Hello everyone,
I've recently been experimenting with a new checkpoint (model) for image generation! This one is designed to handle backgrounds, small details, and hands (my eternal struggle) much better. It’s also supposed to follow more complex prompts more accurately, which opens up a lot of exciting possibilities.
I’ve been trying to get it to match the style I used before. A friend even made a custom LoRA (add-on) in the game’s style to help with that, and honestly, it’s getting really close. Still, I feel like something’s missing. The faces and colors from the old model just had a certain feel to them that this one hasn’t quite nailed yet.
Original:
New:

Let me know what you think :)
Fabien
2025-05-16 18:17:08 +0000 UTCJay Speelman
2025-05-16 15:26:30 +0000 UTC