ComfyUI Node
Analyze CLIP (SAE)
Run a prompt through ComfyUI's CLIP text encoder and encode the residual at one layer through a Qwen-Scope SAE. Use this for FLUX.2 / Qwen-Image / any Qwen-based encoder.
Analyze CLIP (SAE)
- clip
- sae
- features
- info
◄prompta photo of a cat►
◄top_k64►
◄skip_first_tokenfalse►
CategoryQwenScope/Capture
Inputs (5)
| Name | Type | Default | Description |
|---|---|---|---|
| clip | CLIP | — | |
| sae | QSCOPE_SAE | — | |
| prompt | STRING | a photo of a cat | — |
| top_k | INT | 640–4096 | Top-K features kept per token. 0 = ReLU only. |
| skip_first_token | BOOLEAN | false | — |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| features | QSCOPE_FEATURES | — |
| info | STRING | — |