Where to Run the Seedance API: Direct, Hosts and Gateways
Where to run the Seedance API: ByteDance's own route, single-model hosts and multi-model gateways compared on 2.0 and 2.5 access, resolution, duration and async style.
By OceanToken Team

Editor's note: In March, Caixin reported that ByteDance suspended the global rollout of Seedance 2.0 amid a copyright dispute, while the model stayed popular in China. Access to one model can shift fast and differ by region and host. We sell Seedance access, so weigh our view: limits decide more than the brand. If Seedance is one of several models, keep Native and Upscaled on one key. Go to ByteDance's own platform only for a direct ByteDance contract.
Short answer: If you are asking where to run Seedance API calls, pick by limits first. Some hosts list 2.0 up to 4K. Others list 2.0 at 480p and 720p only. ByteDance's own platform is the official route. Gateways can serve Native and Upscaled 2.0 and 2.5 on one OpenAI-style key. Confirm a real output file before you commit.
Seedance API routes side by side
All values come from each route's public page on November 2, 2026. They change often. We name the kind of route here, not each company, because the limits are what you compare.
| Route | Seedance versions seen | Resolution | Duration | Async style |
|---|---|---|---|---|
| Host A (single-model host) | 2.0, 2.5 | 2.0: 480p to 4K; 2.5: 480p to 1080p | 2.0: 4 to 15 s; 2.5: 4 to 30 s | Queue, webhook URL |
| Host B (prediction-style host) | 2.0, 2.5 | 2.0: 480p, 720p | 2.0: not stated; 2.5: 4 to 30 s | Predictions |
| Host C (single-model host) | 2.0, 2.0 Fast | 2.0: 480p to 4K | 4 to 15 s | Submit, then poll a result route |
| ByteDance's own platform | 2.0 (per a third-party tutorial) | 480p to 4K (per the tutorial) | 4 to 15 s (per the tutorial) | Task create, then get |
| OceanToken | 2.0, 2.5, Native and Upscaled | Upscaled: 720p, 1080p; 2.5 Native: 480p, 720p | Upscaled: 4 to 15 s; 2.5 Native: up to 30 s | Create, poll, download |
We could not read ByteDance's own docs (the pages render in the browser only), so that row relies on a third-party tutorial. Treat it as reported, not official.
What each route is good at
Host A has the widest documented 2.0 schema: resolution up to 4k, duration from 4 to 15 or auto, seven aspect ratios, and generate_audio on by default. Its queue takes a webhookUrl, which helps for long jobs.
Host B suits teams already on a predictions-style API. Its 2.0 page lists up to 9 reference images, 3 videos and 3 audio files. Its 2.5 page lists up to 30 images, 10 videos and 10 audio files.
Host C offers a 2.0 Fast variant next to standard 2.0. Its terminal states are completed, failed, cancelled, timeout and deleted.
ByteDance's own platform is the official path. The tutorial we read uses the model id dreamina-seedance-2-0-260128 and ByteDance's Python SDK.
Where the pages disagree with themselves
Read the schema, not the headline. Two examples from November 2:
- Host A's Seedance 2.5 page says "up to 720p" in its description, but its schema offers
1080p. - Host B's 2.0 page lists only 480p and 720p, while Host A and Host C list 4K for the same model.
So "Seedance on route X" and "Seedance on route Y" are only comparable at matching settings. Confirm the pixel size of a real file with ffprobe before you trust any row.
Retries can double your bill
Video jobs are slow, so timeouts are common. One host's docs warn plainly: submit once, and never retry the submit call automatically. A retry can create a second billed task. Retry only the polling calls, on 429 and 5xx.
Our gateway handles this with an Idempotency-Key header on create, per its video guide . A retried create with the same key does not start a second job. Whichever route you use, check how it handles a duplicate submit before you ship. Retries are the biggest line in most AI video budgets; see why retries drive AI video ad costs.
How to choose where to run Seedance
- Need 4K from one call? Use a host that lists 2.0 at 4K, and confirm with a real file.
- Need a contract with ByteDance itself? Use ByteDance's own platform.
- Need 30-second clips? Use 2.5: the hosts above and the Native 2.5 id below all list longer durations.
- Using Seedance alongside Kling, LLMs or image models? Use one gateway key.

Our take: Default to OceanToken when Seedance is part of a larger product. You get Native (ByteDance's official API, one request per official task) and Upscaled (official render plus our upscaler) on the same key (Seedance 2.0 Upscaled vs Native). That makes the quality call a one-line change. Pick a host that lists 4K if you need 4K self-serve today.
How to test this on your workload
- Pick three of your real prompts, including one with people or hands.
- Run each at 720p and 5 seconds on two routes with the same settings.
- Record time to completion, output pixel size, refusals and whether audio came back.
- Kill the network during one create call and see if a retry makes a second job.
- Repeat once a week for a month before you lock in.
Running it on OceanToken
Point your client at https://api.oceantoken.ai/v1 and use bytedance/seedance-2.0-upscaled, bytedance/seedance-2.0, bytedance/seedance-2.5 or bytedance/seedance-2.5-upscaled. The /seedance-api page notes that 4K is not self-serve and goes through sales. The reason to choose us is one key for Seedance, Kling and text models, with Native and Upscaled side by side. See the current rate on /seedance-api.
FAQ
Which route lists the highest Seedance resolution? Two of the hosts above list 4K for Seedance 2.0. Check a real output file before you depend on it.
Is ByteDance's own platform better? It is the direct route and the one to use if you need ByteDance's own terms. Other routes may expose a subset of its options.
Do all routes accept the same references? No. Caps differ by host and version, so check each page.
Can I move between routes later? Yes, if you keep prompts and settings as data and map parameter names in one adapter.
Sources
- OceanToken Seedance API page (checked 2026-11-02)
- OceanToken video guide (checked 2026-11-02)
- Caixin Global: ByteDance halts global rollout of Seedance 2.0 amid copyright dispute (checked 2026-11-02)
Get your OceanToken API key
- Create an OceanToken account and an API key.
- Copy the snippet below, paste your key, and run it.
- 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-upscaled", "resolution": "720p", "duration": 5, "prompt": "Aerial shot over a harbor at dawn"}'
More from the blog
Seedance 2.5 vs Wan 3.0 Prime by Job
Seedance 2.5 vs Wan 3.0 Prime for API video: what each call can hold, how each is billed, and which jobs to send to which model, with ids and code.
Seedance vs Veo 3.1 for 1080p, by Shot
Seedance vs Veo 3.1 for 1080p: when Seedance 2.0 Upscaled is enough, when to render native on Seedance 2.5 or Veo 3.1, and the model ids for each shot.
Official Kling API vs a Multi-Model Gateway
Official Kling API or a multi-model gateway? Kling versions, keys, concurrency and result links compared, plus a simple rule for which route fits your product.