ComfyUI Node
Anima Prompt Data Conditioning
Reads EASYUSE_ANIMA_PROMPT_DATA by dict keys, encodes positive and negative CONDITIONING with CLIP, and applies comfyui-spectrum-ksampler Anima Mod Guidance to the MODEL when enabled. It also creates an empty latent image from prompt-data width and height with batch size fixed to 1. Artist mix modes use Advanced artist fields as artist data and rebuild artist variants through the Anima prompt ordering rules.
Anima Prompt Data Conditioning
- model
- clip
- EASYUSE_ANIMA_PROMPT_DATA
- model
- positive
- negative
- latent_image
◄mod_guidance_modeprompt_data►
◄mod_w_profilestep_i8_skip27►
◄artist_mix_modeprompt_data►
◄artist_mix_start_percent0.50►
◄artist_mix_strength_scale1.00►
◄artist_mix_style_gain1.35►
◄artist_mix_rms_scale_cap2.00►
◄artist_mix_exact_top_k4►
◄artist_mix_cluster_count4►
◄artist_mix_dominant_isolationtrue►
◄artist_mix_dominant_threshold0.25►
CategoryEasyUse Anima/Prompt
Inputs (14)
| Name | Type | Default | Description |
|---|---|---|---|
| model | MODEL | MODEL to pass through or patch with Anima Mod Guidance. | |
| clip | CLIP | CLIP used to encode prompt data and Mod Guidance quality tags. | |
| EASYUSE_ANIMA_PROMPT_DATA | EASYUSE_ANIMA_PROMPT_DATA | Structured prompt data from Anima Prompt Studio Advanced v2. | |
| mod_guidance_mode | COMBO | prompt_data | prompt_data uses the prompt-data boolean, enabled forces Anima Mod Guidance on, and disabled bypasses the model patch. |
| mod_w_profile | COMBO | step_i8_skip27 | Spectrum AnimaModGuidance per-block profile. off bypasses the model patch. |
| artist_mix_mode | COMBO | prompt_data | prompt_data follows EASYUSE_ANIMA_PROMPT_DATA, off/prompt keeps artists inline. off: Cost: 1 positive branch. Keeps artist-field text inline in the positive prompt. prompt: Cost: 1 positive branch. Keeps artist-field text inline in the positive prompt. average: Cost: 1 positive branch. Weighted average of artist conditionings; fastest stable mix. delta_rms: Cost: 1 positive branch. Mixes artist deltas from the base prompt and restores RMS style energy; usually stronger than average. hybrid: Cost: top_k + 1 positive branches. Keeps strongest artists as exact branches and compresses the tail with delta_rms; recommended balance. clustered: Cost: about cluster_count plus dominant artists. Groups similar artist deltas and compresses each cluster; useful for many artists. exact: Cost: N positive branches. Most faithful artist-specific model output mix. composite_exact: Cost: N + 1 positive branches. Adds one composite prompt branch plus exact artist branches. late_exact: Cost: base + N late exact branches. Applies exact mixing only after start. average_late_exact: Cost: 1 average branch plus N late exact branches. Fast early mix, exact late refinement. scheduled_average: Cost: scheduled average branches. Changes artist weights across timestep ranges. |
| artist_mix_start_percent | FLOAT | 0.500–1 | Start percent used by late/scheduled artist mix modes. |
| artist_mix_strength_scale | FLOAT | 1.000–5 | Strength multiplier used by exact artist mix modes. |
| artist_mix_style_gain | FLOAT | 1.350–3 | Style delta gain used by delta_rms, hybrid tail, and clustered compressed branches. |
| artist_mix_rms_scale_cap | FLOAT | 2.001–5 | Maximum RMS energy restore scale for delta_rms compressed artist branches. |
| artist_mix_exact_top_k | INT | 40–64 | Hybrid mode keeps this many strongest artists as exact positive branches. |
| artist_mix_cluster_count | INT | 41–32 | Clustered mode compresses non-dominant artists into this many positive branches. |
| artist_mix_dominant_isolation | BOOLEAN | true | Clustered mode keeps artists above the dominant threshold as exact branches. |
| artist_mix_dominant_threshold | FLOAT | 0.250–1 | Clustered dominant isolation threshold based on normalized artist weight. |
Outputs (4)
| Name | Type | Description |
|---|---|---|
| model | MODEL | MODEL after prompt-data model patches. |
| positive | CONDITIONING | Positive CONDITIONING encoded from prompt data. |
| negative | CONDITIONING | Negative CONDITIONING encoded from prompt data. |
| latent_image | LATENT | Empty latent image created from prompt-data width and height with batch size 1. |