One API · more models on the way

One API forleading AI generation models

A single task-based API: submit a task, poll its status or receive a callback. Outputs are kept for 30 days, failed tasks are refunded automatically, and new models plug into the same interface.

  • 1 key

    for every model we host

  • 100%

    of credits returned on failure

  • 2 min

    between callback recovery sweeps

  • 30 days

    result file retention

Why it holds up

An AI API gateway you can put in production

Access, billing and delivery are handled at the gateway, so your code only deals with the output.

01

One interface

Every model shares the same task API. Switching models means changing the model field.

  • Same endpoint
  • Same request shape
  • Same response envelope
  • Same callbacks

02

Billing you can audit

Credits are held on submit and settled when the task ends. Every movement is in your ledger.

  1. Held on submit
  2. Refunded on failure
  3. Priced by usage
  4. Full transaction log

03

The whole task lifecycle

From submission to delivery, the gateway owns the parts that usually break.

  1. 01Submit
  2. 02Queue
  3. 03Generate
  4. 04Store
  5. 05Callback
  6. 06Query

Models

Available models

1 credit = $0.005. Public per-model pricing, charged by usage.

ByteDanceVideo

Seedance 2.0

seedance-2.0

ByteDance video generation. Text-to-video and first/last-frame image-to-video, 4–15 seconds.

  • Text-to-video
  • Image-to-video
  • Audio

720p text-to-video

33credits/sec

≈ $0.825 for a 5-second clip

ByteDanceVideo

Seedance 2.0 Fast

seedance-2.0-fast

The faster variant with the same request shape.

  • Text-to-video
  • Image-to-video
  • Audio

720p text-to-video

33credits/sec

≈ $0.825 for a 5-second clip

More models on the way

Image, video and other generation models are being added. They use the same API, so your integration does not change.

Video models are open at 720p today. Failed tasks are refunded in full.

Quickstart

From sign-in to first result in three steps

Every model shares one API. Switching models means changing one field.

  1. 1

    Sign in and create a key

    Open the console, create an API key and top up credits.

    Open the console
  2. 2

    Pick a model

    Put the model ID in the model field. The rest of the request stays the same.

    • seedance-2.0
    • seedance-2.0-fast
    See available models
  3. 3

    Submit a task

    Create a task, then poll for its status or wait for the callback.

    Create task reference
Create a task
curl -X POST https://api.xbaiapi.com/api/v1/jobs/createTask \
  -H "Authorization: Bearer xb-your-key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seedance-2.0",
    "input": {
      "prompt": "a cat stretching on a sunlit windowsill",
      "duration": 5
    }
  }'

Details

What you get beyond model access

The parts that cost you time when a model provider misbehaves.

  • 01

    One integration

    Tasks, queries and callbacks work the same for every model. New models are usable as soon as they launch.

  • 02

    Transparent pricing

    Per-model prices are public, and you are charged for what you use. 1 credit = $0.005.

  • 03

    Outputs kept for 30 days

    Every result is copied to our own object storage and stays available for 30 days, well after temporary upstream links expire.

  • 04

    Automatic refunds

    Credits are held on submit and returned in full the moment a task fails.

  • 05

    Callbacks that retry

    If your endpoint is down we retry, and a sweep every two minutes recovers dropped upstream callbacks.

  • 06

    One console

    Keys, tasks, usage and billing in one place, with a per-task record of what was charged.

FAQ

Frequently asked questions

Which models are available?

See Available models above. More models are being added, and each one uses the same API as the rest.

How is usage billed?

Each model has a public unit price, and you pay for what you use. 1 credit = $0.005. The amount is held when you submit and settled when the task ends.

What resolutions can I use for video?

720p today. Higher tiers open once their upstream cost is confirmed — we would rather not sell a tier we cannot cost.

What happens if a generation fails?

The full credit amount is returned automatically. You can see the refund in your transaction history.

How long do result files stay available?

30 days from task creation, then the file is deleted. Copy anything you need to keep to your own storage before then.

Do I have to handle callbacks?

No. Callbacks are optional; without one you can poll the query endpoint with the task ID. If you do set one, failed deliveries are retried.

Get started

One API, every model we add

Sign in, create a key, submit your first task.