Skip to content
UlazAI

UlazAI developer docs

Seedance API documentation: generate, status, 1080p and pricing

Pricing Dashboard Image studio

Seedance 1.5 Pro reference

Seedance API docs

To use the Seedance API, authenticate with a Bearer key, use POST /seedance/generate/ to create jobs, and poll GET /seedance/status/{generation_id}/ until the video is ready. UlazAI currently supports Seedance 1.5 Pro output in 480p, 720p, and 1080p with 4, 8, or 12 second durations.

Core endpoints

POST /seedance/upload/

POST /seedance/generate/

GET /seedance/status/{generation_id}/

Current capabilities

Text-to-video and image-to-video

Up to 2 guide images

Optional native audio

Next routes

Buy credits

See Seedance landing

Open canonical video API docs

Tip

This page answers the full developer flow in the first screen: the generate endpoint, status polling, 1080p support, Bearer auth and the pricing route.

Use /seedance/upload/ only when you need image-guided video generation. For text-only runs, call /seedance/generate/ directly and skip the upload step.

The Seedance model family: 1.5 Pro, 2.0 and 2.5

Seedance is ByteDance's text-to-video and image-to-video model. ByteDance launched the first Seedance in June 2025, released Seedance 2.0 in February 2026, and has since shown Seedance 2.5. Each generation pushes the same core idea further: cinematic motion, stronger prompt-following, multi-shot scenes, image and video references, and optional native audio in a single render. The API guide covers the currently supported Seedance versions without requiring you to set up ByteDance's own platform.

VersionReleased by ByteDanceWhat it adds
Seedance 1.5 Pro2025Text- and image-to-video, up to 2 guide images, optional audio, 480p/720p/1080p. Wired into UlazAI today.
Seedance 2.0Feb 2026Multimodal inputs (text, image, video, audio), multi-camera storytelling, native audio co-generation, longer clips.
Seedance 2.52026The newest step ByteDance has shown, aimed at faster generation and further quality and control gains.

Same code, newer model

UlazAI's public API generates with bytedance/seedance-1.5-pro today. The important part for developers: when a newer Seedance version is enabled here it reuses the same Bearer auth and the same generate-and-poll endpoints below. Your integration does not change — only the model identifier does — so you can build against this reference now and move up the family later.

Seedance 2.5 on this API: model IDs and referencing

Queries for seedance 2.5 api usually want two things: the model identifier to send, and how image references work on the new generation. ByteDance positions Seedance 2.5 around one-take creation and flexible referencing - more input images steering the same scene - while keeping the generate-and-poll job shape that this page documents.

What you send todayWhat changes when 2.5 is enabled here
model: bytedance/seedance-1.5-proOnly the identifier changes; auth and endpoints stay the same.
input_urls with up to 2 guide imagesExpect a wider referencing budget on the 2.x generation; the array shape stays the same.
POST /seedance/generate/ then poll /seedance/status/{id}/Unchanged - the job-based flow is the same across versions.

Until 2.5 is switched on here, production work stays on seedance-1.5-pro: upload the guide image, generate, poll, download. When the newer identifier goes live, re-run your existing job with only the model field changed and compare output before switching a pipeline over.

ByteDance and ARK: the official model, one unified route

People searching for ByteDance Seedance official documentation or the ARK Seedance API are usually looking for two different things: the model itself, and a way to actually call it. Seedance is ByteDance's own model, published through ByteDance's Volcano Engine ARK platform. That route means setting up an ARK account, region access, and raw model endpoints.

UlazAI sits in front of that as a third-party API layer. You get one Bearer key, a hosted image upload step, a shared credit wallet, and the small set of endpoints on this page — no ARK onboarding, no regional workaround, and no VPN or Chinese phone number to reach the model. If you specifically searched ARK API Seedance image upload, the equivalent here is a single call: POST /seedance/upload/ returns a permanent URL that you drop into input_urls on generate (see the upload endpoint below).

ARK concepts mapped to this API

If you already read ByteDance's ARK documentation, the vocabulary is different but the steps are the same. This is the direct translation:

On ByteDance ARK Here What changes for you
ARK account, project and endpoint setup One Bearer key No platform onboarding before the first call.
Host the guide image yourself, pass a public URL POST /seedance/upload/ You can send the raw file; the response gives you a permanent URL.
Model id in the request body model field Same idea; only the identifier changes between versions.
Poll the task until it completes GET /seedance/status/{generation_id}/ Same polling shape, one endpoint.

Image-to-video in two calls

This is the whole image-upload flow. The first call stores the image and returns a URL, the second call uses that URL as visual guidance for the video.

One uploaded still image on the left feeding a sequence of generated video frames on the right 
# 1. upload the guide image
curl -X POST https://ulazai.com/seedance/upload/ \
  -H "Authorization: Bearer $ULAZAI_API_KEY" \
  -F "[email protected]"

# -> {"url": "https://cdn.ulazai.com/uploads/<id>.jpg"}

# 2. generate, passing that URL back in input_urls
curl -X POST https://ulazai.com/seedance/generate/ \
  -H "Authorization: Bearer $ULAZAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
        "prompt": "slow dolly in, warm evening light",
        "input_urls": ["https://cdn.ulazai.com/uploads/<id>.jpg"],
        "resolution": "1080p"
      }'

You can pass up to two URLs in input_urls. Two images are read as a start and end reference, which is the usual way to keep a character or product consistent across the clip. For text-only generation, skip the upload call entirely.

Quick integration summary

Auth Bearer API key from /dashboard/ -> API Keys
Model ID bytedance/seedance-1.5-pro
Durations 4s, 8s, 12s
Resolutions 480p, 720p, 1080p
Aspect ratios 1:1, 21:9, 4:3, 3:4, 16:9, 9:16
Inputs Prompt only, or prompt plus up to 2 image URLs
Polling flow Create a job, store generation_id, then poll status until completed

Public endpoint summary

Method Endpoint When to use it
POST /seedance/upload/ Upload an image first if you want image-to-video guidance
POST /seedance/generate/ Create a Seedance 1.5 Pro generation job
GET /seedance/status/{generation_id}/ Check whether the job is still processing, completed, or failed

Authentication

All API requests require a Bearer token:

Authorization: Bearer YOUR_API_KEY

Get an API key from your dashboard under API Keys. If you still need credits, go to packages first.

POST /seedance/upload/

Upload input images for image-to-video guidance

Upload a guide image before generation if you want Seedance to anchor composition, character framing, or scene setup. The response gives you a permanent URL to pass into input_urls.

Request headers

Authorization: Bearer YOUR_API_KEY
Content-Type: multipart/form-data

Form data parameters

file * (file)

JPG, PNG, or WebP, up to 10MB.

Success response (200)

{
  "success": true,
  "url": "https://media.ulazai.com/seedance_images/u1_abc123.jpg",
  "filename": "seedance_images/u1_abc123.jpg"
}

Example usage (cURL)

curl -X POST https://api.ulazai.com/seedance/upload/ \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "file=@/path/to/image.jpg"
POST /seedance/generate/

Create a Seedance 1.5 Pro generation task

Send either the simplified payload below or the explicit model plus input format. Use the simplified version if you only need the public Seedance route with prompt, optional guide images, and output settings.

Request body (simple)

{
  "prompt": "A cinematic close-up of a chef flipping noodles in slow motion",
  "input_urls": ["https://media.ulazai.com/seedance_images/u1_abc123.jpg"],
  "aspect_ratio": "16:9",
  "resolution": "1080p",
  "duration": "4",
  "fixed_lens": true,
  "generate_audio": false,
  "prompt_directory_optin": true
}

Request body (model/input style)

{
  "model": "bytedance/seedance-1.5-pro",
  "input": {
    "prompt": "A cinematic close-up of a chef flipping noodles in slow motion",
    "input_urls": ["https://media.ulazai.com/seedance_images/u1_abc123.jpg"],
    "aspect_ratio": "16:9",
    "resolution": "720p",
    "duration": "8",
    "fixed_lens": true,
    "generate_audio": true
  }
}

Parameters

prompt * (string)

3 to 2500 characters describing the scene, motion, framing, and audio intent.

input_urls (array)

Optional. Up to 2 image URLs for image-to-video guidance.

aspect_ratio (string)

One of: 1:1, 21:9, 4:3, 3:4, 16:9, 9:16.

resolution (string)

One of: 480p, 720p, 1080p.

duration (string)

4, 8, or 12 seconds.

fixed_lens (boolean)

Enable for a more stable camera view.

generate_audio (boolean)

Enable synchronized audio generation. This increases credit usage.

prompt_directory_optin (boolean)

Optional. Share eligible prompts to the prompt directory and apply a discount when supported.

Success response (200)

{
  "success": true,
  "generation_id": "abc12345-1234-1234-1234-123456789012",
  "task_id": "seedance_task_xyz789",
  "credits_used": 40,
  "directory_discount_applied": true,
  "message": "Video generation started. This may take a few minutes."
}
GET /seedance/status/{generation_id}/

Poll task status

Poll this endpoint until the job leaves processing. On success you get a video URL back; on failure you should inspect the error and retry with a corrected payload.

Processing response

{
  "success": true,
  "status": "processing",
  "message": "Video is being generated..."
}

Success response (200)

{
  "success": true,
  "status": "completed",
  "video_url": "https://media.ulazai.com/seedance_outputs/video.mp4"
}

Code examples: cURL, Python and JavaScript

The full flow is submit → store generation_id → poll status. Here is the generate call in three common stacks; swap in your own key and prompt.

cURL

curl -X POST https://api.ulazai.com/seedance/generate/ \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "A cinematic close-up of a chef flipping noodles in slow motion",
    "aspect_ratio": "16:9",
    "resolution": "1080p",
    "duration": "4"
  }'

Python (requests)

import time, requests

BASE = "https://api.ulazai.com"
HEADERS = {"Authorization": "Bearer YOUR_API_KEY"}

# 1. Start the job
job = requests.post(f"{BASE}/seedance/generate/", headers=HEADERS, json={
    "prompt": "A cinematic close-up of a chef flipping noodles in slow motion",
    "aspect_ratio": "16:9",
    "resolution": "1080p",
    "duration": "4",
}).json()
generation_id = job["generation_id"]

# 2. Poll until it leaves "processing"
while True:
    status = requests.get(f"{BASE}/seedance/status/{generation_id}/", headers=HEADERS).json()
    if status.get("status") == "completed":
        print("Video:", status["video_url"])
        break
    if status.get("status") == "failed":
        raise RuntimeError(status)
    time.sleep(8)  # back off; do not hammer generate

JavaScript (fetch)

const BASE = "https://api.ulazai.com";
const HEADERS = { "Authorization": "Bearer YOUR_API_KEY", "Content-Type": "application/json" };

const job = await fetch(`${BASE}/seedance/generate/`, {
  method: "POST",
  headers: HEADERS,
  body: JSON.stringify({
    prompt: "A cinematic close-up of a chef flipping noodles in slow motion",
    aspect_ratio: "16:9",
    resolution: "1080p",
    duration: "4",
  }),
}).then(r => r.json());

const id = job.generation_id;
let video;
while (!video) {
  const s = await fetch(`${BASE}/seedance/status/${id}/`, { headers: HEADERS }).then(r => r.json());
  if (s.status === "completed") video = s.video_url;
  else if (s.status === "failed") throw new Error(JSON.stringify(s));
  else await new Promise(r => setTimeout(r, 8000));  // poll every ~8s
}
console.log("Video:", video);

Usage flow

  1. Upload one or two guide images with /seedance/upload/ only if you need image-to-video guidance.
  2. Call /seedance/generate/ with your prompt, output settings, and optional input_urls.
  3. Store the returned generation_id.
  4. Poll /seedance/status/{generation_id}/ until status becomes completed.
  5. Use the returned video_url to download, embed, or pass the final asset into your own workflow.

Error codes

400 Invalid request parameters

401 Authentication failed (missing or invalid API key)

402 Insufficient credits

429 Rate limit exceeded

500 Internal server error

Pricing and rate limits

Seedance billing on UlazAI runs on credits, so cost scales with resolution and duration rather than a fixed per-second rate. Use this as a planning guide; the live credit cost is shown in packages and at request time.

OutputTypical useCost driver
480pDrafts and quick iterationsLowest credits per second
720pSocial-ready clipsMid credits per second
1080pFinal deliveryHighest credits per second; longer durations cost more

Rate limits and 429 handling: if you submit many jobs in parallel you may receive 429 Rate limit exceeded. Treat the API as job-based: submit, store the generation_id, and poll /seedance/status/{generation_id}/ every 5-10 seconds rather than hammering generate. On a 429, back off exponentially (for example 2s, 4s, 8s) before retrying.

Seedance 1.5 Pro vs Seedance 2.0: this reference covers bytedance/seedance-1.5-pro, the model currently wired into Video Studio. ByteDance has since introduced Seedance 2.0 as the next generation; when it is enabled here it uses the same generate-and-poll flow and the same auth, so integration code does not change — only the model identifier does.

Frequently asked questions

Does the UlazAI Seedance API support Seedance 2.0 and 2.5?

The API generates with bytedance/seedance-1.5-pro today. ByteDance released Seedance 2.0 in February 2026 and has since shown 2.5. When a newer version is enabled here it uses the same Bearer auth and the same generate-and-poll endpoints, so your integration code does not change — only the model identifier does.

How does UlazAI relate to ByteDance's official Seedance / ARK API?

Seedance is ByteDance's own model, offered through ByteDance's Volcano Engine ARK platform. UlazAI is a third-party API layer that runs Seedance behind one simplified route with a single Bearer key, hosted image upload, and a shared credit wallet — no ARK account or raw model endpoints to manage.

How do I upload an image for Seedance image-to-video?

Send the image as multipart/form-data to POST /seedance/upload/ with your Bearer key. The response returns a permanent URL; pass it (up to two) in the input_urls array on POST /seedance/generate/. Text-only prompts skip the upload step.

How much does the Seedance API cost?

Billing runs on credits, so cost scales with resolution and duration. A short 480p draft costs the fewest credits; a 12-second 1080p render with audio costs the most. The exact cost shows in packages and again as credits_used in the generate response.

Is Seedance available worldwide through the API?

Yes. You call UlazAI's hosted HTTPS endpoints with a Bearer key, so no VPN, regional workaround, or Chinese phone number is needed. Any server that can make an authenticated HTTPS request can submit jobs and poll for results.

Where to go next

Buy credits Seedance 1.5 Pro landing Seedance 2 pricing Video API docs API reference index Open Seedance in Video Studio

Build Seedance API requests around media validation and job status

Validate reference files and supported fields before upload, submit one job and retain its identifier. Poll until a terminal state and show partial upload or generation errors explicitly.

Use the UlazAI request schema and model availability shown in this reference as the integration contract, even when an ARK example uses different field names.

Music Studio Songs, instrumentals and your own lyrics. AI song generator