Krea2:
This is the first Krea2 version of my LoRA. It's still far from perfect and needs more finetuning.
I'm currently working on a new training set. A new version will follow in some weeks.
My current workflow is included in the images. I'm still trying new things. You can change the denoise and/or the blur to get better/worse/other quality and results. Also try different checkpoints.
Qwen (1.0, 1.1):
Since there is currently no LoRa available that is suitable for muscular or very fit women, I have now trained one myself.
It is my first LoRa ever. Accordingly, it is not outstanding, but it is also not unusable. For me, it serves its purpose. And perhaps for others as well.
It works with both Qwen 2512 and older Qwen versions. I can't say whether it also works with Qwen Edit.
To get the best possible results, I recommend a skin enhancement LoRa (Skin Fix, Qwen-edit-skin) as well. I use both together. For 4k upscaling, I recommend a Pony model (e.g., Cyberrealistic).
I will post more examples and a workflow soon.
