ComfyUI Node
✨Video Caption (VLM)
Extract text prompt from video frames using Vision-Language Models
✨Video Caption (VLM)
- video_frames
- video
- caption_text
- preview_frames
◄rule视频复刻与重构►
◄custom_rulefalse►
◄custom_rule_content►
◄user_prompt►
◄sampling_modeAuto (Uniform)►
◄frame_count5►
◄manual_indices►
◄vlm_service智谱/glm-4.6V-Flash►
◄ollama_auto_unloadtrue►
◄seed0►
Category✨Prompt Assistant
Inputs (12)
| Name | Type | Default | Description |
|---|---|---|---|
| rule | COMBO | 视频复刻与重构 | Preset rule for video captioning |
| custom_rule | BOOLEAN | false | Enable custom rule input |
| custom_rule_content | STRING | Custom rule content, only used when Custom Rule is enabled | |
| user_prompt | STRING | Enter additional prompts here, sent with the rule | |
| sampling_mode | COMBO | Auto (Uniform) | Auto: uniformly sample frames. Manual: specify indices via manual_indices |
| frame_count | INT | 51–32 | Number of frames to sample (only for Auto mode) |
| manual_indices | STRING | Specific frame indices for Manual mode, e.g. 0,10,20 or 0-10 | |
| vlm_service | COMBO | 智谱/glm-4.6V-Flash | Select VLM Service |
| ollama_auto_unload | BOOLEAN | true | Auto unload Ollama model after generation |
| seed | INT | 00–18446744073709550000 | Controls randomness. Set to non-fixed mode to force re-execution |
| video_framesopt | IMAGE | The video frames to analyze (IMAGE batch) | |
| videoopt | VIDEO | Compatible with VIDEO type nodes (like VHS) |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| caption_text | STRING | — |
| preview_frames | IMAGE | — |