AI retopology and surface projection for Posing

An AI-assisted posing tool integrated with Nomad that allows users to define joints and body regions from a menu. When a limb is rotated, the AI automatically corrects surface deformation by regenerating adjacent anatomy (e.g., deltoids, pecs, lats) in a structurally coherent way. It would dynamically refine topology on affected areas during posing, then, once the pose is finalized, convert the modified surfaces into clean quad topology and project the result back onto the base mesh.

3 Likes

This idea looks very good!

1 Like

Thank you, if only I have ability to write down these updates myself :slight_smile:

Overall I get the idea it sounds good, but I think I would skip the AI part and wait for actual rigging and weight paint, that would be a big jump forward in posing, it would probably satisfy 95% of the audience here.

I envisioned AI as a topology tool. Initially, you sculpt the character, then define the head and joints using the anatomy definer tool. Subsequently, AI will remesh the object with the appropriate topology, enabling you to move and pose the character. Once you pose the character, AI will also adjust the volumes. For instance, if you rotate the arms in a close position, and you’ve tagged the joint sections and limbs as forearms, hands, shoulders, etc., the AI volume tool will update the volumes of the arms accordingly. So, all I need to concentrate is how character looks and after that in which way to pose it.

I guess Nomad and AI will never come together :joy:

1 Like

@TheManFromKrypton Of course I do understand what You meant, You are thinking about something similar to Mixamo but with AI aid. And it would be somewhat cool, but You have to realize that:

  1. Integrating AI is a lot of overhead in terms of development code model integration with MCP etc. And I do know what I am saying. It would be imensly problematic in many ways. IT would make the app bigger heavier, an more reliant on external web clound and the mercy of big tech.
  2. We all need riging and weight paint anyway at some point in evolution of this app, it should come as a feature first AND as a standalone feature, that does not require connection to web Cloud etc. those 2 things are far less expensive to apply.
  3. People these days, many people (including me although I work with AI on a daily basis) are not happy about the fact that everyone are trying to solve every problem - especially in artistic space, even if it’s digital sculpting - with AI. I get it AI is just a tool but it takes some thinking process from You that I think might be worth developing anyway.
  4. If You use AI to retopology, rig model, put orientation points on the figure etc., all this would require You too upload You work to the cloud (3d files of figure), even if done in background inside the app, and here many…many people are not ok with that, since we all know, this will be used for model training by big corp. And I don’t know how You view this but I’m pretty adamant towards this process.

If You wan’t AI topology tool I’m sure there is a dedicated service online somewhere that does it it will be essentially few steps more perhaps 3-4min overhead in comparison to the whole process, why not just leave it there for people who wan’t to use AI?

haha, probably, but just as good old James Bond said: never say never.

yeah well I agree on cumbersome nature of getting ai dependent. Actually there are other ways to achieve what I suggest, internal solutions. Perhaps being able to retopo with predefined surface topology. I simply define eyes, mouth, ears etc, or fingers, articulations with predefined types and Nomad will give you projected topology for that certain area, then you simply move it around or use move tool to make it fit better there and once you place all those predefined topology sections, you hit guided retopo, it will do the remesh accordingly. After that all you need to do is minor adjustments, if needed.

I can agree with that, just rule out AI from the equation I think is a good idea.

1 Like

Thanks, I mean in 2026 it hurts me to see we still need to do topology and rigging by hand rather than simply defining the articulation and shapes!