Seedance 2.5 API: Official Model IDs and What to Verify Before Integration

Last verified: 2026-08-15. This page documents publicly described Seedance 2.5 API access through ByteDance first-party consoles. AISeedance25 is an independent editorial platform, not a ByteDance API reseller. Seedance and Dreamina are trademarks of their respective owners.

The 30-second answer

Short answer. Seedance 2.5 API access is available through two separate ByteDance consoles that must not be mixed: BytePlus ModelArk (international) under model ID dreamina-seedance-2-5-260628, and Volcano Engine Ark (China) under doubao-seedance-2.5. Billing is token-based; the usage.completion_tokens field returned after a successful generation is the authoritative charge signal. The published BytePlus rate card lists only 480p and 720p output tiers for Seedance 2.5.

If you are integrating Seedance 2.5 into a product, start by probing the correct model ID in the intended account and region, logging completion_tokens on every test, and measuring the downloaded file — not trusting marketing resolution labels alone.

On this page

• Two official API consoles
• Technical facts to verify first
• How token billing actually works
• 480p and 720p — not 4K on the API rate card
• Third-party gateways vs first-party access
• Pre-production integration checklist
• What AISeedance25 is (and is not) for API users
• Frequently asked questions
• Sources and verification status

How token billing actually works

Short answer. BytePlus documents token-based billing for Seedance 2.5. Token consumption grows with output duration, pixel dimensions, and frame rate, and video-input requests can trigger minimum-token floors. Only successfully generated videos are charged.

The provider-described token formula is:

(input_video_seconds + output_seconds) × width × height × frame_rate / 1024

That formula explains why a five-second 720p clip and a thirty-second 720p clip are not linearly comparable at a glance, and why adding a reference video increases token consumption even when the requested output duration stays the same. For integration work, the practical rule is simpler: log usage.completion_tokens on every successful response and build your cost model from observed usage in your account, not from a single published example row.

Failed generations are documented as non-billable on BytePlus — but your engineering time and queue latency are still consumed. Budget for retries when estimating production throughput.

480p and 720p — not 4K on the API rate card

Short answer. The published BytePlus rate card for Seedance 2.5 lists only 480p and 720p. Consumer pages that advertise 4K are not corroborated by the API rate card or any technical report on this page's last-verified date.

This matters for API integrations because your product UI may expose a "4K" export option that the underlying billed route does not actually deliver at native resolution. Always measure pixel dimensions from the delivered file. For the broader marketing-versus-rate-card gap across consumer surfaces, see the Seedance 2.5 vs 2.0 guide — Seedance 2.0 routes on several platforms still support higher published resolution tiers than Seedance 2.5 today.

Third-party gateways vs first-party access

Aggregators and API gateways can simplify multi-model routing, but they are not interchangeable with first-party BytePlus or Volcano Engine access. A gateway may rename the model, cap resolution, add queue layers, or bill on a flat per-second rate that diverges from token accounting.

Before treating a gateway quote as equivalent to direct access, verify in the target account:

1. The callable model ID matches dreamina-seedance-2-5-260628 (international) or doubao-seedance-2.5 (China).
2. The returned usage object includes a token count you can reconcile against the provider docs.
3. The downloaded file's measured resolution matches what your product promises downstream.

For a consumer-route comparison that includes Higgsfield, Dreamina, Artlist, and BytePlus side by side, see the Higgsfield vs direct access guide.

What AISeedance25 is (and is not) for API users

AISeedance25 is an independent editorial platform with a web workspace for Seedance generation. It is not an official ByteDance API distribution channel, not a licensee, and not a reseller of dreamina-seedance-2-5-260628 access.

If your job is product integration, your first-party path is BytePlus ModelArk or Volcano Engine Ark — linked and documented on the Seedance 2.5 official access guide. If your job is evaluating Seedance 2.5 output quality and limits before choosing a route, the Seedance 2.5 capability guide separates provider-described facts from independently tested behaviour.

Paid plans on AISeedance25 purchase credits for generation through routes available on this site, with a server-verified estimate shown before each submission. That is a different product from the BytePlus API contract.

Sources and verification status

Primary sources cross-checked for this page:

• ByteDance Seed — Seedance 2.5 — model owner release and capability description.
• BytePlus — Dreamina Seedance 2.5 product page — international API availability statement.
• BytePlus ModelArk — model pricing — token rates, calculation formula, and resolution tiers.
• Volcano Engine — China-region console for doubao-seedance-2.5.

API availability, model IDs, and pricing sections can change without a public changelog entry. Re-verify in the console before production cutover.

For review standards, see the AISeedance25 editorial policy and testing methodology. Corrections via the contact page.