ComfyUI Node
Qwen3VL 🖤 NightShift
Qwen3VL multimodal chat node. Accepts text, up to 4 images, and a video (image batch). Returns the model response as a string. Note: the 'max_length' widget controls max_new_tokens internally (name kept for saved workflow compatibility).
Qwen3VL 🖤 NightShift
- 🖼️ Image 1
- 🖼️ Image 2
- 🖼️ Image 3
- 🖼️ Image 4
- 🎥 Video
- Response
◄max_longer_edge768►
◄model_name<no models found>►
◄preset_system_prompt— Custom —►
◄user_prompt►
◄system_promptYou are a helpful assistant.►
◄max_length256►
◄temperature0.70►
◄top_p0.90►
◄num_beams1►
◄repetition_penalty1.20►
◄seed666►
◄tf32true►
◄attentionAuto►
◄torch_compilefalse►
◄keep_model_loadedtrue►
◄video_frames2►
◄devicenvidia_gpu►
CategoryNightShift/Qwen3VL
Inputs (22)
| Name | Type | Default | Description |
|---|---|---|---|
| max_longer_edge | INT | 7680–4096 | — |
| model_name | COMBO | <no models found> | 1 options: <no models found> |
| preset_system_prompt | COMBO | — Custom — | 3 options: — Custom —, 🎨 Z-Image Prompt, 🎬 LTX-2.3 Video Prompt |
| user_prompt | STRING | — | |
| system_prompt | STRING | You are a helpful assistant. | — |
| max_length | INT | 2561–4096 | — |
| temperature | FLOAT | 0.700–2 | — |
| top_p | FLOAT | 0.900–1 | — |
| num_beams | INT | 11–8 | — |
| repetition_penalty | FLOAT | 1.200.5–2 | — |
| seed | INT | 6660–18446744073709550000 | — |
| tf32 | BOOLEAN | true | — |
| attention | COMBO | Auto | 4 options: Auto, flash_attention_2, sdpa, eager |
| torch_compile | BOOLEAN | false | — |
| keep_model_loaded | BOOLEAN | true | — |
| video_frames | INT | 22–256 | — |
| device | COMBO | nvidia_gpu | 2 options: nvidia_gpu, cpu |
| 🖼️ Image 1opt | IMAGE | — | |
| 🖼️ Image 2opt | IMAGE | — | |
| 🖼️ Image 3opt | IMAGE | — | |
| 🖼️ Image 4opt | IMAGE | — | |
| 🎥 Videoopt | IMAGE | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| Response | STRING | — |