ComfyUI custom nodes for ByteDance Seedance 2.0 video generation, backed by the MyteamOne API — a third-party reseller that offers Seedance 2.0 (standard + Fast) at lower cost than the official Volcengine pricing.

✨ Features

Text → Video (T2V)
Image → Video (I2V, with optional last frame)
Video → Video (V2V, experimental — see notes below)
Multi-Reference → Video — up to 5 images + reference video + reference audio in one shot

All five nodes wrap the Seedance 2.0 / Seedance 2.0 Fast models behind one shared MyteamOne API Client node, so you only configure your key and endpoints once per workflow.

📦 Install

Option A — ComfyUI Manager

Search for MyteamOne in ComfyUI Manager and install. (Available once the node is published to the ComfyUI Registry.)

Option B — Manual

cd ComfyUI/custom_nodes
git clone https://github.com/YOUR_GH_USERNAME/ComfyUI-MyteamOne
cd ComfyUI-MyteamOne
pip install -r requirements.txt

Restart ComfyUI. The five nodes appear under Add Node → MyteamOne/Seedance2.0.

🔑 Setup

Register at myteamone.agione.pro and get an API key (looks like ak-xxxxxxxx...).
In your workflow, drop a MyteamOne API Client node and fill in:
- api_key — your ak-... key
- base_url — https://myteamone.agione.pro/hyperone/xapi/api/v1
- create_path — /videos
- poll_path — https://myteamone.agione.pro/hyperone/xapi/api/videos/generations/task/{taskid}
Wire the client output into any of the four generation nodes.

Endpoints may change. If the Client node returns 404, check the latest URL with MyteamOne support.

🎬 The five nodes

| Node | Use case | | --- | --- | | MyteamOne API Client | Configure key/endpoints once per workflow | | Seedance 2.0 (Text to Video) | Pure prompt → video | | Seedance 2.0 (Image to Video) | First frame (+ optional last frame) → video | | Seedance 2.0 (Video to Video) | Reference video → restyled video. Experimental — see notes | | Seedance 2.0 (Multi-Reference) | Up to 5 character/scene images + reference video + reference audio in natural-language prompt |

🧠 Tips

Multi-Reference prompts use natural language

Seedance 2.0's multi-reference mode does not use "图1/图2" style ordinals. Describe the references naturally in the prompt — "参考第一张图的角色，在第二张图的场景里…" — and the model figures out which is which.

Keep reference images clean

Never feed storyboard frames with arrows, callouts, or labels drawn on them. The model will faithfully reproduce those annotations as on-screen graphics in the generated video. Use bare character / scene art only.

Real-person video is blocked

ByteDance content policy rejects videos containing real, identifiable people when accessed via third-party APIs (you'll see InputVideoSensitiveContentDetected.PrivacyInformation). This is a hard limit at the model layer, not a node bug. For drama production, use AI-generated characters end-to-end.

Seedance 2.0 Fast does not support 1080p

The nodes fail fast with a clear message if you pick Fast + 1080p. Use 720p (or 480p) with Fast, or switch to standard Seedance 2.0 for 1080p.

Reference videos and audio must be hosted URLs

The media[].url field accepts HTTP/HTTPS URLs only — base64 video doesn't work. Host your videos / audio on Aliyun OSS, Tencent COS, Qiniu, etc., and paste the direct link into video_url / audio_url. Reference images work fine via base64 (wire a ComfyUI IMAGE directly).

📝 Changelog

0.3.1

Updated the default MyteamOne API endpoints to myteamone.agione.pro

0.3.0

Added Seedance 2.0 Multi-Reference node (up to 5 images + ref video + ref audio)
Switched all multi-modal nodes to the documented media[] payload format
Fast + 1080p now fails fast with a clear error message
Updated model IDs to current dreamina values

0.2.0

Text-to-Video, Image-to-Video, Video-to-Video nodes
API Client node with configurable endpoints

⚖️ License

MIT — see LICENSE.

🙏 Acknowledgements

ByteDance for the Seedance 2.0 model
MyteamOne for the API access
The ComfyUI community