[Discussion] Support for other styles

[oddlyspaced]

I've trained a flux 1 dev model with Google emoji styles. Haven't included skin tones while training for now but I do have the base model ready.

Attaching some reference generated images.

With this, I wanted to start a discussion about introducing stylistic choices in the emoji generation pipeline. Do we setup the inference in a manner where the emoji style is extracted from the prompt, or do we create a simple wrapper and setup params for input that would internally switch between models?

Lmk if you need the current model file aswell.

[oddlyspaced]

another point of discussion would be regarding how we would be handling skin tone styles between the emoji styles.

[SpyC0der77]

I found a stable diffusion model on hugging face that is trained on Microsoft's emojis. I will look into using it

[caspersimon]

@oddlyspaced That's great! See the README in /lora for instructions on how to add the Google emoji finetune to this project.

Also see advanced usage on the README in the top directory for how to switch between styles.