# AiuniVid

> AiuniVid is an AI video generation platform with official, direct access to the Dreamina Seedance model family. Users describe a scene in text, or supply reference images and video, and receive a finished clip. Billing is pay-as-you-go credits with no subscription, and the exact credit cost is shown before a generation starts.

## What it does

- Text-to-video, image-to-video, and reference-to-video generation in a browser, plus a public HTTP API.
- Models: Seedance 2.5 (480p/720p/1080p, up to 30 seconds, omni reference and native audio), Seedance 2.0 (480p to 4K, up to 15 seconds), Seedance 2.0 Fast, and Seedance 2.0 Mini.
- Interface and documentation in Simplified Chinese and English.

## Upstream models

Requests run directly on BytePlus ModelArk, not through a reseller or a proxy of the
Dreamina web app. The mapping is published so it can be verified:
`seedance-2.5-*` runs on `dreamina-seedance-2-5-260628`,
`seedance-2.0-*` on `dreamina-seedance-2-0-260128`,
`seedance-2.0-fast-*` on `dreamina-seedance-2-0-fast-260128`,
and `seedance-2.0-mini-*` on `dreamina-seedance-2-0-mini-260615`.

## Pricing

- Credits are bought in packs and never expire: 1000 credits for USD 10, 2500 for USD 25, 5000 for USD 50. There is no subscription tier.
- Cost scales with model, resolution, and duration, and the exact figure is quoted before the job is submitted.
- Credits frozen for a job are released automatically when a generation fails.
- Generated videos may be used commercially.

## Key pages

- [Home](/): product overview and the generation console.
- [Models](/models): the model line-up and what each one is for.
- [Pricing](/pricing): credit packs and per-generation cost.
- [Docs](/docs): API documentation.

## Developer API

- [Machine-readable API reference](/api-reference.md): workflow, required headers, all request fields, public model IDs, task states, and errors.
- [OpenAPI 3.1 contract](/openapi.yaml): canonical endpoint and schema contract.
- [English human documentation](/en/docs) and [Chinese human documentation](/zh/docs).

## Rules for AI integrators

- Use the OpenAPI contract for endpoint and JSON-schema details.
- Use the Markdown reference for the complete integration workflow and constraints.
- Create requests are asynchronous and return a task ID; poll the task endpoint until it reaches a terminal state, or use `callback_url` for a webhook.
- API keys must be used only on a trusted server, and must carry the correct scope (`video:create`, `video:read`, or `files:write`) for each endpoint.
- Respect rate limits: 10 req/min/key for create endpoints, 60 req/min/key for read endpoints. Check `X-RateLimit-Remaining` before retrying.
- `GET /v1/videos/generations` lists tasks with cursor pagination; `GET /v1/credits/balance` returns `available_credits` and `frozen_credits`.
- Verify webhooks with HMAC-SHA256 over `{t}.{raw_body}`. Reject timestamps older than 5 minutes.