seedance-2.0-sz / seedance-2.5-sz

seedance-2.0-sz, seedance-2.0-sz-fast, seedance-2.0-sz-mini and seedance-2.5-sz Creates asynchronous video task using /v1/videos in a single way. request parameters follow Volcengine Seedance The syntax of field for the official multimodular API; references[] for MoonNexAI is recommended for new access Use the common reference format; the server converts it to the multimodal structure required by the model.

Volcengine official parameters compatible

All four SZ public models support the parameters for submitting Volcengine official shapes in POST /v1/videos. That is, Clients can continue to use content[], ratio, resolution, duration, generate_audio et al. field requires only that the model name be changed to the current open SZ model name of MoonNexAI. references[] is the uniform formulation recommended by MoonNexAI for new access and cross-SDK calls.

Entry	Available Model Name	Annotations
`POST /v1/videos`	`seedance-2.0-sz`, `seedance-2.0-sz-fast`, `seedance-2.0-sz-mini`, `seedance-2.5-sz`	MoonNexAI unified entrance; compatible with the syntax of Volcengine official field and validated according to the capabilities of the selected SZ model.
`POST /api/v3/contents/generations/tasks`	Use MoonNexAI public model ID: `seedance-2.0-sz`, `seedance-2.0-sz-fast`, `seedance-2.0-sz-mini`, `seedance-2.5-sz`	The official Volcengine path is compatible with the shape of the parameter; only the request path is replaced, and `model` is still completed on this page ' s public model list.

The following rules for the unification of field:

MoonNexAI Universal field	Volcengine official shape field	Treatment
`prompt`	`text` Entry for `content[]`	Text entries are replaced by service-ends when `prompt` is passed only; text must be available for pure text-to-video.
`references[]`	Media entry for `content[]`	Press `media_type` and `role` to convert to `image_url`, `video_url` or `audio_url`.
`size`	`resolution`	Both must be consistent when they are imported.
`aspect_ratio`	`ratio`	Both must be consistent when they are imported.
`first_frame` / `last_frame`	Media entry `role`	`first_frame`, `last_frame`, cannot be used in a common mix of reference media assets.

The following conflicts return 400 directly before creating task, and will not be silently discarded field:

content appears with references.
prompt appears at the same time as content[].text but does not match.
size appears at the same time as resolution, or aspect_ratio and ratio but does not match the values.
duration appears alongside frames; the SZ model does not support frames using duration.

Visible false, 0, 0.0 and duration: -1 will be retained. service_tier in SZ default;flex does not support. priority, execution_expires_after, safety_identifier et al. field will still be verified in the official format of Volcengine, and the actual validity will be based on model capacity and account configuration.

Model Selection

Models	Positioning	resolution	duration
`seedance-2.0-sz`	Standard Multimodular Video Generation	`480P`, `720P`, `1080P`, `4K`	`4` - `15` seconds, or smart duration `-1`
`seedance-2.0-sz-fast`	Rapid Multimodular Video Generation	`480P`, `720P`	`4` - `15` seconds, or smart duration `-1`
`seedance-2.0-sz-mini`	Lightweight multimodular video generation	`480P`, `720P`	`4` - `15` seconds, or smart duration `-1`
`seedance-2.5-sz`	New generation multimodular video generation	`480P`, `720P`, `1080P`	`4` - `30` seconds, or smart duration `-1`

Actual available models, resolution, duration and aspect ratio are based on GET /v1/models return capability and console.

Capacity boundaries

Capacity	`seedance-2.0-sz` Series	`seedance-2.5-sz`
text-to-video	Support	Support
Picture reference	Up to 9	Maximum 30
Video reference	Up to 3	Maximum 10
Audio Reference	maximum 3; photo or video required	maximum 10; photo or video required
Mixed media assets	Up to 15 references media assets	Up to 50 references media assets
First / last frame	Support	Support
Video Edit / Extension	Support	Based on model capabilities
Realist reference.	Support; content security clearance still required	Support; content security clearance still required

Audio cannot be entered separately. You can submit any legitimate combination of text, pictures, videos, audio, but at least one picture or video is required to submit audio references. media assets URLs must be a stable public media straight chain that can be downloaded without login, cookies or extra authentication Header.

request parameters

Parameters are divided into two columns. The two columns represent the same set of video capabilities: the common column is suitable for MoonNexAI SDK and cross-model access, and the official column is suitable for the direct re-use of Volcengine content[] request. The two columns are not used in the same request.

MoonNexAI Unified Common Parameters	Volcengine-compatible compatible parameters
`model`	`model` (Book still the MoonNexAI public model ID)
`prompt`	`text` entries in `content[]`
`references[]`: `media_type`, `role`, `url`, `alias`	`content[]`: `image_url`, `video_url`, `audio_url` entries, and fill `role`, unique `alias`
`aspect_ratio`	`ratio`
`size`	`resolution`
`duration`, `generate_audio`, `watermark`, `seed`, `callback_url`	Same official field; `camera_fixed`, `return_last_frame`, `omni_reference_task_type`, `output_format` etc.

MoonNexAI Unified Universal Video Parameters

field	Type	Annotations
`model`	string	It's gonna be filled. Four SZ public models.
`prompt`	string	Video description, with `@alias` references `references[]` media assets.
`references[]`	array	Harmonizes the media assets array; entries use `media_type`, `role`, `url` and optional `alias`.
`aspect_ratio`	string	Video aspect ratio, e.g. `16:9`, `9:16`, `1:1` or `adaptive`.
`size`	string	Common resolution field, mapped to the model's `resolution` by the server.
`duration`	integer	2.0 Support `4` - `15` sec, 2.5 Support `4` - `30` sec; `-1` is smart duration.
`generate_audio`	boolean	Whether or not to generate a track.
`watermark`, `seed`, `callback_url`	boolean/integer/string	General control field, whether model capabilities are used.

Volcengine-compatible compatible parameters

field	Type	Annotations
`model`	string	To fill in. Fill in the MoonNexAI public model ID listed on this page.
`content[]`	array	Official multimodular arrays, with `text`, `image_url`, `video_url` or `audio_url` type of entry.
`ratio`	string	Compatible aspect ratio, such as `16:9`, `9:16`, `1:1`, `21:9` or `adaptive`.
`resolution`	string	Compatible resolution, such as `480p`, `720p`, `1080p` or `4k`.
`duration`	integer	Target duration; SZ does not support `frames`.
`generate_audio`, `watermark`, `seed`, `callback_url`	boolean/integer/string	Consistent with the semantic Volcengine-compatible, visible `false`, `0` will be retained.
`camera_fixed`, `return_last_frame`, `omni_reference_task_type`, `output_format`	mixed	Use only when the SZ type is selected.

generateAudio is a historical alias, semantics are the same as generate_audio, and the new client should use the underlined field. Visible false will not be discarded as default value.

Create Example

1. Text-only video generation

bash

curl https://moonnexai.com/v1/videos \
  -H "Authorization: Bearer <MOONNEXAI_API_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seedance-2.0-sz-fast",
    "prompt": "一只橘猫在窗边伸懒腰，镜头缓慢推近，午后自然光，写实风格。",
    "duration": 4,
    "ratio": "16:9",
    "resolution": "480p",
    "generate_audio": true,
    "watermark": false
  }'

2. Single image image-to-video (Closure sound)

bash

curl https://moonnexai.com/v1/videos \
  -H "Authorization: Bearer <MOONNEXAI_API_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seedance-2.0-sz-mini",
    "prompt": "让人物自然转身并向镜头微笑，保持服装和脸部特征稳定。",
    "duration": 4,
    "aspect_ratio": "9:16",
    "resolution": "480p",
    "generate_audio": false,
    "references": [
      {
        "media_type": "image",
        "role": "reference_image",
        "url": "https://example.com/person.png",
        "alias": "人物"
      }
    ]
  }'

3. Picture + Video + Audio Group (to generate audio video)

json

{
  "model": "seedance-2.0-sz",
  "prompt": "使用@人物的外观，参考@动作的运镜，并将@音乐作为背景音乐。",
  "duration": 8,
  "ratio": "16:9",
  "resolution": "720p",
  "generate_audio": true,
  "references": [
    { "media_type": "image", "role": "reference_image", "url": "https://example.com/person.png", "alias": "人物" },
    { "media_type": "video", "role": "reference_video", "url": "https://example.com/motion.mp4", "alias": "动作" },
    { "media_type": "audio", "role": "reference_audio", "url": "https://example.com/music.mp3", "alias": "音乐" }
  ]
}

4. 1 + 1 audio + 1 video (official `content[]` shape and `Asset://`)

Prepare three types of media assets first through /v1/assets/uploads mode=asset target model and confirm each media assets returns all generation_ready=true and preparation_status=ready. Three are still to be used when task is created References for visible writing request body:

json

{
  "model": "seedance-2.5-sz",
  "content": [
    {
      "type": "image_url",
      "role": "reference_image",
      "image_url": { "url": "Asset://asset_image_demo" }
    },
    {
      "type": "audio_url",
      "role": "reference_audio",
      "audio_url": { "url": "Asset://asset_audio_demo" }
    },
    {
      "type": "video_url",
      "role": "reference_video",
      "video_url": { "url": "Asset://asset_video_demo" }
    }
  ],
  "duration": 5,
  "ratio": "16:9",
  "resolution": "720P",
  "generate_audio": true,
  "watermark": false
}

Audio reference must be submitted with a picture or video reference, and cannot be created as a separate audio input task. Three types of media assets can also Change to references[]; the same request does not submit content and references.

Multigraph + MultiVideo + Multi Audio (MoonNexAI Unified Generic Parameters)

The same scene is expressed in a single references[] format, suitable for MoonNexAI SDK and cross-model calls:

json

{
  "model": "seedance-2.5-sz",
  "prompt": "保持 @Image1、@Image2 的主体特征，参考 @Video1、@Video2 的动作，并融合 @Audio1、@Audio2 的声音氛围。",
  "references": [
    { "media_type": "image", "role": "reference_image", "url": "https://example.com/image-1.png", "alias": "Image1" },
    { "media_type": "image", "role": "reference_image", "url": "https://example.com/image-2.png", "alias": "Image2" },
    { "media_type": "video", "role": "reference_video", "url": "https://example.com/video-1.mp4", "alias": "Video1" },
    { "media_type": "video", "role": "reference_video", "url": "https://example.com/video-2.mp4", "alias": "Video2" },
    { "media_type": "audio", "role": "reference_audio", "url": "https://example.com/audio-1.mp3", "alias": "Audio1" },
    { "media_type": "audio", "role": "reference_audio", "url": "https://example.com/audio-2.mp3", "alias": "Audio2" }
  ],
  "duration": 8,
  "aspect_ratio": "16:9",
  "size": "720p",
  "generate_audio": true
}

5a. Multiple images + videos + audio (Volcengine-compatible parameters)

The Volcengine-compatible Seedance shape supports the combination of multiple pictures, videos and audio in the same request. The following example uses a combination of 2 + 2 + 2; please add or delete entries based on the capacity ceiling of the selected SZ model.

json

{
  "model": "seedance-2.5-sz",
  "content": [
    { "type": "text", "text": "保持 @Image1、@Image2 的主体特征，参考 @Video1、@Video2 的动作，并融合 @Audio1、@Audio2 的声音氛围。" },
    { "type": "image_url", "role": "reference_image", "image_url": { "url": "https://example.com/image-1.png" }, "alias": "Image1" },
    { "type": "image_url", "role": "reference_image", "image_url": { "url": "https://example.com/image-2.png" }, "alias": "Image2" },
    { "type": "video_url", "role": "reference_video", "video_url": { "url": "https://example.com/video-1.mp4" }, "alias": "Video1" },
    { "type": "video_url", "role": "reference_video", "video_url": { "url": "https://example.com/video-2.mp4" }, "alias": "Video2" },
    { "type": "audio_url", "role": "reference_audio", "audio_url": { "url": "https://example.com/audio-1.mp3" }, "alias": "Audio1" },
    { "type": "audio_url", "role": "reference_audio", "audio_url": { "url": "https://example.com/audio-2.mp3" }, "alias": "Audio2" }
  ],
  "duration": 8,
  "ratio": "16:9",
  "resolution": "720P",
  "generate_audio": true,
  "output_format": "mp4"
}

2.0 and 2.5 Use the same official content[] structure; only the number of duration, resolution and media assets are to be adjusted to model capabilities.

6. first frame + last frame

json

{
  "model": "seedance-2.0-sz",
  "prompt": "从室内平滑过渡到户外，保持主体和色彩连续。",
  "duration": 6,
  "ratio": "16:9",
  "resolution": "1080p",
  "generate_audio": false,
  "content": [
    { "type": "image_url", "role": "first_frame", "image_url": { "url": "https://example.com/first.jpg" } },
    { "type": "image_url", "role": "last_frame", "image_url": { "url": "https://example.com/last.jpg" } }
  ]
}

7. Text on `content[]` and use smart duration

json

{
  "model": "seedance-2.5-sz",
  "content": [
    { "type": "text", "text": "一段电影感的城市夜景航拍。" }
  ],
  "duration": -1,
  "ratio": "adaptive",
  "resolution": "1080p",
  "seed": 42,
  "camera_fixed": false,
  "generate_audio": true,
  "watermark": false,
  "callback_url": "https://example.com/webhooks/moonnexai"
}

8. Seedance 2.5 Video Edit and Output MOV

json

{
  "model": "seedance-2.5-sz",
  "prompt": "保留主体动作，调整为清晨自然光，并延续原视频的镜头节奏。",
  "duration": -1,
  "ratio": "adaptive",
  "resolution": "1080p",
  "omni_reference_task_type": "edit",
  "output_format": "mov",
  "generate_audio": true,
  "references": [
    {
      "media_type": "video",
      "role": "reference_video",
      "url": "https://example.com/source.mp4",
      "alias": "原视频"
    }
  ]
}

All four SZ public models use duration. Do not send frames or set unnecessary fields to null. 2.5 edit / extend requests must satisfy the table above. Even if creation succeeds, the task may fail asynchronously if references or prompts do not match the selected task type.

Reference media assets role

`role`	Purpose
`reference_image`	Normal Picture Reference
`first_frame`	Picture of first frame
`last_frame`	Picture of last frame
`reference_video`	Actions, images or mirror reference video
`reference_audio`	Audio reference; must be submitted with pictures or videos

@alias in prompt is used only to describe the semantics of media assets, and the service is still based on the structured role of references[] or content[]. When strict first and last frames is required, first_frame/last_frame should be used in a visible manner.

Use anonymously accessible URLs from your own storage or ready MoonNexAI Asset://asset_xxx references. Public URLs do not require /v1/assets/uploads; use the asset API when hosting, reuse or managed asset lifecycles are needed. Both methods remain subject to format, duration, size and content checks.

Query task and Download Results

Save returned id or task_id, query status after surrender, read result address only if status is succeeded and no error:

bash

curl https://moonnexai.com/v1/videos/<TASK_ID> \
  -H "Authorization: Bearer <MOONNEXAI_API_KEY>"

Successful response may return the result address in result_url, video_url, url, output.url, output[].url or content.video_url. Please use the complete URL that API actually returns; it may be either an approved direct media address or a media address managed by MoonNexAI. Do not collage, replace or delete signature parameters.

Only successful response can download with compatible content API invert when URLs are not available:

bash

curl -L https://moonnexai.com/v1/videos/<TASK_ID>/content \
  -H "Authorization: Bearer <MOONNEXAI_API_KEY>" \
  -o result.mp4

Do not read a URL for pending, submitted, running, failed or cancelled. Read error after failure. Result URLs may contain temporary signatures and expire; download or archive successful results promptly.

billing Method

This family is billed by Seedance token usage: the model's token price multiplied by task token usage. Tasks containing video references use the video-input billing category; other tasks use the category without video input. Check the model list, console and billing records for current prices.

Conversation APIs

Embeddings

Audio

Models

GPT Image

Gemini Nano

Ideogram Image

Midjourney Task

Seedance-2 Video

Moon-2 Video

seedance-2.0-sz / seedance-2.5-sz

Volcengine official parameters compatible

Model Selection

Capacity boundaries

request parameters

MoonNexAI Unified Universal Video Parameters

Volcengine-compatible compatible parameters

Create Example

1. Text-only video generation

2. Single image image-to-video (Closure sound)

3. Picture + Video + Audio Group (to generate audio video)

4. 1 + 1 audio + 1 video (official `content[]` shape and `Asset://`)

Multigraph + MultiVideo + Multi Audio (MoonNexAI Unified Generic Parameters)

5a. Multiple images + videos + audio (Volcengine-compatible parameters)

6. first frame + last frame

7. Text on `content[]` and use smart duration

8. Seedance 2.5 Video Edit and Output MOV

Reference media assets role

Query task and Download Results

billing Method

Relevant Pages

Ideogram Image

Midjourney Task

Seedance-2 Video

Moon-2 Video

seedance-2.0-sz / seedance-2.5-sz ​

Volcengine official parameters compatible ​

Model Selection ​

Capacity boundaries ​

request parameters ​

MoonNexAI Unified Universal Video Parameters ​

Volcengine-compatible compatible parameters ​

Create Example ​

1. Text-only video generation ​

2. Single image image-to-video (Closure sound) ​

3. Picture + Video + Audio Group (to generate audio video) ​

4. 1 + 1 audio + 1 video (official content[] shape and Asset://) ​

Multigraph + MultiVideo + Multi Audio (MoonNexAI Unified Generic Parameters) ​

5a. Multiple images + videos + audio (Volcengine-compatible parameters) ​

6. first frame + last frame ​

7. Text on content[] and use smart duration ​

8. Seedance 2.5 Video Edit and Output MOV ​

Reference media assets role ​

Query task and Download Results ​

billing Method ​

Relevant Pages ​

seedance-2.0-sz / seedance-2.5-sz

Volcengine official parameters compatible

Model Selection

Capacity boundaries

request parameters

MoonNexAI Unified Universal Video Parameters

Volcengine-compatible compatible parameters

Create Example

1. Text-only video generation

2. Single image image-to-video (Closure sound)

3. Picture + Video + Audio Group (to generate audio video)

4. 1 + 1 audio + 1 video (official `content[]` shape and `Asset://`)

Multigraph + MultiVideo + Multi Audio (MoonNexAI Unified Generic Parameters)

5a. Multiple images + videos + audio (Volcengine-compatible parameters)

6. first frame + last frame

7. Text on `content[]` and use smart duration

8. Seedance 2.5 Video Edit and Output MOV

Reference media assets role

Query task and Download Results

billing Method

Relevant Pages