by tuantruong on 12/27/24, 10:50 AM with 3 comments
I noticed many iPhone users can't access the new AI emoji features because they have older devices. So I spent 48 hours building an accessible alternative using open-source images models.
The system uses a fine-tuned Flux AI model (based on https://github.com/kijai/ComfyUI-FluxTrainer) to generate custom emojis from text. It runs efficiently enough to work on a basic web interface - no device requirements or OS updates needed.
Just type what emoji you want (like "happy pizza doing backflip"), and it generates a vector emoji with transparent background in about 5 seconds. Works on any device with a web browser, including older iPhones.
iOS and android versions are coming, and would love your feedbacks on the tools. Love to hear if something else interesting you would like to see on the too,
Cheers, //TT
by maalber on 12/27/24, 1:30 PM
That being said, I like the use-case and it seems like nice work in a short amount of time. I have a couple of questions that I am curious about
1. Can you elaborate a bit more on your fine-tuning process? Did you "just" feed the model a bunch of regular emojis? Have you considered using any RLHF/DPO approaches?
2. You mention you generate a vector emoji. As far as I know the flux model just generates bitmaps, how do you handle that conversion?