A comprehensive finetune of Flux.2-Klein 4B/9B, that works on distilled versions.
Art style transfer using 4 reference images:
Workflow
Upscale any anime image, prompt at 2-4MP:
Workflow
ControlNet any single subject image:
Workflow
...And many other edit tasks: Real life to anime, change clothes, prompt new pose, change expression, hairtstyle, pupils, Compose two subjects, convert to 3d model, anthropomorphize, etc.
Dataset (1500 image pairs CONTAINS NSFW/Drawn nudity)
Description
Doubled the dataset.
FAQ
Comments (18)
Will the version 4b be updated later?
Great lora! I cleaned some old screenshots using this.
The workflow links aren't working. Maybe catbox is just down? but maybe it's worth zipping them into another file to add to this lora. There's a way to add workflows here.
thanks for the heads up, ill fix it.
I've been going over your dataset, and I just have to say that this is crazy impressive. You put so much work into this. An astounding amount of work. Thank you for sharing it. I've learned some things about how to train edit loras.
What is it, exactly, if you don't mind explaining? What's an edit lora? It changes the style to anime like?
@Polygon it means you can make edits to the image like "change the hair color" "change the pose" by giving it a reference.
https://www.youtube.com/watch?v=Y9foxm9OYEU
@memnoid can this lora colorize sketch manga?Tag?
@Ovoooooo it can colorize yes, but it might mangle text
@memnoid need the prompt?or just prompt "color this image"?
How much VRAM is needed in total for 9B?
preferably 24gb, but if you can offload, 16gb takes about 1-2 min per prompt at 1 Megapixel
How does the reference chaining work in your style transfer workflow? is it a kind of hack to increase the references influence on the gen? Because it just looks like you're adding the same stitched image 4 times as reference.
yes exactly
THIS IS AMAZING!
Please let me know what you used to train it!! AI Toolkit? any tips for settings?
Can you please give an example of the controlnet edit. I'm getting not good results on the shared workflow.



