An illustrative model mix that incorporates my Noiseform LoRA and other custom LoRAs to produce something that can do it all with a pleasant gradient aesthetic.
Recommended Settings:
Steps: 25-30
CFG Scale: 7
Sampler: DPM++ 2M Karras
Use Highres.Fix
Utilize negative embeddings such as EasyNegative, bad_prompt, and occasionally badhandv4.
Additional Notes:
Beta model: Changed the formula for this mix. Increased consistency, prompt accuracy, color and contrast.
Alpha model: There is still room for improvement in its capabilities, including the NoiseForm effect and overall consistency. I'm releasing it just to see what it can do.
You can achieve satisfactory results with simpler prompts, but you may need to add emphasis, such as (yellow shirt:1.2), to ensure colors are interpreted more accurately.