Update 3.0!
Trained using a brand new high quality dataset captioned using natural language for Flux Dev. This is an experimental first release, and additional improvements will be added over the coming days. I have an updated version training at the moment.
Please enable notifications on the top-right to see when updates are released :)
Use keywords "fca_style" and "anime".
Longer training time at a lower learning rate with more images, and also mixing in some regularisation images. Please note the new activation keywords of "Flat colour anime style image showing".
Flux Schnell is faster than Flux Dev but the outputs are lower quality, so expect some issues with hands and text. For best results use the v3.4 Flux Dev model listed at the top.
Go ahead and upload yours!
Your query returned no results β please try removing some filters or trying a different term.