns it into a 3D model there are many of them right now and there is Roden AI built on Clay a large scale generative model which resembles a text model rather than a 2d to 3D model more like diffusion in 3D if you will because that's how it works trained on thousands of 3D models data sets it denoises 3D data in compress late in space but what does that mean this is a good question but who cares the only thing interesting here to note is that it is a new approach which is computationally way more efficient the latest Road in AI from hyper human I think is something that has a lot of potential to start creating assets for example had to hyper human. deos.com register for every account and loging click on the Roden Tab and now you can see the text input field which dumbles as an image uploader to upload your images whichever way you choose to proceed it won't affect the generation process but if you choose to upload your own image make sure the image has one subject which is a monocolor background here I will use a prompt to describe something say like old books medieval stacked and click generate if you like the result you can hit the green check button if not you can refresh a new image until you find something that you like now hit generate after upload a geometry generation window will appear with more parameters and you can somewhat guide the generation process choose a quality setting and even redo the mass generation as many times as you want till you get something that you like for instance you can highlight parameters like whether the object is symmetrical or not and whether you want sharp edges or smooth edges and so on and so forth also if you have multiple images you can pick between each image and how much weight it can contribute to the generation process most of these you don't really need to change I'm just listing them so you know what you're looking at the default settings are probably enough for most users and all you need to know is confirm and redo once you're satisfied with the result hit confirm and your mesh will be generated next you will apply a texture from an image like in the previous step and you can confirm the selection or redo the texturing once everything is set and done you can pack your model by choosing only the base model multiple LOD or high poly version you can also choose between multiple formats and texture quality of up to 4K and that's it this is a generation workflow you can check some of the featured models generated using this tool and it looks really promising I think it would be very useful for people who are trying to create somewhat of a stylized style or content that doesn't require super close-ups or even animations for YouTube like storytelling videos and so on it can also be used for concepting and quick ideation and the fact is all these models are generated using quads instead of triangles is a big Pro for me and a big stop in AI development so if you want to check those two and try it for yourself you can click the link in the description down below I hope you guys found this video useful and informative if you did please give it a thumbs up also please subscribe to this channel thank you guys very much for watching and I will see you in the next one

Channel: InspirationTuts

Views: 10,713

Keywords: AI-generated 3D assets, HyperHuman, RodinAI, AI modeling, text prompt 3D generation, AI image input, 3D geometry generation, AI texture application, generative AI, 2D to 3D AI, AI in 3D art, HyperHuman ChatAvatar, AI tools for 3D artists, CLAY generative model, 3D model denoising, AI-generated models, 3D asset creation, 3D texturing, AI in animation, AI in concept art, AI-generated quads, AI-driven workflows, future of AI in 3D, AI-generated high-quality 3D assets.

Id: uzgaNuFqhWU

Length: 6min 37sec (397 seconds)

Published: Thu Jun 06 2024