Sooo... this looked better in my imagination, lol.

The outfit change itself actually works surprisingly well. Just the hair is giving me problems. (Click above to watch the video that randomly cycles through outfits and hairstyles.)

It's easier to tell AI image generation "recolor this outfit" within a mask, than to do the same with hair. AI is great at creating new outfits, but terrible at creating new hair styles to spec.

Given that this is just a proof-of-concept for now, I did not spend the additional hours for each haircut to carefully make the background transparent (which would've reduced the "paper cut" effect that remains in many of the hair styles.)

Other reasons why the end result video looks a bit janky... that have nothing to do with the texture change process...

Loading a new outfit takes at least 2-3 seconds on my computer, which leads to a break in the video. And worse... the lipsyncing seems briefly suspended.

So I had to play the same animation several times, each time with multiple random outfit changes, then cut it together, hoping to omit all the gaps and as much of the non-lip-synced parts as possible.

All this took way more time than anticipated.

I continue to work on the process as I have time... and of course AI image models get better every single week, so the end result will look better and better with time.