Generate Videos

February 23, 2026 (October 23, 2026)

Table of contents

  1. Input Modes
    1. Aspect ratio auto-detection
  2. Model Capabilities
    1. What is new in seedance-2.5
    2. Content on seedance-2.5
  3. Request Headers
  4. Request Body
  5. Responses
  6. Model
  7. Examples
  8. Try It

Generate videos using Dreamina AI models from text prompts with optional image frames. All video generation is asynchronous — this endpoint returns immediately with a job ID. Poll GET /videos/jobid for completion, or use replyUrl webhook for automatic callbacks.

Video generation typically completes within 60-180 seconds depending on the model and duration. The long_video and edit modes on seedance-2.5 take considerably longer — a 21-second edit measured 13 minutes and a 35-second long video 17, so use replyUrl rather than polling for those.

Input Modes

The input mode is automatically determined from the provided parameters:

Mode API Value Trigger Description
Prompt prompt No image refs Text-to-video generation
First Frame first_frame firstFrameRef provided Video starts from uploaded image
First/Last Frame end_frame firstFrameRef + endFrameRef Video transitions between two images
Multi-Frame multi_frame frame_N_imageRef params 2-10 keyframe images with per-frame prompts
Omni Reference unified_edit omni_N_imageRef/omni_N_videoRef/omni_N_audioRef in prompt Interleaved text, image, video, and audio references (Seedance 2.5, 2.0, 2.0 Fast, 2.0 Mini, Wan 3.0; HappyHorse 1.1 images only)
Long Video long_video duration above 30 31-180 second generation (seedance-2.5 only)
Edit edit editVideoRef provided Re-cut an existing video (seedance-2.5 only)

Aspect ratio auto-detection

In first_frame and end_frame modes the aspect ratio is taken from the uploaded image, and the closest standard ratio (21:9, 16:9, 4:3, 1:1, 3:4, 9:16) is used. The ratio parameter cannot be supplied alongside image refs. When using firstFrameRef together with endFrameRef, both images must share the same aspect ratio.

Model Capabilities

Model Region Durations Resolution Input Modes
seedance-2.5 US, CA 4-30s, or 31-180s in long video mode 480p (default), 720p, 1080p Prompt, First/Last Frame, Omni Reference, Long Video, Edit
seedance-2.0 US, CA 4-15s 720p, 1080p, 4k Prompt, First/Last Frame, Omni Reference
seedance-2.0-fast US, CA 4-15s 720p, 1080p, 2k, 4k (above 720p, upscaled from 720p) Prompt, First/Last Frame, Omni Reference
seedance-2.0-mini US, CA 4-15s 720p, 1080p, 2k, 4k (above 720p, upscaled from 720p) Prompt, First/Last Frame, Omni Reference
seedance-1.5-pro US, CA 5, 10, 12s 720p, 1080p (1080p is CA only) Prompt, First/Last Frame
seedance-1.0-pro CA 5, 10s 1080p Prompt, First Frame
seedance-1.0-mini US 5, 10s 720p Prompt, First Frame, Multi-Frame
seedance-1.0-fast CA 5, 10s 720p, 1080p Prompt, First Frame, Multi-Frame
happyhorse-1.1 CA 4-15s 720p (default), 1080p Prompt, First Frame, Omni Reference (images only)
wan-3.0 CA 4-15s 720p (default), 1080p Prompt, First/Last Frame, Omni Reference

Both regions run the same resolutions on the Seedance 2.0 and 2.5 models. seedance-2.5 runs 1080p on CA since August 14, 2026 and on US since October 23, 2026. seedance-2.0-fast and seedance-2.0-mini render at 720p, and Dreamina upscales 1080p, 2k and 4k from that 720p video (as of October 23, 2026). 480p arrived with seedance-2.5 on August 7, 2026. On seedance-1.5-pro, 1080p is CA only. See Setup Dreamina for per-region setup and pricing.

happyhorse-1.1 (by Alibaba) and wan-3.0 (by Tongyi Lab) run on CA accounts as of October 23, 2026. Both make 4 to 15-second videos with sound at 720p or 1080p. HappyHorse takes up to 9 reference images, and Wan up to 10 images, 5 videos and 5 audio clips.

What is new in seedance-2.5

  • Up to 30 seconds in one call, twice the 15-second ceiling of the 2.0 family.
  • Up to 50 omni references — 30 images, 10 videos, 10 audio clips — against the 2.0 family’s 12.
  • Defaults to 480p, which costs roughly half of 720p.
  • 1080p, added on CA August 14, 2026 and on US October 23, 2026.
  • Long video — pass a duration of 31 to 180 and the generation switches to long video mode, six times the ordinary 30-second ceiling. Works from a plain prompt or with omni references. Priced on its own, higher, per-second rate.
  • Edit — pass editVideoRef and Seedance re-cuts an existing 4-30 second video instead of generating a new one, keeping its motion and framing while applying whatever the prompt asks for. Combine it with omni_N_imageRef references to bring new elements in.
  • Attaching an omni video reference makes a generation cheaper per second at 480p on both regions (as of October 21, 2026, US 720p no longer has the lower rate). Reference videos may be up to 30.2 seconds each and are billed on top of the output duration.

Content on seedance-2.5

Markedly more permissive than the 2.0 family:

  • Real human faces are accepted. A photographic portrait works directly as omni_N_imageRef or firstFrameRef, and the identity carries through the clip. The 2.0 family rejects them, so the face-grid workaround in the Seedance 2.0 real-faces tutorial is not needed here.
  • Cinematic violence renders. Stylised action — fight scenes, impacts, blood spray — is generated rather than refused.
  • Recognisable third-party content is refused, as fail_code 2053. It is judged against the finished video two to four minutes into generation, not against your prompt text — so a prompt naming no person and no property can still fail if it describes one, such as a costumed superhero, a trademarked landmark, or a character written as an unmistakable celebrity. Failed generations are not charged.

https://api.useapi.net/v1/dreamina/videos

Request Headers 

Authorization: Bearer {API token}
Content-Type: application/json
# Alternatively you can use multipart/form-data
# Content-Type: multipart/form-data

Request Body 

{
  "prompt": "A serene mountain landscape at sunset with camera slowly panning right",
  "model": "seedance-2.0",
  "ratio": "16:9",
  "duration": 5
}
  • account is optional. Specific account to use. Auto-inferred from image refs if provided. If omitted, a random account with free capacity is picked among those whose region offers model.
  • prompt is required for prompt mode. Optional for first_frame and end_frame. Cannot be used with multi_frame mode (use frame_N_prompt instead). Maximum 20000 characters on wan-3.0, 15000 on seedance-2.5, 2500 on happyhorse-1.1, 5000 on every other model.
  • model is optional, the AI model to use (default: seedance-2.0). Available models depend on account region — see Model Capabilities table.
  • ratio is optional, video aspect ratio (default: 16:9). One of 21:9, 16:9, 4:3, 1:1, 3:4, 9:16, plus 4:5, 5:4 and 9:21 on happyhorse-1.1. wan-3.0 takes 16:9, 4:3, 1:1, 3:4 and 9:16. Cannot be specified when image refs are provided (image aspect ratio is used).
  • duration is optional, video duration in seconds (default: 5). Valid values depend on model — see Model Capabilities table. On seedance-2.5 any value from 31 to 180 switches the request to long video mode, which is priced separately. Cannot be used with editVideoRef, where the output always matches the source video.
  • resolution is optional, one of 480p, 720p, 1080p, 2k or 4k. Each model accepts only some of these — see the Resolution column in Model Capabilities above. Defaults to 720p, except seedance-2.5 which defaults to 480p.
  • firstFrameRef is optional, imageRef from POST /assets/account for the starting frame. Triggers first_frame mode.
  • endFrameRef is optional, imageRef for the ending frame. Triggers end_frame mode. Requires firstFrameRef.
  • editVideoRef is optional, videoRef from POST /assets/account for the video to re-cut. Triggers edit mode, seedance-2.5 only. The source must be 4-30 seconds. The output inherits the source’s length and aspect ratio, so duration and ratio cannot be supplied alongside it. May be combined with omni_N_imageRef/omni_N_videoRef references, mentioned in the prompt as usual.
  • frame_N_imageRef is optional (N=1-10), imageRef for keyframe N. Triggers multi_frame mode when at least 2 frames are provided.
  • frame_N_prompt is optional, per-frame prompt for multi_frame mode. Maximum 5000 characters.
  • frame_N_duration is optional, per-frame duration in seconds for multi_frame mode (1-6, default: 5).
  • replyUrl is optional, webhook URL for job status callbacks. Receives POST requests with the job record on submission and on completion/failure. Callback body has the same JSON shape as GET /videos/jobid response.
  • replyRef is optional, custom reference string passed back in webhook callbacks.
  • omni_N_imageRef, omni_N_videoRef and omni_N_audioRef are optional assetRef values from POST /assets/account, and trigger Omni Reference mode. Mention each one in the prompt as @imageN, @videoN or @audioN — for example "Make @image1 dance like @video1 while @audio1 plays". Supported by seedance-2.5, seedance-2.0, seedance-2.0-fast, seedance-2.0-mini, wan-3.0, and happyhorse-1.1 (images only); per-model counts are in the reference-limit table below.
  • maxJobs is optional, override max concurrent jobs for this request (1-50).

Omni Reference constraints:

  • Only supported by seedance-2.5, seedance-2.0, seedance-2.0-fast, seedance-2.0-mini, wan-3.0, and happyhorse-1.1 (images only)
  • Prompt must contain @imageN, @videoN, and/or @audioN placeholders matching the provided omni_N_*Ref params
  • A reference that is uploaded but never mentioned in the prompt is ignored, and is not billed
  • The Seedance 2.0 models and wan-3.0 need at least one image or video reference. Audio references alone are refused with 400. seedance-2.5 accepts audio alone
  • Cannot be combined with firstFrameRef, endFrameRef, or frame_N_imageRef

Reference limits differ by model:

Limit seedance-2.5 Seedance 2.0 family wan-3.0 happyhorse-1.1
Image references 30 9 10 9
Video references 10 3 5 —
Audio references 10 3 5 —
Total materials combined 50 12 20 9
Total video duration across all refs 30.2s 15.4s 15s —
Total audio duration across all refs 30.2s 15s 15s —
Individual video 1.8-30.2s, 300-6000px 2-15.4s, 200-2160px 1-15s, 240-4096px —
Individual audio 1.8-30.2s 2-15s 1-15s —

Uploads via POST /assets/account are accepted up to the loosest of these limits (video 1.8-30.2s, audio 1.8-30.2s, video max 50 MB, audio max 15 MB), so one asset can be reused across models. The per-model limits above are checked when you submit the generation.

Multi-frame constraints:

  • Minimum 2 frames, maximum 10 frames
  • No gaps allowed (frame_1 through frame_N must be contiguous)
  • All frame images must have the same aspect ratio
  • Top-level prompt cannot be used (use frame_N_prompt instead)
  • firstFrameRef/endFrameRef cannot be combined with multi_frame
  • Total duration of non-last frames must equal a valid model duration

Responses

  • 200 OK

    Job created successfully. Video is generating in the background.

    {
  "jobid": "j0223140530123456789v-u12345-CA:[email protected]:dreamina",
  "type": "video",
  "status": "created",
  "model": "seedance-2.0",
  "created": "2026-02-23T14:05:30.123Z",
  "request": {
    "prompt": "A serene mountain landscape at sunset with camera slowly panning right",
    "model": "seedance-2.0",
    "ratio": "16:9",
    "duration": 5,
    "inputMode": "prompt"
  },
  "response": {
    "forecastCost": 125
  },
  "code": 200
}

    Poll GET /videos/jobid for completion status, or use replyUrl for webhook callbacks.

  • 400 Bad Request

    Validation error.

    {
  "error": "Parameter model (invalid-model) valid values: seedance-2.5, seedance-2.0, seedance-2.0-fast, seedance-2.0-mini, seedance-1.5-pro, seedance-1.0-mini, seedance-1.0-pro, seedance-1.0-fast, happyhorse-1.1, wan-3.0"
}

  • 401 Unauthorized

    Invalid API token.

    {
  "error": "Unauthorized"
}

  • 402 Payment Required

    Subscription expired or insufficient credits.

    {
  "error": "Account has no subscription or subscription expired"
}

  • 403 Forbidden

    One of two reasons, and ret says which.

    ret 5006: the account has no Dreamina subscription that includes this model. Dreamina runs it only on a paid plan, and credits bought separately don’t count. Subscribe the account, or use one that has a plan.

    {
  "error": "Account US:[email protected] needs a paid Dreamina subscription for seedance-2.5: Dreamina refused the job because the account has no plan that includes it (credits bought separately don't count). Subscribe at https://dreamina.capcut.com/pricing/dreamina-price, or use an account that has one. Dreamina: credit miss role",
  "ret": "5006",
  "errmsg": "credit miss role",
  "account": "US:[email protected]",
  "model": "seedance-2.5"
}

    ret 4010: Dreamina asks for its one-time “Before you continue” confirmation on this account (currently Seedance 2.x). We accept it automatically and resubmit, so you only get this response when that did not clear it. Retry once. If it persists, sign in at dreamina.capcut.com with the account, start a generation with the same model, click Agree, then retry.

    {
  "error": "Dreamina asks for a one-time confirmation on account US:[email protected] and our automatic confirmation did not clear it. Retry once. If it persists, sign in at https://dreamina.capcut.com with this account, start a seedance-2.0 generation, click Agree on \"Before you continue\", then retry. Dreamina: Couldn't generate. Refresh and agree to the confirmation before you continue.",
  "ret": "4010",
  "errmsg": "Couldn't generate. Refresh and agree to the confirmation before you continue.",
  "failCode": "4010",
  "account": "US:[email protected]",
  "model": "seedance-2.0"
}

  • 429 Too Many Requests

    All accounts at maximum capacity. Wait for current jobs to complete or increase maxJobs.

    {
  "error": "All accounts at capacity"
}

  • 596 Session Error

    Account session expired. Re-add the account using POST /accounts with correct credentials.

    {
  "error": "Session expired"
}

Model 

{
  jobid: string                    // Unique job identifier
  type: 'video'                    // Job type
  status: 'created'                // Initial status
  model: string                    // Model used
  created: string                  // ISO 8601 timestamp
  request: {
    prompt?: string
    model: string
    ratio?: string                 // "16:9", "9:16", etc.
    duration?: number              // Seconds
    inputMode: string              // "prompt" | "first_frame" | "end_frame" | "multi_frame" | "unified_edit" | "long_video" | "edit"
    resolution?: string            // "480p", "720p", "1080p", "2k" or "4k"
    firstFrameRef?: string         // assetRef (image) for first frame
    endFrameRef?: string           // assetRef (image) for end frame
    frame_1_imageRef?: string      // Keyframe refs (multi_frame mode)
    frame_1_prompt?: string
    frame_1_duration?: number
    frame_2_imageRef?: string
    frame_2_prompt?: string
    frame_2_duration?: number
    omni_1_imageRef?: string       // Omni image refs (use @image1 in prompt)
    omni_2_imageRef?: string
    omni_1_videoRef?: string       // Omni video refs (use @video1 in prompt)
    omni_1_audioRef?: string       // Omni audio refs (use @audio1 in prompt)
    replyUrl?: string              // Webhook URL
    replyRef?: string              // Custom reference
  }
  response: {
    forecastCost: number           // Estimated generation time in seconds, NOT a price
  }
  code: number                     // HTTP status code
  error?: string                   // Error message
}

Examples 

  • # Text-to-video
curl -X POST \
     -H "Authorization: Bearer YOUR_API_TOKEN" \
     -H "Content-Type: application/json" \
     -d '{
       "prompt": "A serene mountain landscape at sunset",
       "model": "seedance-2.0",
       "ratio": "16:9",
       "duration": 5
     }' \
     "https://api.useapi.net/v1/dreamina/videos"

# Image-to-video (first frame — ratio auto-detected from image)
curl -X POST \
     -H "Authorization: Bearer YOUR_API_TOKEN" \
     -H "Content-Type: application/json" \
     -d '{
       "prompt": "Camera slowly pans across the scene",
       "model": "seedance-2.0",
       "firstFrameRef": "CA:[email protected]:w685:h900:s86866-uri:tos-useast5-i-wopfjsm1ax-tx/abc123",
       "duration": 5
     }' \
     "https://api.useapi.net/v1/dreamina/videos"

# Omni Reference (image + audio)
curl -X POST \
     -H "Authorization: Bearer YOUR_API_TOKEN" \
     -H "Content-Type: application/json" \
     -d '{
       "prompt": "@image1 slowly moving forward while @audio1 plays in background",
       "model": "seedance-2.0",
       "ratio": "16:9",
       "duration": 4,
       "omni_1_imageRef": "CA:[email protected]:w2560:h1440:s580914-uri:tos-alisg-i-wopfjsm1ax-sg/abc123",
       "omni_1_audioRef": "CA:[email protected]:d14040-vid:v10762g50003d77l5uvog65nl4esjs2g"
     }' \
     "https://api.useapi.net/v1/dreamina/videos"

  • const token = 'YOUR_API_TOKEN';
const apiUrl = 'https://api.useapi.net/v1/dreamina/videos';

// Text-to-video
const response = await fetch(apiUrl, {
  method: 'POST',
  headers: {
    'Authorization': `Bearer ${token}`,
    'Content-Type': 'application/json'
  },
  body: JSON.stringify({
    prompt: 'A serene mountain landscape at sunset',
    model: 'seedance-2.0',
    ratio: '16:9',
    duration: 5
  })
});

const result = await response.json();
console.log('Job created:', result.jobid);

// Poll for completion
const poll = async (jobid) => {
  while (true) {
    const res = await fetch(`https://api.useapi.net/v1/dreamina/videos/${jobid}`, {
      headers: { 'Authorization': `Bearer ${token}` }
    });
    const job = await res.json();
    console.log('Status:', job.status);

    if (job.status === 'completed') {
      console.log('Video URL:', job.response.videoUrl);
      return job;
    }
    if (job.status === 'failed') throw new Error(job.error);

    await new Promise(r => setTimeout(r, 10000));
  }
};

const completed = await poll(result.jobid);

  • import requests
import time

token = 'YOUR_API_TOKEN'
api_url = 'https://api.useapi.net/v1/dreamina/videos'

headers = {
    'Authorization': f'Bearer {token}',
    'Content-Type': 'application/json'
}

# Text-to-video
data = {
    'prompt': 'A serene mountain landscape at sunset',
    'model': 'seedance-2.0',
    'ratio': '16:9',
    'duration': 5
}

response = requests.post(api_url, headers=headers, json=data)
result = response.json()
print(f"Job created: {result['jobid']}")

# Poll for completion
jobid = result['jobid']
while True:
    job = requests.get(
        f'https://api.useapi.net/v1/dreamina/videos/{jobid}',
        headers={'Authorization': f'Bearer {token}'}
    ).json()
    print(f"Status: {job['status']}")

    if job['status'] == 'completed':
        print(f"Video URL: {job['response']['videoUrl']}")
        break
    if job['status'] == 'failed':
        raise Exception(job.get('error'))

    time.sleep(10)

Try It