Tencent has just released FLUX SPRO, which aligns diffusion models with fine-grained human preferences. It enables faster training (taking less than 10 minutes for FLUX.1.dev) through direct trajectory alignment, avoids reward hacking, and supports controllable fine-tuning.

I created this workflow simply to figure out which one—FLUX Dev, Krea, or SPRO—better matches my aesthetic preferences. Please feel free to leave your comments and share your thoughts!

1. Pick Your Model: Generate images with FLUX.dev/FLUX Krea/FLUX SPRO—just select your preferred model from the group selector below.​

2. Customize Settings: Adjust width, height, and text prompt; you can also set model-specific parameters in their corresponding groups.

Run it directly here: https://www.runninghub.ai/post/1967123991443935234?inviteCode=rh-v1254

Sign up via my invite link to get 1000 bonus points + 100 daily points—this lets you run this workflow for several free times a day.
Sign up now: https://www.runninghub.ai/?inviteCode=rh-v1254