CLIP Skip Configuration for Model Optimization

{{YOUR_PROMPT}}, high quality, detailed

CLIP Skip Configuration Guide:

CLIP Skip 1 (Default):
- Uses all CLIP text encoder layers
- Best for: SD 1.5 base models, realistic models
- Characteristics: Strongest prompt adherence, detailed interpretation

CLIP Skip 2:
- Skips last CLIP layer
- Best for: Anime models (NAI-based), stylized models
- Characteristics: More artistic freedom, anime models trained this way

CLIP Skip 3-4:
- Rarely used, very loose prompt following
- Best for: Experimental, abstract art
- Characteristics: High creativity, may ignore prompt details

Model-Specific Recommendations:
- Realistic Vision, SD 1.5: CLIP Skip 1
- Anything v5, Counterfeit: CLIP Skip 2
- NovelAI models: CLIP Skip 2
- SDXL models: CLIP Skip 1 (different architecture)

Settings Location:
- Automatic1111: Settings > Stable Diffusion > Clip Skip
- Or use --clip-skip flag
- Can be set per-generation in some UIs

Negative prompt: {{STANDARD_NEGATIVES}}