Developer Guide · Updated August 29, 2026
Seedance 2.5 API: Status & Access Guide
Seedance 2.5 is live through OpenRouter and fal.ai. Compare verified model IDs, supported inputs, current 480p and 720p pricing, and the asynchronous integration pattern before you spend on a generation.
API Status — August 29, 2026
OpenRouter is currently the lower-cost public path. Its published Seedance 2.5 price is about $0.2311/second at 720p versus fal.ai’s roughly $0.473/second. That puts a 5-second 720p clip near $1.16 on OpenRouter versus about $2.31–$2.37 on fal.ai. Verify live provider pricing before production because rates and schemas can change.
Live Seedance 2.5 API Providers
Publicly verifiable options and current prices, with hosted web interfaces kept separate from raw developer APIs.
Volcano Engine / BytePlus (ByteDance)
Enterprise access nowFirst-party access, enterprise agreements only. With the consumer launch complete (July 31, 2026), public API availability is the next step to watch.
OpenRouter
Live — lower published priceModel slug bytedance/seedance-2.5. Verified August 29, 2026: from about $0.1028/sec at 480p and about $0.2311/sec at 720p. Supports 4–30 second clips and optional generated audio.
fal.ai
Live — 2.5 API availableThree live endpoints: bytedance/seedance-2.5/text-to-video, image-to-video, and reference-to-video. Verified August 29, 2026. Published pricing is about $0.473/sec at 720p or $0.2205/sec at 480p.
OpenArt
Hosted UI live (reported)Reports hosting Seedance 2.5 generation in its web UI as of launch day. A hosted interface, not a raw API — useful for testing outputs, not for integrations.
Other inference platforms
Multiple providers now liveReplicate, BytePlus and other inference providers also surface Seedance 2.5. Compare exact model capabilities, output resolution, queue reliability and pricing before integrating.
The Seedance 2.5 API Flow
OpenRouter and fal.ai both use an asynchronous job pattern: submit, poll or receive a callback, then store the completed result.
1
Submit a generation job
POST your structured prompt with parameters (duration, aspect ratio, resolution, references). The API returns a job ID immediately — video generation is asynchronous everywhere.
2
Poll job status
Check the job status endpoint (or receive a webhook). Generation time scales with length and resolution; 30-second clips will take meaningfully longer than 5-second ones.
3
Download and store the result
Completed jobs return a video URL, usually time-limited. Download and store it on your side — do not hotlink provider URLs in production.
The pattern in pseudocode (provider-agnostic)
// 1. Submit — structured prompt in, job ID out
const job = await POST('/video/generate', {
model: 'bytedance/seedance-2.5', // OpenRouter model slug
prompt: structuredPrompt, // subject + camera + lighting + style
duration: 30, // 2.5: up to 30s single segment
aspect_ratio: '16:9',
});
// 2. Poll until terminal state
let status;
do {
await sleep(POLL_INTERVAL);
status = await GET(`/video/jobs/${job.id}`);
} while (status.state === 'queued' || status.state === 'processing');
// 3. Persist the result on your side
if (status.state === 'completed') await download(status.video_url);Illustrative pattern only — endpoint paths and field names vary by provider. Consult your provider’s official documentation for exact schemas.
API Spend Is Prompt Quality, Measured in Dollars
In a UI you can eyeball a bad generation and tweak. Over an API, every under-specified prompt is a billable clip you throw away. At 30 seconds per generation, 2.5 raises the stakes: one vague prompt wastes six times the compute of a 5-second test.
The fix is structural, not creative: every prompt should carry subject, action beats, camera movement, lighting, style, duration, and aspect ratio. PromptMotion generates prompts in exactly this format — paste them into your API payloads, or read the Seedance prompt guide to build them by hand.
Frequently Asked Questions
Is there a public Seedance 2.5 API?
Yes. OpenRouter lists bytedance/seedance-2.5 on its video generation API, while fal.ai offers separate text-to-video, image-to-video, and reference-to-video endpoints. First-party access also runs through ByteDance's Volcano Engine / BytePlus.
Which providers offer the Seedance 2.5 API?
OpenRouter and fal.ai both provide public Seedance 2.5 API access. OpenRouter uses one bytedance/seedance-2.5 model slug through its asynchronous video endpoint. fal.ai exposes text-to-video, image-to-video, and reference-to-video endpoints. ByteDance's Volcano Engine / BytePlus is the first-party path.
What does the Seedance 2.5 API cost?
As verified August 29, 2026, OpenRouter advertises Seedance 2.5 from about $0.1028 per second at 480p and about $0.2311 per second at 720p. fal.ai publishes about $0.2205 per second at 480p and $0.473 per second at 720p. A 5-second 720p clip is therefore about $1.16 on OpenRouter versus about $2.31–$2.37 on fal.ai. Exact cost varies by resolution, aspect ratio, duration, and input type.
How do I integrate the Seedance 2.5 API?
Both providers use an asynchronous job flow: submit the prompt and generation parameters, retain the returned job ID and polling URL, poll or receive a webhook, then download the completed video. Keep OPENROUTER_API_KEY or FAL_KEY on the server and never expose it in browser code.
How does prompt format affect Seedance API results?
Heavily. API generations are unattended — there is no UI preview loop — so structured prompts with explicit subject, camera movement, lighting, style, duration, and aspect ratio are the difference between usable output and wasted spend. PromptMotion generates prompts in exactly this structure, which can be passed directly into API calls.
Ship Your Prompts Before You Ship Your Integration
Build and test structured prompts free today — 3 per day, no credit card — before a billable 2.5 API call. Then follow the Seedance 2.5 workflow guide to turn that prompt into a controlled generation.
PromptMotion is an independent third-party AI prompt optimization tool and is not affiliated with, endorsed by, or associated with ByteDance Ltd., Volcano Engine, OpenRouter, fal.ai, or any API provider named on this page. “Seedance” is a trademark of ByteDance Ltd. Provider expectations reflect the publicly observable Seedance 2.0 rollout and are not announcements. Always consult providers’ official documentation and pricing.