Seedance2 ComfyUI
https://github.com/anil-matcha/seedance2-comfyui- Category
- AI Tools
- Rank
- No. 1116Tools index
Previous survey · No. 1136 ·
- Pricing
- Open Source
- Type
- TOOL
- Builder
- Anil-matcha
- GitHub
- 40 stars
- Date
About
ComfyUI custom nodes for ByteDance Seedance 2.0 video generation — drive text-to-video, image-to-video, and video extension from your local ComfyUI workflows.
What it does
This node pack turns ComfyUI graphs into clients for MuAPI-hosted Seedance generation. Nodes upload reference media, submit jobs, poll for completion, and return video URLs, preview frames, request identifiers, or saved frame tensors. Separate flows cover prompts, reference images, mixed media, character sheets, character continuity, extension, and Mini generation.
Why it's ranked here
The pack covers more than basic prompt generation. Mixed image, video, and audio references, character-sheet handoff, extension, and video saving make it useful for composed workflows. The catch is architectural: ComfyUI orchestrates remote MuAPI jobs rather than running the model locally, and Seedance 2.5 requires another package.
What's good
Inputs and outputs compose naturally inside ComfyUI. Reference images upload automatically, completed jobs expose identifiers for extension, and the saver returns both a disk path and frame tensors. Authentication can come from a wired key or existing MuAPI CLI configuration. Two example graphs demonstrate basic generation and the character-sheet workflow.
Tradeoffs
Generation requires a MuAPI key, network access, and sufficient service credits. Jobs poll every ten seconds for up to fifteen minutes, so a graph can remain occupied during generation. Media uploads and video downloads add transfer time. Native Seedance 2.5 support lives in a separate repository, and this package does not provide local model inference.
How to use it well
Use it when ComfyUI is your visual orchestration layer and Seedance 2.0 outputs must feed later image or video nodes. It is especially suited to reference-heavy clips, reusable character sheets, extensions, and frame extraction. Choose the separate companion pack for Seedance 2.5, or another interface when you need standalone automation, assistant access, or local inference.
Technical notes+
__init__.py merges mappings from seedance2_nodes.py and seedance2_video_saver.py. In seedance2_nodes.py, requests use the x-api-key header against https://api.muapi.ai/api/v1; _submit starts jobs, _poll checks prediction results every 10 seconds with a 900-second ceiling, and _upload_image converts ComfyUI tensors to JPEG before upload. _first_frame downloads the completed video and uses OpenCV, returning a 64 by 64 zero tensor on failure. seedance2_video_saver.py downloads MP4 output, chooses a collision-free numbered filename, and supports frame caps, initial-frame skipping, and interval sampling. requirements.txt declares requests, Pillow, NumPy, Torch, and OpenCV. Seedance2_T2V_Example.json and Seedance2_ConsistentCharacter_Example.json provide importable workflow examples.
Observed
- License
- MIT
- Primary language
- Python
- Interface
- ComfyUI custom nodes backed by the MuAPI HTTP API
- Installation surface
- ComfyUI Manager via Git URL, or manual clone plus pip requirements installation
- Authentication
- MuAPI key node, per-node key field, or MuAPI CLI configuration file
- Declared runtime dependencies
- requests, Pillow, NumPy, Torch, and OpenCV
- Example packaging
- Two ComfyUI workflow JSON examples are included
Read from README.md, requirements.txt, __init__.py, seedance2_nodes.py, seedance2_video_saver.py, LICENSE, Seedance2_T2V_Example.json, Seedance2_ConsistentCharacter_Example.json.
What it can do
Generate video from text prompts
Text description → Video file
Convert image to video
Static image → Video file
Extend existing video length
Video file → Longer video file
Create custom video generation workflows
ComfyUI workflow configuration → Executable video generation pipeline
Process video generation locally
Video generation request → Locally generated video
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.