V.4.0
*insert long-winded description here* Also, pineapples are now oversized resins or whatever those things are, bad economy.
V.3.0
By far the most convoluted training i have ever done. From to artificial to low resolution to illustration ( don't even know how that happened ) and the list goes on.
Everything that matters is that i have my little toy to offset my other loras if they don't behave.
The only thing that is still open are more nature photos, landscapes and most important stupid loras.
More time to play with other peoples models.
V2.0
A slightly different approach. Merged smaller lora batches into the checkpoint and extracted a lora from this. A lot better to manage, since with so many pictures and different themes it takes to long to convervege. With less pictures it's faster and easier to determine when it's enough training.
Very happy with the result. It takes the edge of pictures that in some cases still look to artificial or where it looks like a character got photoshopped in.
I would keep the strength low ( around 0.7 ish ) for anything people or animal related, because it can fuck up anatomy and chroma is good enough to handle the rest.
Even though Chroma is more than capable of producing this stuff by itself, needed to balance the fuckups i usually train and for combining loras with each other. Not finished yet, but will most likely pick it up in between to continue training till it converges, which by my calculation will take around 5 million years, roughly.
Kinda works with normal Flux models but reduced influence, at least in Comfy, no guarantee for your personal setup and version.
Trained on version HD rev 0.1
And still no category for Chroma...
Description
Different approach and more pictures.
FAQ
Comments (13)
Chroma now has its own model category on Civitai.
https://civitai.com/search/models?baseModel=Chroma&sortBy=models_v9
images in showcase looks craaazy wild, been trying and in fact sadly seems to presents some body deformities unless I keep it at 7 or just 8,but the quaility is still impressive, btw, did you notice any difference with clip skip 1 and 2 ?
Never tried clip skip in Comfy, so i don't know if there's a difference.
Nice👍🏻
Looking forward to an anatomical fix for lora. Especially hand fix
Combat it with DPMPP 2M scheduler and bong_tangent sampler. 9/10 works well.
Try adding these in the positive and negative prompts:
Positive: anatomically correct hands. normal hands.
Negative: broken fingers. 6 fingers. six fingers. 4 fingers. four fingers. supernumerary fingers. anatomically inaccurate. body horror. mutated body. disfigured.
@mmdd2543 I've since found that this is mostly a symptom of an undercooked image, it needs more steps. Unless the prompt is just straight up causing inferrence of multiple hands or something weird.
@Diecron312 I haven't tried this lora but I have used 250 steps with Chroma and Chroma Radiance and still had bad anatomy frequently, including hands. Wan 2.2 gives me better results than Chroma but it has a lot of limitations, even with loras.
@ss9999 It's been a while, but the main thing that can help with Chroma is using a better sampler (implicit is best). But it gets slow as hell.