Workflow based on official ComfyUI Flux.2 workflow with Qwen3-VL-Instruct integrated to generate JSON prompts
Description
FAQ
Comments (8)
those qwen-vl nodes have issues.
What issues? If you don't have enough VRAM, pick smaller Qwen-VL model and/or quantization... Tested on two different machines with 5090 & 4090 - no issues.
@SensoryLens I have rtx 6000 pro and a4000 pro dual GPU workstation with 500gb RAM - not the issue. The Qwen-VL nodes are flawed because the get a kwargs error (but when that is made to work they are inflexible because it wants the models to be in non standard location and forces you to download twice or create symlinks (but it doesn't like them). Your workflow is flawed because it uses an outdated version of the node and the node needs to be rebuilt / recreated before use.
@EricRollei21 I am always use latest/nightly of all my nodes. Outdated it not. About "non-standard" location - my machines are Linux and since I got multiple active Comfyui version I indeed use extra_model_paths.yaml for redirects.
@SensoryLens If you are the author of the Qwen-VL nodes, then next time I try them, I'll post information as issues on githhub.
@EricRollei21 I am not the author, just user. There are few similarly named nodes BTW. Please report the issue on github - maybe someone else experiencing the issues as well.
i cant manage to have the qwen3 vl node to work. Always the same error about transformers not up to date. i tried different versions of transformers, no way to solve this issue :(
@bertranddespas845 Please follow QwenVL installation requirements on their github. Also, if you are on Windows, the setup might be different than linux..