ComfyUI Node
TS Qwen 3
Run Qwen 3 VL locally over image, video and text to caption, describe or answer.
TS Qwen 3
- image
- video
- generated_text
- processed_image
◄model_namehuihui-ai/Huihui-Qwen3.5-2B-abliterated►
◄custom_model_id►
◄hf_token►
◄system_presetPrompts enhance►
◄prompt►
◄seed42►
◄max_new_tokens512►
◄precisionauto►
◄attention_modeauto►
◄offline_modefalse►
◄unload_after_generationfalse►
◄enabletrue►
◄max_image_size1024►
◄video_max_frames16►
◄custom_system_prompt—►
CategoryTS/LLM
Inputs (17)
| Name | Type | Default | Description |
|---|---|---|---|
| model_name | COMBO | huihui-ai/Huihui-Qwen3.5-2B-abliterated | Pick a model from the list. Choose 'Custom (manual)' for a third-party model. |
| custom_model_id | STRING | HuggingFace repo id (e.g. 'Qwen/Qwen2-VL-7B-Instruct') or a full local path. | |
| hf_token | STRING | HuggingFace token (Write/Read) for downloading models. Leave empty for public models. WARNING: saved in the workflow JSON in plain text — do not share a file that contains the token. | |
| system_preset | COMBO | Prompts enhance | System-prompt preset. Controls the model's behavior and response style. |
| prompt | STRING | Your request (prompt) to the model. | |
| seed | INT | 420–18446744073709550000 | Seed for reproducible generation. |
| max_new_tokens | INT | 51264–8192 | Maximum number of tokens in the response (output length). |
| precision | COMBO | auto | Weight precision. 'auto' picks the best option. int4/int8 require bitsandbytes to be installed. |
| attention_mode | COMBO | auto | Attention implementation. 'flash_attention_2' is faster and more memory-efficient but needs a compatible GPU. |
| offline_mode | BOOLEAN | false | Forbid downloads. Use only files already present in the models/LLM folder. |
| unload_after_generation | BOOLEAN | false | Unload the model from memory right after generation. Saves VRAM but slows down repeated runs. |
| enable | BOOLEAN | true | Enable processing. When disabled, images pass through to the output unchanged. |
| max_image_size | INT | 102464–4096 | Maximum image side length. Larger resolutions require more VRAM. |
| video_max_frames | INT | 164–256 | How many video frames to feed the model. More frames = better context but higher memory use. |
| imageopt | IMAGE | Input image. | |
| videoopt | IMAGE | Input video stream (batch of images). | |
| custom_system_promptopt | STRING | Your custom system prompt. Used when 'system_preset' is set to 'Your instruction'. |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| generated_text | STRING | Text generated by the model. |
| processed_image | IMAGE | Input image passed through (unchanged). |