ComfyUI Node
Anima CLIP Block Merger
A ComfyUI node in HaDeS/Anima Mergers with 8 inputs and 2 outputs.
Anima CLIP Block Merger
- clip1
- clip2
- CLIP
- info
◄base_ratio0.50►
◄vocab_embeddings0.50►
◄syntax_rigidity0.50►
◄semantic_focus0.50►
◄style_abstraction0.50►
◄final_projections0.50►
CategoryHaDeS/Anima Mergers
Inputs (8)
| Name | Type | Default | Description |
|---|---|---|---|
| clip1 | CLIP | — | |
| clip2 | CLIP | — | |
| base_ratio | FLOAT | 0.50-10–10 | 0.0 = CLIP 1, 1.0 = CLIP 2 |
| vocab_embeddings | FLOAT | 0.50-10–10 | Lexical token embeddings (embed_tokens) |
| syntax_rigidity | FLOAT | 0.50-10–10 | Early layers (~0-35% depth, 0-9 on Qwen3-0.6B) |
| semantic_focus | FLOAT | 0.50-10–10 | Middle layers (~35-72% depth, 10-19 on Qwen3-0.6B) |
| style_abstraction | FLOAT | 0.50-10–10 | Late layers (~72-100% depth, 20-27 on Qwen3-0.6B). Measured: carries almost all of a text encoder's effect |
| final_projections | FLOAT | 0.50-10–10 | Final normalization layer (norm.weight / final_layer_norm) |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| CLIP | CLIP | — |
| info | STRING | — |