I know its been a while, but I'm happy to complete my first Krea 2 lora based on my charcoal Dall-E dataset I made when I did both SDXL and Flex lora models.

I gave this lora a good play and it turned up some pretty good results that I'm happy with.

To use this lora the trigger word is cha0al, I wasn't sure if I needed a trigger word or not as many different workflows say different things about it.

The strength works well from 0.3 to 0.8. But have a play around with it as sometimes the look isn't really their. But I did notice.

• 0.8 - 1.0: Best for close-up portraits, character shots, and simple scenes.

• 0.5 - 0.75: Best for wide-angle camera shots, complex sci-fi/fantasy backgrounds, and detailed environments.

For CFG strengths, the lora was trained on 3.5. But I have noticed that sometimes using CFG 1 images can come out as failed to correctly render objects.

For prompting I would use something like this to get the full use out of the lora. But coming up with other charcoal prompts could also work.

((black and white charcoal drawing)), cha0al, {your subject description here}, <lora:Krea 2 Charcoal style v1:0.7>, black and white, breathtaking charcoal illustration, highly detailed, 4k, (textured paper), charcoal texture

For example, ((black and white charcoal drawing)), cha0al, wide-angle shot of a lone knight standing on a cliff edge looking over a misty forest, <lora:Krea 2 Charcoal style v1:0.65>, black and white, breathtaking charcoal illustration, highly detailed, 4k, (textured paper), charcoal texture

Let me know what you think, as this is my first Krea 2 lora using AI toolkit locally.