Seedance 2.5 API: Draft-to-1080p Video Generation

Overview

Seedance 2.5 is a video generation model for prompt-only video, first-and-last-frame animation, and multimodal reference workflows. Use this page when you need up to 30-second video generation with optional image, video, or audio references, or when you want to review low-cost 480p Drafts before producing selected results in 1080p.

Capability	Value
Model ID	`seedance-2-5`
Modes	`text-to-video`, `first_and_last_frames`, `omni_reference`
Prompt	Required non-empty string; no API-side maximum length
Resolutions	`480p`, `720p`, `1080p`
Durations	Any integer from `4` through `30` seconds, or `-1` for dynamic duration
Aspect ratios	`auto`, `16:9`, `4:3`, `1:1`, `3:4`, `9:16`, `21:9`
Reference images	1-2 URLs for `first_and_last_frames`; 1-30 URLs for `omni_reference`
Reference video/audio	1-10 URLs each for `omni_reference`; each file 2-30 seconds, total max 30 seconds per media type
Draft workflow	Generate a 480p Draft with `draft=true`, then create one or more 1080p Finals with `draft_task_id`
Output	Video URL array in `resultJson.resultUrls`

In omni_reference mode with video references, the generation system infers whether your task is reference generation (creating a new video guided by reference media) or video editing (modifying elements within an existing reference video):

Video editing (modifying original clips): If your prompt instructs changes to a reference video (such as replacing subjects, altering outfits, or modifying scene elements), set duration to -1 and aspect_ratio to auto so output attributes adapt to the source clip.
Reference generation (creating new clips): If you only want to borrow motion, lighting, or style to create a new scene, describe the features to reference rather than commanding edits to the source footage.
Recommended fallback: If your prompt could be interpreted as editing or you are unsure how the model will classify it, pass duration: -1 and aspect_ratio: "auto" to prevent parameter mismatch failures.

Endpoint and authentication

Base URL:

https://api.apixo.ai/api/v1

Method	Endpoint	Purpose
`POST`	`/generateTask/seedance-2-5`	Submit a generation task
`GET`	`/statusTask/seedance-2-5?taskId={taskId}`	Poll task status and retrieve results

All requests require your APIXO API key:

Authorization: Bearer YOUR_API_KEY

Submit requests also require:

Content-Type: application/json

Recommended: Draft first, Final second

Save about 57% on a four-candidate workflow and roughly double iteration efficiency. Generate inexpensive 480p Drafts for review, then convert only the selected Draft into a 1080p Final. This avoids repeatedly paying for full-resolution candidates while preserving the Draft’s generation conditions, composition, motion, duration, and aspect ratio for the Final generation.

Step 1: Generate a 480p Draft

Set input.draft=true. resolution may be omitted because Draft generation is fixed to 480p.

curl -X POST "https://api.apixo.ai/api/v1/generateTask/seedance-2-5" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "request_type": "async",
    "input": {
      "draft": true,
      "mode": "text-to-video",
      "prompt": "a blue ball rolls from left to right across a white table",
      "duration": -1,
      "aspect_ratio": "16:9",
      "sound": true
    }
  }'

Poll the returned task ID until the Draft reaches state=success. Keep this APIXO task ID; it is the value used as draft_task_id in the next step.

Step 2: Generate the 1080p Final

Send the successful Draft task ID. resolution may be omitted because Final generation is fixed to 1080p.

curl -X POST "https://api.apixo.ai/api/v1/generateTask/seedance-2-5" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "request_type": "async",
    "input": {
      "draft_task_id": "DRAFT_TASK_ID",
      "output_format": "mp4",
      "watermark": false
    }
  }'

The Final inherits the Draft’s generation conditions. Do not resend prompt, reference media, mode, duration, aspect_ratio, or sound. A successful Draft can be reused to create multiple independent Finals for 7 days. Each Final returns its own task ID and is charged to the API key that submits that Final request.

Why this workflow saves about 57%

For a no-video-reference workflow that compares four candidates of the same duration:

Four direct 1080p candidates: 4 × $0.800 = $3.200 per output second
Four 480p Drafts + one 1080p Final: (4 × $0.140) + $0.800 = $1.360 per output second
Savings: ($3.200 - $1.360) / $3.200 = 57.5%

The exact savings depend on candidate count, output duration, and whether video references are used. The largest efficiency gain comes from reviewing several Drafts and generating 1080p only for selected results.

Copy-paste async quickstart

This minimal request submits a text-to-video task and returns a taskId.

curl -X POST "https://api.apixo.ai/api/v1/generateTask/seedance-2-5" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "request_type": "async",
    "input": {
      "mode": "text-to-video",
      "prompt": "a cinematic night city drone shot with dramatic lighting",
      "resolution": "720p",
      "duration": 8,
      "aspect_ratio": "16:9",
      "sound": true
    }
  }'

Successful response:

{
  "code": 200,
  "message": "success",
  "data": {
    "taskId": "task_12345678"
  }
}

Save the taskId; you need it to poll for the final result.

Poll for result

curl -X GET "https://api.apixo.ai/api/v1/statusTask/seedance-2-5?taskId=task_12345678" \
  -H "Authorization: Bearer YOUR_API_KEY"

Processing response:

{
  "code": 200,
  "message": "success",
  "data": {
    "taskId": "task_12345678",
    "state": "processing",
    "createTime": 1767965610929
  }
}

Success response:

{
  "code": 200,
  "message": "success",
  "data": {
    "taskId": "task_12345678",
    "state": "success",
    "resultJson": "{\"resultUrls\":[\"https://file.apixo.ai/video.mp4\"]}",
    "createTime": 1767965610929,
    "completeTime": 1767965890929,
    "costTime": 280000
  }
}

Failed response:

{
  "code": 200,
  "message": "success",
  "data": {
    "taskId": "task_12345678",
    "state": "failed",
    "failCode": "GenerationError",
    "failMsg": "Generation failed",
    "createTime": 1767965610929,
    "completeTime": 1767965620132
  }
}

Parse resultJson after state becomes success:

const payload = JSON.parse(data.resultJson);
const videoUrls = payload.resultUrls;

Request body

Text-to-video

{
  "request_type": "async",
  "input": {
    "mode": "text-to-video",
    "prompt": "a slow tracking shot through a foggy pine forest at dawn",
    "resolution": "720p",
    "duration": 10,
    "aspect_ratio": "16:9",
    "sound": true
  }
}

Dynamic duration

Use duration=-1 when you want the generation system to decide the final output length.

{
  "request_type": "async",
  "input": {
    "mode": "text-to-video",
    "prompt": "a continuous cooking tutorial scene with natural camera transitions",
    "resolution": "720p",
    "duration": -1,
    "aspect_ratio": "auto",
    "sound": true
  }
}

First-and-last-frames

{
  "request_type": "async",
  "input": {
    "mode": "first_and_last_frames",
    "prompt": "turn the still frames into a smooth cinematic camera move",
    "image_urls": [
      "https://example.com/frame-start.png",
      "https://example.com/frame-end.png"
    ],
    "resolution": "720p",
    "duration": 5,
    "aspect_ratio": "auto",
    "sound": true
  }
}

Omni-reference

{
  "request_type": "async",
  "input": {
    "mode": "omni_reference",
    "prompt": "create a stylish product ad using the visual references and voice rhythm",
    "image_urls": [
      "https://example.com/product-shot.png"
    ],
    "video_urls": [
      "https://example.com/ref-motion.mp4"
    ],
    "audio_urls": [
      "https://example.com/ref-audio.mp3"
    ],
    "resolution": "480p",
    "duration": 6,
    "aspect_ratio": "auto",
    "sound": true,
    "output_format": "mp4",
    "watermark": false
  }
}

Parameters

string

default:"async"

Result delivery mode. Omit this field or use async for polling with statusTask, or use callback for webhook delivery.

string

Required when request_type is callback. Must be a public HTTPS URL that can receive the final task payload. See Webhooks.

object

required

Seedance 2.5 input parameters.

Show properties

boolean

default:"false"

Set to true to generate a 480p Draft. When enabled, resolution defaults to 480p and any other explicit resolution is rejected. Cannot be used together with draft_task_id.

string

APIXO task ID of a successful Seedance 2.5 Draft. Creates a 1080p Final that inherits the Draft’s generation conditions. The Draft must be no more than 7 days old and may be reused for multiple Finals. Cannot be used together with draft=true.

string

Generation mode. Required for ordinary generation and Draft creation; omit it when draft_task_id is used. Supported values: text-to-video, first_and_last_frames, omni_reference.

string

Text prompt describing the desired video. Required for ordinary generation and Draft creation; inherited and therefore prohibited when draft_task_id is used. The API trims whitespace and rejects empty prompts. When referencing video inputs, prompts that instruct modifications to the source video (video editing) require duration=-1 and aspect_ratio=auto. Passing -1 and auto is also recommended whenever editing intent might be ambiguous. Maximum length: 10000 characters.

string

default:"720p"

Output resolution. Ordinary generation defaults to 720p. Draft generation is fixed to 480p; Final generation with draft_task_id is fixed to 1080p. Values are trimmed and normalized to lowercase.

integer|string

default:"5"

Output duration in seconds. Supports integers from 4 through 30, including numeric strings such as "10", or -1 for dynamic duration. For new duration=-1 tasks, pre-charge uses 30 billable seconds without video references, or 30 + max(actual total reference video seconds, 20) with them. After success, the same billing rule is recalculated using actual output duration and any excess charge is refunded. See Billing for existing-task compatibility.

string

default:"auto"

Output aspect ratio. Supported values: auto, 16:9, 4:3, 1:1, 3:4, 9:16, 21:9. Use auto for first_and_last_frames and for prompts that may trigger video editing.

boolean

default:"true"

Whether to generate audio with the video. Must be a boolean when provided.

string

default:"mp4"

Output container format. Supported values: mp4, mov.

boolean

default:"false"

Whether to add a watermark. Must be a boolean when provided.

boolean

default:"false"

Enables content safety checking for this request. API requests default to false; the APIXO playground may use its own default. Must be a boolean when provided.

string[]

Reference image URLs. Required for first_and_last_frames with 1-2 images. Optional for omni_reference with up to 30 images.

string[]

Reference video URLs for omni_reference. Supports 1-10 URLs when provided. Each video must be 2-30 seconds, and total reference video duration cannot exceed 30 seconds.

string[]

Reference audio URLs for omni_reference. Supports 1-10 URLs when provided. Each audio file must be 2-30 seconds, and total reference audio duration cannot exceed 30 seconds.

Mode	Required media	Optional media	Billing note
`text-to-video`	None	None	Uses the no-video-reference rate.
`first_and_last_frames`	`image_urls` with 1-2 images	None	Uses the no-video-reference rate and requires `aspect_ratio=auto`.
`omni_reference`	At least one of `image_urls`, `video_urls`, or `audio_urls`	Any supported reference media	Uses video-reference billing when `video_urls` is provided.

Draft and Final input rules

Request	Required input	Allowed additional input	Fixed resolution
480p Draft	`draft=true`, plus the normal mode-specific generation fields	All ordinary fields supported by the selected mode	`480p`
1080p Final	`draft_task_id`	`resolution`, `output_format`, `watermark`	`1080p`

For Final requests, generation fields such as prompt, reference URLs, mode, duration, aspect_ratio, sound, and nsfw_checker are inherited from the Draft and must not be resent. draft=true and draft_task_id are mutually exclusive.

In omni_reference, audio-only requests are supported. The video-reference rate only applies when video_urls is provided.

Use public, directly accessible URLs for all reference media. Video and audio durations are checked before the task is submitted.

image_urls examples use URL strings. The API also accepts object items with a non-empty url field and normalizes them to URL strings.

Reference generation vs. video editing

When providing reference videos in omni_reference mode, Seedance 2.5 infers whether your goal is to generate an entirely new video inspired by the source or to edit the source clip directly.

Task Intent	Prompt Focus	Required Parameters	Example Prompt
Reference Generation (Create a new video)	Describe which characteristics to borrow from the reference (e.g., camera motion, character appearance, or rhythm) to compose a new scene.	Any valid `duration` (e.g. `5`, `10`) and `aspect_ratio` (e.g. `16:9`).	”A futuristic neon metropolis at dusk with rain-slicked streets, borrowing the dynamic drone tracking shot and pacing from @video1.”
Video Editing (Modify existing footage)	Instruct direct changes to elements in the source clip (e.g., swapping objects, modifying clothes, removing elements, altering backgrounds).	Must set `"duration": -1` and `"aspect_ratio": "auto"`.	”In @video1, change the person’s blue jacket into a black leather biker jacket while keeping all movement and environment unchanged.”

Best practice when intent is ambiguous: If your prompt combines new scene descriptions with elements referenced from an existing video, or if you are unsure whether the generation system will classify the prompt as editing, set "duration": -1 and "aspect_ratio": "auto". This allows the engine to adapt dynamically to the reference footage rather than failing due to fixed duration or aspect ratio constraints.

Response format

Submit task response

POST /generateTask/seedance-2-5 returns a task ID when the task is accepted:

integer

API status code. 200 means the task was accepted.

string

Human-readable status message.

string

Unique task identifier used with the status endpoint.

Status response fields

string

Unique task identifier.

string

Current task state: pending, processing, success, or failed.

string

JSON string containing generated video URLs in resultUrls. Present when state is success.

string

Machine-readable failure code. Present when state is failed.

string

Human-readable failure message. Present when state is failed.

integer

Task creation timestamp in Unix milliseconds.

integer

Task completion timestamp in Unix milliseconds. Present after completion.

integer

Processing duration in milliseconds. Present after completion when timing data is available.

object

Final billing snapshot under data.billing, returned on successful terminal status and callback responses for new tasks. Older tasks may omit this object; see Billing for compatibility rules.

string

Settlement status. The final snapshot uses final.

string

Opaque identifier of the billing rule captured when this task was created. Do not infer the rule from polling time.

number

Actual total reference-video duration in seconds. Zero without video references; this is not raised to the minimum billable input.

number

For new tasks with video references, ceil(outputVideoSeconds * 2 / 3), rounded upward to a whole second. Zero without video references or under the original rule.

number

The larger of inputVideoSeconds and minimumInputVideoSeconds; zero without video references.

number

Output duration used for billing, in seconds. For duration=-1, this is the actual generated duration, rounded half up to 3 decimal places. For a Final, it is the source Draft’s actual output duration.

number

Total billed seconds: outputVideoSeconds + billableInputVideoSeconds.

string

Duration unit. The value is second.

string

Additional field for settled duration=-1 tasks: round half up to 3 decimal places. May be absent for fixed-duration and Final tasks.

integer

Number of reference videos included in inputVideoSeconds. Returned for settled duration=-1 tasks; may be absent for fixed-duration and Final tasks.

integer

Number of output videos included in outputVideoSeconds. Returned for settled duration=-1 tasks; may be absent for fixed-duration and Final tasks.

Webhook callback mode

Use callback mode when your backend should receive the final result automatically instead of polling.

curl -X POST "https://api.apixo.ai/api/v1/generateTask/seedance-2-5" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "request_type": "callback",
    "callback_url": "https://your-server.com/webhooks/apixo",
    "input": {
      "mode": "first_and_last_frames",
      "prompt": "turn the still frames into a smooth cinematic camera move",
      "image_urls": [
        "https://example.com/frame-start.png",
        "https://example.com/frame-end.png"
      ],
      "resolution": "720p",
      "duration": 5,
      "aspect_ratio": "auto",
      "sound": true
    }
  }'

The callback payload uses the same top-level shape as status polling with code, message, and data. Successful callbacks include the final state and resultJson. For new tasks, including fixed-duration, duration=-1, Draft, and Final tasks, successful terminal callbacks include the same data.billing object returned by status polling. See Webhooks for delivery requirements and retry behavior.

Billing

Seedance 2.5 is billed per second. The selected resolution and whether omni_reference includes video_urls determine the per-second unit price.

For iterative production, generate 480p Drafts first and create 1080p only from the selected Draft. In the four-candidate example above, this reduces generation spend by 57.5% and can roughly double iteration efficiency by avoiding repeated full-resolution generations.

Configuration	APIXO price
`480p`, without video reference	`$0.140 / second`
`480p`, with video reference	`$0.085 / second`
`720p`, without video reference	`$0.315 / second`
`720p`, with video reference	`$0.190 / second`
`1080p`, without video reference	`$0.800 / second`
`1080p`, with video reference	`$0.500 / second`

Formula without video reference:

total cost = output duration * unit price

Formula with video reference:

billable seconds = output seconds + max(actual total reference video seconds, ceil(output seconds * 2 / 3))
total cost = billable seconds * unit price

text-to-video and first_and_last_frames always use the without-video-reference rate. omni_reference uses the with-video-reference rate only when video_urls is provided. Image and audio references do not add extra billable seconds. ceil rounds upward to the next whole second. The minimum applies once to the total reference-video duration, not separately to each video. For example, 5 output seconds with 3 actual reference seconds bills 5 + max(3, 4) = 9 seconds; 6 output seconds with 2 actual reference seconds bills 10 seconds. This rule applies to tasks created after the billing update. Existing tasks retain the rule captured at creation: with video references, (output duration + actual total reference video duration) * unit price; without video references, output duration only. Per-second prices, custom prices, and user discounts are unchanged. Polling or receiving a callback after the update does not change an existing task’s rule. For new duration=-1 tasks, APIXO pre-charges using an estimated 30-second output:

Without video references: 30 * unit price.
With video references: (30 + max(actual total reference video seconds, 20)) * unit price, because ceil(30 * 2 / 3) = 20.

After success, APIXO recalculates billable seconds using the actual output duration and the task’s captured rule, then refunds any excess using the effective unit price captured at creation, including applicable custom prices and discounts. The final charge is rounded to four decimal places before subtracting it from the original charge to calculate the refund. For example, with 3 reference-video seconds, pre-charge covers 50 seconds; if actual output is 5 seconds, final billing covers 9 seconds. Older automatic-duration tasks retain their original input-plus-output rule for both pre-charge and settlement. Draft creation is charged as a normal 480p generation using the same formulas and, for duration=-1, the same pre-charge and settlement rules. Each Final is a separate new task charged at the 1080p rate to the API key that submits it:

Without video references: use the source Draft’s actual output seconds.
With video references: Draft actual output seconds + max(Draft actual total reference video seconds, ceil(Draft actual output seconds * 2 / 3)).
For an automatic-duration Draft, use its final settled actual output duration, rounded half up to three decimal places. A Draft settled at 10.080 seconds with 3 actual reference-video seconds gives 10.080 + max(3, 7) = 17.080 billable seconds for a new Final.
Use the original actual reference duration, not the Draft’s minimum or billable input duration. Each reuse is billed independently. A Final created after the update uses the new rule even if its source Draft used the original rule; the Draft’s own charge is unchanged.

Successful terminal responses for new tasks include the final billing snapshot under data.billing; it is no longer limited to duration=-1. Fixed-duration, Draft, and Final tasks include the common fields. Automatic-duration tasks include additional rounding and video-count fields, as in this example:

{
  "billing": {
    "settlementState": "final",
    "inputVideoSeconds": 25.371,
    "minimumInputVideoSeconds": 17.000,
    "billableInputVideoSeconds": 25.371,
    "outputVideoSeconds": 25.042,
    "totalBillableSeconds": 50.413,
    "unit": "second",
    "rounding": "round half up to 3 decimal places",
    "inputVideoCount": 1,
    "outputVideoCount": 1
  }
}

Successful terminal status and callback responses for new tasks include data.billing. Duration values use seconds with three decimal places; minimumInputVideoSeconds is a whole-second ceiling. inputVideoSeconds remains the actual input duration, while billableInputVideoSeconds is the charged input duration. The snapshot remains stable on repeated polling. Older tasks may omit billing or the new fields; clients must not require them for historical tasks. totalBillableSeconds is outputVideoSeconds + billableInputVideoSeconds. Without video references, actual, minimum, and billable input seconds are all zero. Under the original rule, the minimum is zero and billable input equals actual input. Older automatic-duration tasks can still return a final snapshot under their original rule. For current route and market comparison pricing, see Pricing.

Latency and polling

Seedance 2.5 tasks are asynchronous. Actual latency may vary by prompt complexity, duration mode, resolution, reference media accessibility, route queue load, and storage transfer time.

Workload	Typical generation time	Recommended first poll	Poll interval
Text-to-video or first-and-last-frames	Varies by prompt, duration, and resolution	120s after task creation	10s-20s
Omni-reference with image/video/audio inputs	Varies by reference count, media duration, and accessibility	180s after task creation	10s-20s

For production workloads, use callback mode to avoid frequent polling while video tasks run.

Result URLs are temporary. Download and store important outputs promptly after task completion. Rate limits and concurrency can vary by account, API key, and route. If you receive 429, slow down requests and retry with backoff. For account-level details, see System APIs.

Errors and troubleshooting

HTTP errors

Code	Meaning	What to do
`400`	Invalid request body, missing `input`, unsupported `mode`, missing `prompt`, invalid `resolution`, invalid `duration`, invalid `aspect_ratio`, invalid boolean value, invalid `output_format`, or invalid media array	Fix the request before retrying
`401`	Missing or invalid API key	Check the `Authorization` header
`402`	Insufficient balance or quota	Add balance or switch account/key
`403`	Key or route cannot access the model	Check permissions and route strategy
`404`	Task not found when polling	Check the `taskId`
`429`	Rate limit or concurrency limit reached	Retry with exponential backoff
`500`	Server error or unknown task failure	Retry with backoff
`502`	Generation service error	Retry with backoff
`504`	Generation timeout	Retry or use callback mode for long-running jobs

Validation notes

Condition	Backend behavior
Missing `input`	Request fails before task creation.
Missing, non-string, or empty `prompt`	Request fails before task creation.
Missing `mode`	Request fails before task creation.
Unsupported `mode`	Request fails before task creation. Supported values are `text-to-video`, `first_and_last_frames`, and `omni_reference`.
Missing `resolution`	Defaults to `720p`.
Unsupported `resolution`	Request fails before task creation. Supported values are `480p`, `720p`, and `1080p`.
Missing `duration`	Defaults to `5`.
`duration` outside `4` through `30` and not `-1`	Request fails before task creation.
`duration=-1`	New tasks pre-charge 30 seconds without video references, or `30 + max(actual total reference video seconds, 20)` with them; after success, recalculate using actual output and refund any excess. Existing tasks keep their original rule.
`draft=true` with omitted `resolution`	The request generates a 480p Draft.
`draft=true` with a resolution other than `480p`	Request fails before task creation.
`draft=true` together with `draft_task_id`	Request fails before task creation.
`draft_task_id` with omitted `resolution`	The request generates a 1080p Final.
`draft_task_id` with a resolution other than `1080p`	Request fails before task creation.
`draft_task_id` with inherited generation fields	Request fails before task creation. Only `resolution`, `output_format`, and `watermark` may accompany `draft_task_id` inside `input`.
`draft_task_id` references an unsuccessful, expired, or incompatible task	Request fails before charging the Final.
Missing `aspect_ratio`	Defaults to `auto`.
Unsupported `aspect_ratio`	Request fails before task creation.
Video-editing prompt intent with `duration` not set to `-1` or `aspect_ratio` not set to `auto`	The generation task can fail; use `duration=-1` and `aspect_ratio=auto` for these prompts.
`sound`, `watermark`, or `nsfw_checker` is not a boolean	Request fails before task creation.
Unsupported `output_format`	Request fails before task creation. Supported values are `mp4` and `mov`.
`first_and_last_frames` without `image_urls`	Request fails before task creation.
`first_and_last_frames` with more than 2 images	Request fails before task creation.
`first_and_last_frames` with `aspect_ratio` other than `auto`	Request fails before task creation.
`omni_reference` without `image_urls`, `video_urls`, or `audio_urls`	Request fails before task creation.
`omni_reference` with more than 30 images, 10 videos, or 10 audio files	Request fails before task creation.
Reference video or audio shorter than 2 seconds or longer than 30 seconds	Request fails before task creation.
Total reference video duration or total reference audio duration over 30 seconds	Request fails before task creation.

Task failure codes

failCode identifies the task failure. Common values include:

Fail code	Meaning	What to do
`SensitiveContent`	Prompt or input/output content was rejected by safety checks	Change the prompt or reference media
`PromptInvalid`	Prompt was invalid or rejected during generation	Rewrite the prompt and retry
`ImageFormatIncorrect`	Reference image format could not be processed	Use a public, direct image URL in a common image format
`RateLimited`	Rate limit was reached	Retry with exponential backoff
`Timeout`	Generation timeout	Retry later or use callback mode
`StreamError`	A generation service error occurred	Retry with backoff
`Unknown error`	An unknown task failure occurred	Retry with backoff or contact support with the `taskId`

See Error Codes for the full error reference.

​Overview

​Endpoint and authentication

​Recommended: Draft first, Final second

​Step 1: Generate a 480p Draft

​Step 2: Generate the 1080p Final

​Why this workflow saves about 57%

​Copy-paste async quickstart

​Poll for result

​Request body

​Text-to-video

​Dynamic duration

​First-and-last-frames

​Omni-reference

​Parameters

​Draft and Final input rules

​Reference generation vs. video editing

​Response format

​Submit task response

​Status response fields

​Webhook callback mode

​Billing

​Latency and polling

​Errors and troubleshooting

​HTTP errors

​Validation notes

​Task failure codes

​Related links

Overview

Endpoint and authentication

Recommended: Draft first, Final second

Step 1: Generate a 480p Draft

Step 2: Generate the 1080p Final

Why this workflow saves about 57%

Copy-paste async quickstart

Poll for result

Request body

Text-to-video

Dynamic duration

First-and-last-frames

Omni-reference

Parameters

Draft and Final input rules

Reference generation vs. video editing

Response format

Submit task response

Status response fields

Webhook callback mode

Billing

Latency and polling

Errors and troubleshooting

HTTP errors

Validation notes

Task failure codes

Related links