Accidentally caught some folks off guard at the OMI Developer hour today with the default model in this project 😅 I'd suggest having a less risque default model and default clothing choices.

The project itself is looking great though! Nice work!

Right now the export contains multiple skeletons. We really want all of our assets in the final export to be assigned to a single skeleton, and for the final file to only have one skeleton. This will hugely reduce file size.

Can try to port to wasm if wanted.

Randomize is a common feature across character creation programs, it'd be worth exploring a way to do this in avatar creator

Lots of dead code, and state is currently a bit messy. For a production release we need to go through and make sure we've cleared any dead files, dead code, consolidated everything as tightly as possible.

Currently character is in T pose. Should be in an idle pose and animated. We can get free poses from ActorCore, hopefully we don't need a whole retargeting thing.

This can be very complicated. If we have animation matching the character (which is rigged using Reallusion skeleton) then this is not so hard, but if we don't we need a retargeting solution.

I think that the best approach is to integrate Ossos and use the retargeting example provided here:
https://github.com/sketchpunklabs/ossos/blob/main/examples/threejs/003_retarget_and_springs.html

We might need to make a new retargeting profile for Character Creator character, since it is different than Mixamo rig which is generally standard.

VRM support was merged into meshoptimizer https://github.com/infosia/meshoptimizer/tree/VRM

We want to add this and support it. Making out files smaller and more performant is important!

Currently we are handling metadata inside the mint function. This isn't extensible or easy to change, so we should make it a separate JSON config

We also need to make sure the that if we are using IC we conform to the popular metadata conventions, separate from ethereum

When the characters were loaded in as GLB they worked well. We've imported them as VRM and they were facing backwards. After fixing that, there is still an issue where the models are more forward than they need to be. They should behave the same as the base character.

We have removed .scss / .css to simplify the codebase and make a React-only component. However, a few things got broken and don't look great, and need to be fixed.

Right now our base assets are hosted on AWS. We should host them on the IC, this could be as simple as hosting them inside of the canister that holds our app, but we don't want to commit these assets to Git

Current the vrm base body assets are 48 megabytes. They can be much, much smaller with a bit of optimization.

-- Make sure that they are a single combined material
-- The head needs to be a separate mesh, and the blendshapes need to be removed from the body, this will reduce file size by 60%

Right now we are holding morph target data to morph the clothing to fit each body.

When we export the model, we want to make sure that the morph targets are baked to final position and removed-- they add massive amounts of data to the final file.

We want to keep blendshapes on the face, so we are just removing these additional modifier morphs that the character creator uses. The only morph targets remaining should be on the face, and they should only be on the head.

If we could add any helper shell scripts and cmd line stuff that newbs can paste in, that will be helpful

Currently we are using DFINITY stuff but we want this to be totally optional. We don't want users to need to dfx deploy before launching.

• Add .env var for USE_DFINITY
• Dynamically import dfx/dfinity/etc into js code
• Test with and without env var

https://github.com/oguzeroglu/TextureMerger

We will need to think about things like--

• Making sure all of our textures fit on one atlas and there is no wasted space
• Making sure more important parts get more space, or is UV proportionate
• Transparency objects are separated into a separate material

End ideal would be a single material and texture set (diffuse, normals, other PBR if contains)

Right now we have a light <-> dark slider

Doesn't look great

As @fire suggested, a palette of nice skin tones would probably be better (ReadyPlayerMe does this)

Right now we have the body selection on the top left.. it's weird.

We want to add it here:

Character isn't tpose exported? It's weird. We need to make sure she doesn't have a pose applied or something, and is standing on the ground flat.