QwenVL-F is a ComfyUI node in the QwenVL-F category from ComfyUI-QwenVL-F. Inputs: model_name, quantization, attention_mode, preset_prompt. Outputs: STRING…

Fork of ComfyUI-QwenVL: Qwen-VL nodes (Qwen2.5-VL / Qwen3-VL, Transformers and GGUF backends) for text generation, image understanding, and video analysis. Adds local-only model discovery — models are picked from models/text_encoders and models/LLM with no automatic downloading — plus explicit mmproj selection, up to 3 image inputs on the Advanced 

ComfyUI-QwenVL-F

model_name

quantization

attention_mode

preset_prompt

custom_prompt

max_tokens

enable_thinking

keep_model_loaded

seed

image

video

QwenVL-F

RESPONSE

Name	Type	Default	Description
model_name	COMBO	(no models found — see console)	Pick a Transformers checkpoint already present under models/text_encoders or models/LLM. Nothing is downloaded automatically — copy the model folder in yourself, then reload ComfyUI.
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.
preset_prompt	COMBO	🖼️ Detailed Description	Built-in instruction describing how Qwen-VL should analyze the media input.
custom_prompt	STRING		Optional override—when filled it completely replaces the preset template.
max_tokens	INT	51264–32768	Maximum number of new tokens to decode. Larger values yield longer answers but consume more time and memory.
enable_thinking	BOOLEAN	false	Enable thinking mode for Qwen3-VL Thinking models. When disabled, the model skips chain-of-thought reasoning and responds directly. Has no effect on non-Thinking models.
keep_model_loaded	BOOLEAN	true	Keeps the model resident in VRAM/RAM after the run so the next prompt skips loading.
seed	INT	11–4294967295	Seed controlling sampling and frame picking; reuse it to reproduce results.
imageopt	IMAGE		—
videoopt	IMAGE		—

QwenVL-F

Inputs (11)

Outputs (1)