Da F*ck we were not read, text to 3d are good now
A week ago i made a video about how polygon modeling was dying, and my point was about how you rarely have to model anything from scratch now days, unless if you want a custom or stylized model.
but if you want realistic characters, applications like meta humans can generate customizable human characters for unreal engine or addons like human generator and fashionista can do the same for blender, all these give you customization tools to change the character into what you want.
and if you want something for the background you can use libraries like render people to get scanned and animated models to use, they also give you some free ones to test out, the quality is similar to what you get with the procedural crowd addon for blender.
another point i made was how procedural generation was also reducing the need for polygon modeling in that you can use these procedural generators for non character models like plants, buildings, roads, rocks and more, there generators can make those things with just a few clicks and get hundreds of variations of what you want.
and for everything else that needs to be hand modeled but not going to be the hero or star of your movie, like background characters for cartoon animation, there artists on sites like cgtrader.com and turbo squid who have made hundreds of characters in different styles that are already rigged and some even come with animation, these are packages sold with 100s of models, some are entire villages with house, trees, characters, cars and everything you can think, all premade and ready to go for just a few 100s of dollars.
on that video, there were some counter arguments made in the comments, most of which i agree with like this,
but that was before this Ai generator called ProlificDreamer a high-fidelity and diverse text to 3d generator was released.
AI 3d Modeling
this Ai is capable of generating a wide range of models and texturing them.
when it comes to modeling, the AI starts with an approximation of the model and progressively adds detail and polygons as it goes through its training data.
it can also understand how to deal with transparent and translucent parts of the model which other techniques like photogrammetry or 3d scanning struggle with. it does not only stop at generating the mesh, it also does texture generation.
Ai Texture Generation
this Ai is built on top of text to image models like stable diffusion thats why its so good at texture detail generation.
when it comes to texturing the model, first the Ai produces the 3d models, and then it generates the uvs that it progressively textures to make a high detailed texture that is mapped onto the model.
these tools may not be able to generate stylized characters yet, but i imagine given the right training data it may be possible.
in the future production studios like marvel or Disney could hire artists to design or model a set of stylized characters and then the Ai can use that to generate more in the same style.
what is left would be rigging and animating, this may sound discouraging to new and senior artists and it is discouraging but pretending its not happening would make it even a bigger problem than it is.
on top of that we are starting to see early generations of text to video Ai generators, and here are some samples
i dont think there are alot of artists prepared for this, but i see alot of opportunities on the horizon, one of the samples shows a bunch of explosions with smoke and debri, there is not alot of consistence between each frame but this is still early stages of the Ai.
Imagine a time where you dont have to worry about smoke simulation and rendering anymore, this would save you alot of time as an artist and let you focus on developing the important parts of your story