Nodes/comfyui-fvmtools/Qwen 2.1 Pipe
ComfyUI Node

Qwen 2.1 Pipe

Bundles Qwen Image 2.1 MODEL, CLIP (qwen_image text encoder) and VAE into one pipe. Sources, strongest first: the single inputs, an OBVPM bundle (fields named model/clip/vae), an upstream pipe.

By ping1979ping·Created 6 months ago·Updated 2 days ago· 1
Qwen 2.1 Pipe
  • pipe
  • bundle
  • model
  • clip
  • vae
  • pipe
  • model
  • clip
  • vae
CategoryFVM Tools/Face

Inputs (5)

NameTypeDefaultDescription
pipeoptFVM_QWEN_PIPEAn existing Qwen 2.1 pipe, e.g. from another Qwen 2.1 Pipe node or a detailer's pipe output. Lowest priority: its model / clip / vae are only used for fields that neither the single inputs nor the bundle provide. Handy for swapping out just one part, e.g. pipe in + a LoRA-patched model -> new pipe.
bundleoptOBVPM_BUNDLEOBVPM bundle (Bundle node) with fields named exactly model, clip and vae - e.g. the 'BUN Qwen' bundle in the V6 workflow. Beats the pipe input, loses to the single model / clip / vae inputs. Other fields in the bundle are ignored.
modeloptMODELQwen Image 2.1 diffusion model (UNETLoader, e.g. qwen_image_2.1_int8_convrot). Highest priority - overrides the bundle and the pipe. A LoRA-patched model can go here too.
clipoptCLIPQwen3-VL text encoder, loaded with CLIPLoader type 'qwen_image' (e.g. qwen3vl_8b_int8_convrot). Highest priority. Not the qwen3.5 prompt-enhancer models - those only work with Generate Text.
vaeoptVAEQwen Image 2.1 VAE (qwen_image_2.1_vae_bf16). Highest priority. The older Qwen Image VAE does not fit the 2.1 model. Note: this VAE decodes RGBA; the detailer drops the alpha channel itself.

Outputs (4)

NameTypeDescription
pipeFVM_QWEN_PIPE—
modelMODEL—
clipCLIP—
vaeVAE—