ComfyUI Node
🌗 Moon Anima Regional Patcher (Advanced)
A ComfyUI node in MoonNodes with 13 inputs and 3 outputs.
🌗 Moon Anima Regional Patcher (Advanced)
- model
- clip
- mask_list
- patched_model
- POSITIVE
- NEGATIVE
◄positive_textduo, outdoors, oil painting
BREAK
anthro lynx
BREAK
anthro tiger►
◄negative_textlow quality, blurry, deformed►
◄prompt_modebase_only►
◄conditioning_dropoff1.00►
◄start_percent0.00►
◄end_percent1.00►
◄base_strength0.80►
◄cross_mask_strength1.00►
◄self_mask_strength0.00►
◄base_ratio0.00►
CategoryMoonNodes
Inputs (13)
| Name | Type | Default | Description |
|---|---|---|---|
| model | MODEL | The Anima diffusion model to patch with regional conditioning. | |
| clip | CLIP | The CLIP / Qwen text encoder used to tokenize and encode prompts. | |
| mask_list | MASK | List of spatial masks corresponding to regional prompt zones (Zone 0, Zone 1, ...). | |
| positive_text | STRING | duo, outdoors, oil painting BREAK anthro lynx BREAK anthro tiger | Multi-line prompt. Use 'BREAK' on its own line or between phrases to separate the base prompt from regional prompts. |
| negative_text | STRING | low quality, blurry, deformed | Negative prompt applied across the generation. |
| prompt_mode | COMBO | base_only | - base_only: passes only prompt 0 (base) to POSITIVE output. - concat_text: replaces BREAK with a newline for a single natural language encoding pass. - concat_conditioning: concatenates individual conditioning tensors. - merge_average: blends/averages all conditionings into a single embedding. - comfy_area_conditioning: outputs standard ComfyUI area-conditioning list with spatial masks attached. |
| conditioning_dropoff | FLOAT | 1.000–1 | At what point in generation (0.0 to 1.0) the prompt switches to 'base_only'. Default 1.0 uses the selected mode all the way through. Setting to e.g. 0.40–0.60 uses concatenated/merged conditioning for early poses/composition, then drops off to clean base conditioning to eliminate noise and refine sharp details. |
| start_percent | FLOAT | 0.000–1 | Sampling percentage (0.0–1.0) when regional conditioning starts being applied. |
| end_percent | FLOAT | 1.000–1 | Sampling percentage (0.0–1.0) when regional conditioning stops being applied. |
| base_strength | FLOAT | 0.800–1 | How much the global base prompt affects regional zones. Higher values blend more shared style/lighting into regions; lower values isolate the regional prompt. |
| cross_mask_strength | FLOAT | 1.000–1 | Cross-attention isolation between distinct regions (Zone A vs Zone B). 1.0 blocks bleed-through; lower values allow soft cross-regional influence. |
| self_mask_strength | FLOAT | 0.000–1 | Spatial self-attention isolation between regions. 0.0 maintains global scene coherence (shared lighting, perspective); higher values isolate spatial patches. |
| base_ratio | FLOAT | 0.000–1 | Direct blend ratio with the un-partitioned base generation. 0.0 is pure regional output. Values > 0 blend in the global base image (runs the model twice per step). |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| patched_model | MODEL | — |
| POSITIVE | CONDITIONING | — |
| NEGATIVE | CONDITIONING | — |