ComfyUI Node
CLIP Text Encode (Differential)
Encodes text with CLIP and injects seed-driven variation to avoid seed collapse on z-image.
CLIP Text Encode (Differential)
- clip
- CONDITIONING
◄text—►
◄seed0►
◄noise_std0.18►
◄token_dropout0.10►
◄per_token_gain_std0.12►
◄global_offset_std0.08►
◄pooled_noise_std0.05►
◄pooled_global_offset_std0.04►
◄preserve_magnitudetrue►
◄orthogonalize_noisetrue►
◄min_std_scale0.001►
Categoryconditioning
Inputs (12)
| Name | Type | Default | Description |
|---|---|---|---|
| text | STRING | Prompt text to encode with seed-dependent perturbations. | |
| clip | CLIP | CLIP model used for encoding. | |
| seed | INT | 00–9223372036854776000 | Seed used for the differential noise. |
| noise_std | FLOAT | 0.180–1 | Scale of seed-driven noise added to token embeddings (per-token Gaussian). |
| token_dropoutopt | FLOAT | 0.100–0.8 | Chance to zero-out entire token embeddings (seeded) to create variation. |
| per_token_gain_stdopt | FLOAT | 0.120–1 | Std of multiplicative gain noise per token (1 + N(0, std)), adds structured jitter. |
| global_offset_stdopt | FLOAT | 0.080–1 | Gaussian shift shared by all tokens to steer conditioning per seed. |
| pooled_noise_stdopt | FLOAT | 0.050–1 | Noise applied to pooled_output so refiner/backbones also see variation. |
| pooled_global_offset_stdopt | FLOAT | 0.040–1 | Global pooled shift; helps video/backbones that rely more on pooled_output. |
| preserve_magnitudeopt | BOOLEAN | true | Re-normalize after perturbation to keep overall embedding statistics stable. |
| orthogonalize_noiseopt | BOOLEAN | true | Project noise off the original token directions to keep semantics while adding diversity. |
| min_std_scaleopt | FLOAT | 0.0010–0.1 | Lower bound for std/mean used to scale noise so low-variance prompts still get variation. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| CONDITIONING | CONDITIONING | Seed-sensitive conditioning embedding. |