ComfyUI Node
VLM Prompt Refiner (kppb)
A ComfyUI node in conditioning/klein with 21 inputs and 4 outputs.
VLM Prompt Refiner (kppb)
- character_ref
- scene_ref
- prop_ref
- refined_prompt
- image_caption
- filename_prefix
- video_prompt
◄ollama_urlhttp://localhost:11434►
◄modelhuihui_ai/qwen3-vl-abliterated:32b-instruct-q8_0►
◄modedescribe & enhance►
◄positive_prompt►
◄prompt_json►
◄edit_prompt►
◄system_prompt►
◄temperature0.30►
◄seed-1►
◄preserve_identitytrue►
◄unload_modeltrue►
◄use_claude_codefalse►
◄claude_modelopus►
◄trigger_wordohwx►
◄sfw_promptfalse►
◄generate_video_promptfalse►
◄motion_prompt►
◄audio_prompt►
Categoryconditioning/klein
Inputs (21)
| Name | Type | Default | Description |
|---|---|---|---|
| character_ref | IMAGE | Character reference image — identity/likeness is extracted from this | |
| ollama_url | STRING | http://localhost:11434 | — |
| model | STRING | huihui_ai/qwen3-vl-abliterated:32b-instruct-q8_0 | — |
| mode | COMBO | describe & enhance | 4 options: describe & enhance, image edit aware, caption only, dataset generation |
| scene_refopt | IMAGE | Additional reference for scene, pose, lighting, or outfit | |
| prop_refopt | IMAGE | Product/prop reference — handbag, sunglasses, sponsored item, etc. | |
| positive_promptopt | STRING | Connect from Prompt Builder's positive_prompt — used as fallback if VLM returns empty | |
| prompt_jsonopt | STRING | Connect from Prompt Builder's prompt_json output | |
| edit_promptopt | STRING | Connect from Image Edit Composer's edit_prompt output | |
| system_promptopt | STRING | — | |
| temperatureopt | FLOAT | 0.300–2 | Lower = more focused, higher = more creative |
| seedopt | INT | -1-1–2147483647 | -1 for random, or set for reproducible output |
| preserve_identityopt | BOOLEAN | true | Append identity lock phrase to reinforce likeness preservation for the diffusion model |
| unload_modelopt | BOOLEAN | true | Unload Ollama LLM from VRAM after inference. Turn ON for localhost (frees VRAM for Klein). Turn OFF for remote/RunPod (avoids slow reload between iterations) |
| use_claude_codeopt | BOOLEAN | false | Use Claude Code CLI instead of Ollama — does both stages in one shot with images |
| claude_modelopt | COMBO | opus | Claude model to use (sonnet recommended for speed/quality balance) |
| trigger_wordopt | STRING | ohwx | Trigger word for LoRA training captions (dataset generation mode only) |
| sfw_promptopt | BOOLEAN | false | Soften the output prompt to pass content safety filters on hosted models like nanobanana/Gemini. Leave off for local models. |
| generate_video_promptopt | BOOLEAN | false | Generate a Wan 2.6 I2V motion prompt from the scene settings. Turn off to save LLM overhead on smaller models. |
| motion_promptopt | STRING | — | |
| audio_promptopt | STRING | — |
Outputs (4)
| Name | Type | Description |
|---|---|---|
| refined_prompt | STRING | — |
| image_caption | STRING | — |
| filename_prefix | STRING | — |
| video_prompt | STRING | — |