Spaces:

huggingchat

/

chat-ui

like 1.09k

Running

Flux dev (better quality, a little slow but number of steps can be turned down):
https://huggingface.co/spaces/black-forest-labs/FLUX.1-dev

Flux schnell (faster, only requires 4 steps):
https://huggingface.co/spaces/black-forest-labs/FLUX.1-schnell

I've provided zero space links of both models. Flux dev is much better at aesthetic and realism, midjourney level. Flux schnell is a distilled model optimised for 4 step image generation, but It's noticeably inferior to the previous one.
These both are 12b parameters models and much better than the current model which is being used in Huggingchat.

Here's a quick comparison:
(Prompt 1: a tiny astronaut hatching from an egg on the moon)
(Prompt 2: a cat holding a sign that says hello world)
(Prompt 3: A young woman taking a boring selfie with her phone in front of a mirror, typical of a 2015 Snapchat post.)

Flux dev:

Flux schnell: (1-2 attempts)

Current model: (2-4 attempts)