Skip to contentOceanToken

5 min read

Seedance ComfyUI Guide: Partner Nodes, Templates, Limits

How to run Seedance 2.0 in ComfyUI with partner nodes: templates, login and credit rules, real-person checks, Comfy's limits, and when to use the API.

By OceanToken Team

Seedance ComfyUI Guide: Partner Nodes, Templates, Limits

Editor's note: ComfyUI is the best place we know to design a Seedance shot chain, because you can see every input and swap one node at a time. It is not where we would run a nightly batch of 500 clips. This page covers ComfyUI's own partner nodes, which bill in Comfy credits. It is not about the OceanToken API, which we cover at the end.

Short answer: Seedance ComfyUI support comes through partner nodes. The model runs on the provider's side, so you need no local GPU or weights. You must sign in to a Comfy account and hold credits. Comfy ships templates for text-to-video, reference-to-video, first-last-frame, 4K reference, storyboard and a cheaper Mini variant.

What the Seedance ComfyUI partner nodes are

Comfy describes partner nodes as "special nodes that connect to external API services." For Seedance 2.0, the compute happens at the provider. Your machine only builds the graph and sends the request (Comfy docs ).

The docs list these templates:

Template What it does
Text-to-Video (T2V) Prompt only
Reference-to-Video (R2V) Prompt plus image, video or audio references
First-last-frame-to-video (FLF2V) Fixes the start and end frames
R2V 4K Renders the final clip at 4K
Storyboard to Video Builds frames with image nodes, then animates them
Mini T2V and Mini R2V Lighter, faster variant at a lower per-run cost

The Comfy hub page for the T2V model lists 4 to 15 second clips, seven aspect ratios plus Auto, and 480p, 720p, 1080p and 4K (Comfy hub ).

Seedance 2.0 ComfyUI workflow: setup steps

  1. Update ComfyUI. Comfy recommends the nightly build. Missing nodes usually mean an old install or a failed import at startup.
  2. Sign in under Settings, then User.
  3. Buy credits under Settings, then Credits.
  4. Open Templates, search for Seedance, and start with T2V.
  5. Run one short 720p clip to confirm login, credits and network.
  6. Swap in R2V or FLF2V once the base graph works.

If you run on Comfy Cloud, note one catch. Cloud gets new core nodes only after a ComfyUI stable release, so a template from the docs may be missing there for a while.

Login, credits and network rules

These come from Comfy's partner nodes overview , checked 2026-11-02:

  • Credits are prepaid. There is no free usage, and refunds are not supported.
  • Credits expire. Monthly credits end with the billing period. Top-up credits expire one year after purchase.
  • No bring-your-own key. You can't plug in your own ByteDance key yet. Comfy says the feature is still being considered.
  • Login is local only. Direct login works only from 127.0.0.1 or localhost. On a LAN box or a remote server, use a ComfyUI Account API Key.
  • Your inputs leave the machine. Prompts and images go to the third-party provider, which applies its own moderation.

ComfyUI ByteDance nodes for real people

Real faces need extra steps. The real-human guide  uses the ByteDance Create Image/Video Asset node:

  1. Upload a portrait to the asset node and run the graph.
  2. ComfyUI returns a verification link from ByteDance.
  3. Complete the liveness check, which takes under a minute.
  4. Save the Group ID (the person) and the Asset ID (that image).

The check uses no extra credits. Next time, enter the saved Group ID with a new photo of the same person. If it matches, you skip the check. References go up to 9 images, 3 videos and 3 audio clips.

If a normal R2V run hits a protected real person, the node returns an error. Switch to the Real Human template instead of working around it.

Where ComfyUI helps most

The graph lets you chain shots without leaving the canvas. Make a first frame with an image node, feed it to FLF2V, take the last frame of clip A, and use it to start clip B.

[Image node] -> first frame --> [Seedance FLF2V] -> clip A -> [Get last frame]
                                                                    |
[Prompt B] --------------------> [Seedance from last frame] <-------+  -> clip B

Chains drift. Each hop adds small changes, so go back to a clean reference every few shots. Check each last frame before you feed it forward, since a bad frame spoils every later shot.

Troubleshooting a graph that won't run

Symptom Likely cause Fix
Nodes show as missing Old ComfyUI or failed import Update, restart, read the startup log
Run stops at once Not signed in, or zero credits Check Settings, then User and Credits
Login fails on a server Not on localhost Use a ComfyUI Account API Key
Error on a portrait reference Protected real person Use the Real Human template
Template missing on Cloud Waiting for a stable release Run locally on nightly, or wait

Keep a small known-good graph saved on its own. After an update breaks something, load it first. You'll know in a minute whether the fault is yours or the new version's.

How to test this on your workload

  1. Build a three-clip FLF2V chain for one product shot.
  2. Run the same idea as one single 15-second T2V clip.
  3. Compare subject continuity and the cost shown in credits.
  4. Repeat the chain with Mini and note what changes.
  5. Port the winning settings into one API call and compare the output.

Our take: Use ComfyUI for look-dev, shot chaining and anything you want to see before you commit. Once a job is batch, scheduled or triggered by another system, move it to code. Retries, logging and cost tracking are simpler there.

Running it on OceanToken from code

Prototype the graph in ComfyUI, then port the settings to an API call. OceanToken serves Seedance at https://api.oceantoken.ai/v1 with POST /v1/videos. The Native ids bytedance/seedance-2.0 and bytedance/seedance-2.5 run on ByteDance's official API. The -upscaled ids add OceanToken's own upscaling. See the current rate on /seedance-api.

Two notes from the video guide  help when you port a graph. A first frame with no aspect_ratio makes the clip follow the frame's shape. The -upscaled models accept 1 to 9 reference images and no frames, so FLF2V chains belong on the Native ids.

FAQ

Do I need a GPU for Seedance in ComfyUI? No. Partner nodes run the model on the provider's side.

Can I use my own ByteDance API key in ComfyUI? Not yet. Comfy says partner nodes don't support your own key.

Why are the Seedance nodes missing? Usually an old ComfyUI or a failed import. On Comfy Cloud, new nodes arrive after a stable release.

Does the liveness check cost credits? No. Comfy says verification uses no extra credits.

Is Mini priced separately? Comfy says Mini runs at a lower per-run cost on the same pricing line as the full model. Check Comfy's pricing page for figures.

Sources

  1. Comfy docs: Seedance 2.0 partner nodes  (checked 2026-11-02)
  2. Comfy docs: Seedance 2.0 real human  (checked 2026-11-02)
  3. Comfy docs: partner nodes overview  (checked 2026-11-02)
  4. Comfy hub: Seedance 2.0 text-to-video  (checked 2026-11-02)
  5. OceanToken Seedance API page (checked 2026-11-02)
  6. OceanToken docs: video guide  (checked 2026-11-02)

Get your OceanToken API key

  1. Create an OceanToken account and an API key.
  2. Copy the snippet below, paste your key, and run it.
  3. Check the current rate on /seedance-api.
curl https://api.oceantoken.ai/v1/videos \
  -H "Authorization: Bearer $OCEANTOKEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "bytedance/seedance-2.0", "prompt": "A paper boat drifts down a rain gutter, close tracking shot", "resolution": "720p", "seconds": 5}'

Get API Key 

  • Seedance
  • AI video
  • Tutorial
  1. 7 min read

    A Unified API for Chat, Image and Video

    Building on a unified API for chat, image and video: sync vs async calls, billing units, retries, timeouts and storage, with three calls and a launch checklist.

    • AI video
    • LLM API
    • Tutorial
  2. 7 min read

    Kimi K3 API International Access Guide

    Use the Kimi K3 API internationally: sign up with email, pay by card, create one key and call moonshotai/kimi-k3 with an OpenAI SDK. Steps, code and checks.

    • Kimi
    • LLM API
    • Tutorial
  3. 6 min read

    Call Tencent Hy3 With the OpenAI SDK

    Call Tencent Hy3 with the OpenAI SDK outside China: base URL, model id, streaming, tool calls, reasoning_effort values, Hy3 vs Hy3 preview, and token billing.

    • Hunyuan
    • LLM API
    • Tutorial