hongfz16/AvatarCLIP ? reverse-engineered prompt

Build me the AvatarCLIP project so I can generate and animate a 3D avatar from a text prompt.

I want to be able to type something like “a tall skinny female soldier arguing” and get a believable 3D character with the right body shape, appearance, and motion, then export it in a format I can open in Blender or Unity. Please wire up the full flow for avatar generation, animation, and FBX conversion, and make the setup as easy as possible for someone following the README.

If anything depends on extra models or data, add clear instructions for downloading and placing them in the right folders, and make sure the scripts run without confusing manual fixes. If you need to check current install or dependency docs online, do that too.

Also add a simple way to preview results and verify the pipeline is working end to end.