Seedance 2.0

Seedance 2.0 is ByteDance’s latest video generation model available through Upmore API. It supports text-to-video, image-to-video, multimodal reference input, video editing, video extension, and synchronized audio generation.

Key capabilities

Feature	Description
Text-to-video	Generate video from text prompts
Image-to-video (first frame)	Use an image as the first frame
Image-to-video (first + last frame)	Use two images as first and last frames
Multimodal reference	Combine images, videos, and audio as references (1–9 images, up to 3 videos, up to 3 audio clips)
Video editing	Modify elements in an existing video using reference images
Video extension	Extend and concatenate reference videos
Audio generation	Auto-generate synchronized voice, sound effects, and background music
Web search	Enhance generation with real-time internet content (text-to-video only)
Return last frame	Retrieve the last frame of generated video

Output specifications

Property	Value
Resolution	480p, 720p, 1080p
Aspect ratio	16:9, 4:3, 1:1, 3:4, 9:16, 21:9, adaptive
Duration	4–15 seconds
Format	mp4

Workflow

POST /v1/video/generations  →  task_id
Poll GET /v1/video/generations/{task_id}  →  status
When status = "succeeded"  →  download video URL (valid 24 hours)

Examples

[!WARNING] Read this before using the examples below. Through the Upmore gateway (/v1/video/generations), the example bodies on this page do not work as-is: top-level content, duration, generate_audio, ratio, … are envelope fields, and only model, prompt, images, seconds and metadata are honored. See Passthrough via metadata — that is the format that actually reaches the model.

Passthrough via `metadata`

The gateway forwards everything inside metadata to the model verbatim — this is how you use the full native Seedance API (roles, duration, all options) through Upmore. Send a JSON body like this:

{
  "model": "doubao-seedance-2-0",
  "prompt": "A cat playing piano in a sunlit room, cinematic lighting",
  "metadata": {
    "content": [
      { "type": "image_url", "image_url": { "url": "https://example.com/first.png" }, "role": "first_frame" },
      { "type": "image_url", "image_url": { "url": "https://example.com/character.png" }, "role": "reference_image" },
      { "type": "image_url", "image_url": { "url": "https://example.com/last.png" }, "role": "last_frame" }
    ],
    "duration": 8,
    "resolution": "1080p",
    "ratio": "16:9",
    "generate_audio": true,
    "watermark": false,
    "seed": 42,
    "camera_fixed": false
  }
}

Rules and pitfalls:

Images go in metadata.content with a role, not in top-level images. Top-level images are forwarded as plain reference images without role, so first-frame/last-frame locking and character references do not work through that path. If metadata.content is present, it replaces any top-level images entirely — do not mix the two.
Duration goes in metadata.duration. A top-level duration field is dropped by the gateway envelope; only metadata.duration (or top-level seconds as a string) controls the video length. Sending "duration": 45 at the top level silently falls back to the model default.
The prompt goes in the top-level prompt field. It is injected into the content array as the single text item; text items inside metadata.content are dropped. metadata.model is ignored (removed before passthrough).
Use a JSON body, not multipart form-data — the content array only round-trips correctly as JSON.
return_last_frame is currently not usable through this endpoint: the model generates the last frame, but task polling only returns video_url.
Billing is unaffected: resolution tiers and video-input surcharges are read from metadata, and settlement recalculates from the actual token usage reported by the model.

Text-to-video

curl https://api.upmore.net/v1/video/generations \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "doubao-seedance-2-0",
    "content": [
      {
        "type": "text",
        "text": "A cat playing piano in a sunlit room, cinematic lighting"
      }
    ],
    "generate_audio": true,
    "ratio": "16:9",
    "duration": 8
  }'

import requests
import time

headers = {"Authorization": "Bearer YOUR_API_KEY", "Content-Type": "application/json"}

# Step 1: create task
resp = requests.post(
    "https://api.upmore.net/v1/video/generations",
    headers=headers,
    json={
        "model": "doubao-seedance-2-0",
        "content": [
            {"type": "text", "text": "A cat playing piano in a sunlit room, cinematic lighting"}
        ],
        "generate_audio": True,
        "ratio": "16:9",
        "duration": 8
    }
)
task_id = resp.json()["task_id"]

# Step 2: poll for result
while True:
    result = requests.get(
        f"https://api.upmore.net/v1/video/generations/{task_id}",
        headers=headers
    ).json()
    status = result["data"]["status"]
    if status == "SUCCESS":
        print("Video URL:", result["data"]["data"]["content"]["video_url"])
        break
    elif status == "FAILED":
        print("Failed:", result["data"].get("fail_reason"))
        break
    time.sleep(5)

Multimodal reference (image + video + audio)

cURL

curl https://api.upmore.net/v1/video/generations \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "doubao-seedance-2-0",
    "content": [
      {
        "type": "text",
        "text": "Extend the video with a smooth camera pan, festive atmosphere"
      },
      {
        "type": "image_url",
        "image_url": {"url": "https://example.com/character.jpg"},
        "role": "reference_image"
      },
      {
        "type": "video_url",
        "video_url": {"url": "https://example.com/clip.mp4"},
        "role": "reference_video"
      },
      {
        "type": "audio_url",
        "audio_url": {"url": "https://example.com/bgm.mp3"},
        "role": "reference_audio"
      }
    ],
    "generate_audio": true,
    "ratio": "16:9",
    "duration": 11
  }'

Video editing

cURL

curl https://api.upmore.net/v1/video/generations \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "doubao-seedance-2-0",
    "content": [
      {
        "type": "text",
        "text": "Replace the water bottle in the video with the perfume bottle from the image, keep camera movement unchanged"
      },
      {
        "type": "image_url",
        "image_url": {"url": "https://example.com/perfume.jpg"},
        "role": "reference_image"
      },
      {
        "type": "video_url",
        "video_url": {"url": "https://example.com/original.mp4"},
        "role": "reference_video"
      }
    ],
    "generate_audio": true,
    "ratio": "16:9",
    "duration": 5
  }'

Web search enhanced (text-to-video only)

cURL

curl https://api.upmore.net/v1/video/generations \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "doubao-seedance-2-0",
    "content": [
      {
        "type": "text",
        "text": "Macro shot of cherry blossoms in spring, petals falling slowly"
      }
    ],
    "generate_audio": true,
    "ratio": "16:9",
    "duration": 11,
    "tools": [{"type": "web_search"}]
  }'

Parameters

Parameter	Type	Required	Description
`model`	string	Yes	`doubao-seedance-2-0`
`content`	array	Yes	Input content array (text, image_url, video_url, audio_url items)
`content[].type`	string	Yes	`text`, `image_url`, `video_url`, or `audio_url`
`content[].text`	string	Text items	Text prompt (max 500 Chinese chars / 1000 English words)
`content[].image_url.url`	string	Image items	Image URL, Base64 data URI, or `asset://<ID>`
`content[].video_url.url`	string	Video items	Video URL or `asset://<ID>` (mp4/mov, max 50 MB, 2–15s)
`content[].audio_url.url`	string	Audio items	Audio URL, Base64 data URI, or `asset://<ID>` (wav/mp3, max 15 MB)
`content[].role`	string	Conditional	`first_frame`, `last_frame`, `reference_image`, `reference_video`, `reference_audio`
`generate_audio`	boolean	No	Generate synchronized audio. Default: `true`
`resolution`	string	No	`480p`, `720p`, or `1080p`. Default: `720p`
`ratio`	string	No	`16:9`, `4:3`, `1:1`, `3:4`, `9:16`, `21:9`, `adaptive`. Default: `adaptive`
`duration`	integer	No	4–15 seconds. Default: `5`
`tools`	array	No	`[{"type": "web_search"}]` for web search (text-to-video only)
`watermark`	boolean	No	Add watermark. Default: `false`

[!NOTE] The table above documents the native Seedance parameters (as carried inside metadata). Through the gateway envelope, pass each of them as metadata.<name> — e.g. metadata.duration, metadata.content[].role — as described in Passthrough via metadata.

Input modes

Mode	Content items	`role` values
Text-to-video	1× text	—
Image-to-video (first frame)	text (optional) + 1× image_url	`first_frame` or omit
Image-to-video (first + last frame)	text (optional) + 2× image_url	`first_frame` + `last_frame`
Multimodal reference	text (optional) + image/video/audio	`reference_image`, `reference_video`, `reference_audio`
Video editing	text + image_url + video_url	`reference_image` + `reference_video`
Video extension	text + video_url(s)	`reference_video`

Note: First frame, first+last frame, and multimodal reference are mutually exclusive — do not mix them in the same request.

Resolution pixel values

Resolution	16:9	4:3	1:1	3:4	9:16	21:9
480p	864x496	752x560	640x640	560x752	496x864	992x432
720p	1280x720	1112x834	960x960	834x1112	720x1280	1470x630
1080p	1920x1080	1664x1248	1440x1440	1248x1664	1080x1920	2208x944

API Reference

View the interactive API playground.

Getting started

Native Compatible Endpoints

Model APIs

Key capabilities

Output specifications

Workflow

Examples

Passthrough via `metadata`

Text-to-video

Multimodal reference (image + video + audio)

Video editing

Web search enhanced (text-to-video only)

Parameters

Input modes

Resolution pixel values

API Reference

​Key capabilities

​Output specifications

​Workflow

​Examples

​Passthrough via metadata

​Text-to-video

​Multimodal reference (image + video + audio)

​Video editing

​Web search enhanced (text-to-video only)

​Parameters

​Input modes

​Resolution pixel values

API Reference

Key capabilities

Output specifications

Workflow

Examples

Passthrough via `metadata`

Text-to-video

Multimodal reference (image + video + audio)

Video editing

Web search enhanced (text-to-video only)

Parameters

Input modes

Resolution pixel values