What’s wild is that it was probably not even 2 years ago that I was all excited that ChatGPT was able to generate the coordinates of a 3D cube in space.
Now we’re at 3D models from text prompts and image uploads.
This is an opossum. I used Gemini (which uses Imagen for image generation) to create the line drawings (“b&w line drawing of an opossum body in profile”) and this demo of Tripo3d to create a 3D model of the opossum based on the drawing. Having done this a few times now it seems the AI 3D model generators function better with simple inputs. So, after each new drawing that Gemini spat out, I asked it to simplify the next one. The #3 drawing is the one used to generate the model. The mesh is 100,000 faces and obviously very tessellated but retopology tools in Max work decent enough to get a nice quadflow for my purposes.
I see that Tripo3D also now does AI generated character rigging on 3D models. I honestly don’t know. Reject AI completely or use it for the tedious stuff? But what’s tedious to me is not tedious to the absolutely crazy person who loves to rig and skin characters, but who hates painting textures. It’s not ok for this person to use AI to paint textures but it’s ok for me to use it to generate a character rig? Or a 3D model? What’s going on here?