ComfyUI Extension: ComfyUI-Tessiiiz-WanVideoToolkit
Run ComfyUI workflows without the setup
No installs, no CUDA version roulette, no GPU sitting idle on your bill. Bring a workflow and run it in the browser.
ComfyUI Wan 2.1 toolkit with visual multi-ref node, positioned reference gallery, and VACE multi-ref to video nodes.
Looking for a different extension?
Custom Nodes (0)
README
ComfyUI Wan 2.1 Toolkit
This pack is now rebuilt around a cleaner Wan VACE flow:
Wan Positioned Ref GalleryWan VACE Multi-Ref To VideoWan 2.2 VACE Multi-Ref To Video
Old nodes from earlier versions of this pack were removed from the node list.
Main Workflow
1. Wan Positioned Ref Gallery
This is the new visual multi-ref node.
It is designed for the exact "throw many refs in, preview them, and assign frames" workflow:
- drag and drop multiple images
- live preview inside the node
- reorder refs with left/right buttons
- remove refs
- set each frame position directly in the card
- output one
IMAGEbatch and oneSTRINGpositions value
2. Wan VACE Multi-Ref To Video
This is the new main node.
It is designed as a cleaner replacement for the built-in WanVaceToVideo, but adds:
- start frame pinning
- end frame pinning
- positioned middle frames
- multi-reference image batches
- control video frame filtering
Important:
- Multi Ref frames are applied through the pinned timeline path.
- The internal
reference_imagepath now uses only one primary reference frame for compatibility with Wan VACE runtime. - If you want many reference frames, put them in
positioned_frameswith positions, not all into a global reference slot.
It keeps the same output shape:
positivenegativelatenttrim_latent
So you can connect it to your sampler the same way as before.
2b. Wan 2.2 VACE Multi-Ref To Video
This is the Wan 2.2 path.
It keeps the same multi-ref timeline workflow, but is now easier to drop into a normal Wan image-to-video graph.
Internally it does two things:
- creates the normal
positive,negative, andlatentoutputs through the built-inWanImageToVideopath - creates
vace_embedsthrough the Wan 2.2WanVideoVACEEncodepath
Use it when your workflow is based on Wan 2.2 video nodes.
It outputs:
positivenegativelatentvace_embeds
Recommended wiring for Wan 2.2:
- Connect
wan_vaefrom your Wan 2.2 VAE loader. If your build exposes a plainvae, that also works. - Connect
control_video,start_image,positioned_frames,end_image, andextra_reference_imagesthe same way as the 2.1 node. - Connect
positive,negative, andlatentto your sampler like a normal Wan graph. - Use
vace_embedsonly if your downstream Wan 2.2 workflow has a node that consumes VACE embeddings.
3. Wan Multi Ref Stack
This older batch stacker is still available if you want a simple manual image combiner.
4. Wan Video Freeze Frames
Optional post-process node for freezing the first few frames to reduce Wan startup flicker.
Recommended Wiring
- Connect
positive,negative, andvae. - Connect
control_videoif needed. - Connect
start_image. - Connect
end_image. - Use
Wan Positioned Ref Galleryfor the middle frames. - Connect:
Wan Positioned Ref Gallery.images->Wan VACE Multi-Ref To Video.positioned_framesWan Positioned Ref Gallery.positions->Wan VACE Multi-Ref To Video.positions_of_positioned_frames
- Typical positions output:
20 30 50 90
- If you also have extra reference images, use
Wan Multi Ref Stackand connect it toextra_reference_images. - Send
positive,negative, andlatentoutputs to your sampler path like a normal Wan VACE workflow.
Useful Inputs
frames_to_keep_in_control_videoExamples:
1 20 30
1:16
positions_of_positioned_framesExamples:
1 20 30 L
Install
Put this folder in:
ComfyUI/custom_nodes/ComfyUI-Tessiiiz-WanVideoToolkit
Then restart ComfyUI.
Run ComfyUI workflows without the setup
No installs, no CUDA version roulette, no GPU sitting idle on your bill. Bring a workflow and run it in the browser.