Krea usage tips
Use the raw model with the turbo lora applied to it at 0.6 weight. This looks much better than turbo by itself. Use about 12 steps instead of 8. For 1 CFG use euler, beta / beta57 scheduler. (feel free to find a better balance of lora weight vs steps.) https://huggingface.co/Comfy-Org/Krea-2/blob/main/loras/krea2_turbo_lora_rank_64_bf16.safetensors
Use this vae: https://huggingface.co/spacepxl/Wan2.1-VAE-upscale2x its better quality. It needs either kijai's diffusers WF or this: https://github.com/spacepxl/ComfyUI-VAE-Utils Still playing with other merges as well. (merge qwen and the wan vae tunes at like 0.5)
Use the uncensor Lora: https://civitai.red/models/2728234/krea2filterbypass?modelVersionId=3067151 Comparison of loras vs node (lora is better): https://files.catbox.moe/9zpoqs.png It also helps with expressions and other prompt responsiveness. The censorship hurt more than just nudity.
It can use reference images: https://github.com/ethanfel/ComfyUI-Krea2TextEncoder
And be more specific with your prompts. For example. "Teen Titans" gives you crappy Teen Titans Go. But "Teen Titans (2003)" gives you the good version. Same for other series / characters / people. Try changing your capitalization / specifying the series / sometimes the year.
It knows TONS of artists by name, a undersung feature. Use "In the style of" or "Painted by", ect...
For mixing characters use "Cosplaying as" instead of "Wearing X's outfit" or try something like "With the body of X and the head of X in the style of X..." https://files.catbox.moe/sveyky.png Be specific.
For even better realism try this lora: https://civitai.red/models/2727284/realism-enhancer-krea2?modelVersionId=3065628 For more uncensored stuff try: https://civitai.red/models/2688234/realism-engine-ideogram-4-krea-2?modelVersionId=3067451
https://www.reddit.com/r/StableDiffusion/comments/1ueybt8/some_important_krea_usage_tips_ive_found_not_seen/