Seedance 2.5 is live on EvoLinkTry Seedance 2.5
Seedance 2.5 Opens to Developers August 7, With EvoLink Access
Product Launch

Seedance 2.5 Opens to Developers August 7, With EvoLink Access

Jessie
Jessie
COO
June 23, 2026
Updated on August 7, 2026
10 min read
Short answer: ByteDance released Seedance 2.5 as a model and creator product on July 31, 2026, then opened the public developer API on August 7, 2026. Seedance 2.5 is available on EvoLink from the same day for text-to-video, image-to-video, and reference-to-video workflows. The release brings up to 30 seconds of audio-video generation in one pass, as many as 30 image, 10 video, and 10 audio references, and timestamp-level editing.
For EvoLink users, the August 7 release means the model can be evaluated without adding a separate provider integration. Use the Seedance 2.5 model page to start generating, compare current pricing, and choose the route that matches your input workflow. This article records the dated release and availability history.
Last verified: August 7, 2026.
Try Seedance 2.5 on EvoLink

Seedance 2.5 Status on August 7, 2026

Seedance 2.5 reached the market in two steps: creator access on July 31 and public developer API access on August 7. That distinction matters because a model announcement, a consumer product, a provider API, and an EvoLink generation route are different release layers.

Release layerStatus on August 7, 2026What it means
Official model releaseReleased by ByteDance Seed on July 31Seedance 2.5 is an official production model, not a preview name
Jimeng AI and Doubao ProCreator rollout began July 31Users can generate through ByteDance product interfaces, subject to account and regional access
Volcano Ark and BytePlus ModelArkPublic developer API released August 7API teams can begin provider-specific integration and live validation
EvoLink generation routeLive from August 7Generate through text-to-video, image-to-video, or reference-to-video with one EvoLink account and API key

The upstream API and EvoLink route are both live, but channel-specific contracts still matter. Do not substitute a Volcano Ark or BytePlus model ID inside an EvoLink request, and do not assume that every creator-product control appears in every API workflow.

What the August 7 API Release Changes

Before August 7, teams could study the model, prepare requests, and test creator-product workflows, but they could not treat Seedance 2.5 as a generally available developer dependency. The public API release changes the next action from preparation to verification.

API teams can now evaluate whether Seedance 2.5 solves a real production constraint:

  • whether a continuous 20-30 second scene reduces stitching and continuity failures;
  • whether a larger reference package keeps people, products, environments, and voices consistent;
  • whether timestamp-level instructions make longer prompts easier to direct;
  • whether editing preserves approved footage instead of forcing a complete rerun;
  • whether the cost per accepted output is better than the cost of repeated short generations.
The release does not make Seedance 2.5 the automatic default for every video job. Short hooks, rapid variants, action-heavy clips, and cost-sensitive drafts may still fit Seedance 2.0 or another video route better.

Confirmed Seedance 2.5 Capabilities

ByteDance's official release confirms the following model capabilities:

CapabilityConfirmed scopeProduction value
Single-pass durationUp to 30 seconds of audio-video generationBuild a complete ad beat, dialogue exchange, or continuous product story in one generation
Multimodal referencesUp to 30 images, 10 videos, and 10 audio filesAssign identity, product, motion, environment, style, and sound references within one brief
Editing controlTimestamp-level editing of audio and video contentTarget a specific time segment without describing the entire video again
Professional referencesGreen-screen, camera-perspective, reference-based, and clay-render workflowsGuide composition, subject motion, spatial layout, and environment changes
Native audio-video generationAudio and video are generated within the same model workflowCoordinate dialogue, performance, ambient sound, and visible action

These are model capabilities, not a promise that every API route exposes every control in exactly the same way. Teams should validate the parameter contract on the provider or gateway route they actually plan to use.

Seedance 2.5 should also not be budgeted as a native 4K API route. The currently verified provider pricing scope covers 480p and 720p output. Treat higher-resolution marketing language on creator-product pages separately from the API contract.

Availability by Channel

ChannelAvailable useWhat to verify before relying on it
Jimeng AIBrowser-based creator generationAccount access, regional availability, credits, and product-specific controls
Doubao ProVideo generation inside the Doubao productAccount rollout and available settings
DreaminaInternational creator-product experienceRegional access and whether its controls match Jimeng
Volcano ArkChina-facing developer APIAccount eligibility, callable model ID, quotas, request contract, and billing
BytePlus ModelArkInternational developer APIRegion, account eligibility, request acceptance, limits, and token accounting
EvoLinkLive text-to-video, image-to-video, and reference-to-video routesChoose the correct EvoLink model ID, confirm current pricing, and measure output acceptance for your workload
Creator products, provider APIs, and the EvoLink route are separate Seedance 2.5 release layers
Creator products, provider APIs, and the EvoLink route are separate Seedance 2.5 release layers

The safest rule is simple: use the contract for the exact channel you intend to call. A working Jimeng generation does not define the BytePlus request schema, and upstream provider identifiers and token rates do not replace EvoLink model IDs or EvoLink pricing.

EvoLink's value is not merely adding another model name. A unified API gateway lets teams keep model selection, asynchronous task handling, usage reporting, billing, and fallback policy in one integration while choosing the best video route for each workload.

Seedance 2.5 is available on EvoLink through three production routes:

EvoLink model IDUse it when
seedance-2.5-text-to-videoThe scene starts from a written brief without source media
seedance-2.5-image-to-videoOne start image or a start-and-end image pair should anchor the result
seedance-2.5-reference-to-videoImages, videos, or audio should control identity, motion, environment, style, or sound

The recommended launch path is:

  1. Open the Seedance 2.5 model page and choose the workflow that matches your source material.
  2. Create or reuse an EvoLink API key and use the corresponding EvoLink model ID.
  3. Run a fixed evaluation set using your own people, products, scripts, and source media.
  4. Record successful tasks, failed tasks, queue time, charges, accepted seconds, and retry reasons.
  5. Keep Seedance 2.0 or another verified model as an explicit fallback until 2.5 passes repeated production tests.
Then use the Seedance 2.5 prompt library to prepare a repeatable first test.

Dated Seedance 2.5 Release Timeline

DateEventDeveloper meaning
June 23, 2026Seedance 2.5 was previewed at Volcano Engine FORCETeams learned the model direction, but public access was not established
July 31, 2026ByteDance Seed officially released Seedance 2.5The model and initial creator-product rollout became official
July 31, 2026Jimeng AI and Doubao Pro rollout beganCreator access became available through ByteDance product interfaces
August 7, 2026Public developer API released through Volcano Ark and BytePlus ModelArkAPI teams could move from preparation to live provider use
August 7, 2026Seedance 2.5 became available on EvoLinkTeams could use one EvoLink key for the new model while retaining existing task handling and fallback routes

This timeline will continue to separate upstream release events from production maturity. Future updates should be driven by verified pricing changes, API-contract changes, or original production testing rather than by launch-day marketing alone.

Production Checks Before You Scale

CheckWhy it matters
Successful generation with the exact route and model IDDocumentation alone does not prove request acceptance
Duration, resolution, and media limitsCreator-product controls may differ from the API contract
Actual charge and returned usageProvider token rates are not the same as the cost of one finished video
Failed-task and moderation billingFailure policy changes the real cost of experimentation
Median and p95 queue timeA visually strong model can still miss production deadlines
Character, product, and audio consistency across retriesSelected demos do not show repeatability
Fallback routeProduction should not stop if a new route is limited or unavailable
Measure cost per accepted output, not just list price. A higher-priced generation can be efficient when it avoids multiple reruns; a lower unit price can still be expensive when long clips repeatedly fail review.

FAQ

Has Seedance 2.5 been released?

Yes. ByteDance Seed officially released the model on July 31, 2026. The public developer API followed on August 7, 2026.

What is the Seedance 2.5 API release date?

The public developer API release date is August 7, 2026. This is separate from the July 31 model and creator-product release.

Yes. Seedance 2.5 is available on EvoLink from August 7, 2026, through text-to-video, image-to-video, and reference-to-video routes. Open the Seedance 2.5 model page to start a generation or review current pricing.
Use seedance-2.5-text-to-video, seedance-2.5-image-to-video, or seedance-2.5-reference-to-video, depending on the input workflow. Do not substitute an upstream provider model ID in an EvoLink request.

Where can developers access Seedance 2.5?

The upstream developer channels are Volcano Ark for China-facing access and BytePlus ModelArk for international access. Eligibility, regions, quotas, and accepted parameters can still vary by account and channel.

Can Seedance 2.5 generate 30-second videos?

Yes. ByteDance's official release states that Seedance 2.5 supports up to 30 seconds of audio-video generation in one pass, plus multi-round extension. Confirm which duration and extension controls are exposed on the route you use.

How many references does Seedance 2.5 support?

The official model release lists up to 30 images, 10 video clips, and 10 audio clips in one request. Provider-specific file formats, sizes, and duration limits still need route-level verification.

Does the Seedance 2.5 API support 4K?

Do not plan an API workflow around 4K. The currently verified provider pricing scope covers 480p and 720p output. Creator-product marketing and API output tiers should be treated as separate claims.

How much does Seedance 2.5 cost?

Upstream providers publish token-based rates, while EvoLink pricing is route-specific. Use the Seedance 2.5 model page for the current EvoLink price instead of converting an upstream token rate into an assumed per-video cost.

Should production teams replace Seedance 2.0 immediately?

No. Run matched tests first. Keep Seedance 2.0 or another verified route as a fallback until Seedance 2.5 improves the target workload's accepted-output rate, total production cost, or editing effort over repeated jobs.

Sources

Ready to Reduce Your AI Costs by 89%?

Start using EvoLink today and experience the power of intelligent API routing.