ComfyUI Node
QwenVL Prompt Enhancer
A ComfyUI node in 🧠aistudynow/QwenVL with 14 inputs and 1 output.
QwenVL Prompt Enhancer
- ENHANCED_OUTPUT
â—„model_nameQwen3-0.6Bâ–ş
â—„quantizationNone (FP16)â–ş
â—„attention_modeautoâ–ş
â—„use_torch_compilefalseâ–ş
â—„deviceautoâ–ş
â—„prompt_textâ–ş
â—„enhancement_styleđź“ť Enhanceâ–ş
â—„custom_system_promptâ–ş
â—„max_tokens256â–ş
â—„temperature0.70â–ş
â—„top_p0.90â–ş
â—„repetition_penalty1.10â–ş
â—„keep_model_loadedtrueâ–ş
â—„seed1â–ş
Category🧠aistudynow/QwenVL
Inputs (14)
| Name | Type | Default | Description |
|---|---|---|---|
| model_name | COMBO | Qwen3-0.6B | Pick the Qwen-VL checkpoint. First run downloads weights into models/LLM/Qwen-VL, so leave disk space. |
| quantization | COMBO | None (FP16) | Precision vs VRAM. FP16 gives the best quality if memory allows; 8-bit suits 8–16 GB GPUs; 4-bit fits 6 GB or lower but is slower. |
| attention_mode | COMBO | auto | auto tries flash-attn v2 when installed and falls back to SDPA. Only override when debugging attention backends. |
| use_torch_compile | BOOLEAN | false | Enable torch.compile('reduce-overhead') on supported CUDA/Torch 2.1+ builds for extra throughput after the first compile. |
| device | COMBO | auto | Choose where to run the model: auto, cpu, mps, or cuda:x for multi-GPU systems. |
| prompt_text | STRING | Prompt text to enhance. Leave blank to just emit the preset instruction. | |
| enhancement_style | COMBO | đź“ť Enhance | 6 options: đź“ť Enhance, đź“ť Refine, đź“ť Creative Rewrite, đź“ť Detailed Visual, đź“ť Artistic Style, đź“ť Technical Specs |
| custom_system_prompt | STRING | — | |
| max_tokens | INT | 25632–1024 | — |
| temperature | FLOAT | 0.700.1–1 | — |
| top_p | FLOAT | 0.900–1 | — |
| repetition_penalty | FLOAT | 1.100.5–2 | — |
| keep_model_loaded | BOOLEAN | true | — |
| seed | INT | 11–4294967295 | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| ENHANCED_OUTPUT | STRING | — |