Video generation

POST /v1/videos: text to video, image to video and first/last frame mode — with async mode for long renders.

Updated on Sep 02, 2026

Generates video from text, from an image, or interpolating between a first and last frame — the mode that transforms two images into a narrative scene. There are 24 models in the catalog (GET /v1/models, endpoint: "/v1/videos"), charged per second of video or per execution.

Rendering takes 1 to 4 minutes. To avoid holding the connection, use async mode: "async": true responds immediately with a job_id.

POSThttps://api.hinow.ai/v1/videosBearer

Renders a video and returns the public URL of the mp4.

Parâmetros

modelstring· bodyobrigatório
Namespaced model ID (e.g., `bytedance/seedance-2.0-mini`, `google/veo-31`).
promptstring· bodyobrigatório
The scene and movement, in text. With frames, describe the MOVEMENT between them.
imagesstring[]· body
Public URLs. One image = animate from it; **two = first and last frame** (in this order), in `image_to_video` models.
image_modestring· body
`frames` (default up to 2 images) or `reference` (3+ images become style/subject reference, in models that support it).
durationnumber· body
Seconds of video, within the model's limit.
asyncboolean· body
`true` returns `202` with `job_id` immediately — always recommended. See [Media Jobs](/en/models/quick-reference/media-jobs).

Respostas

200Rendered video (sync mode)

{
  "success": true,
  "data": {
    "urls": ["https://s3.us1-stlouis.hinow.ai/hi-storage/image_to_video/.../clip.mp4"],
    "model": "bytedance/seedance-2.0-mini",
    "category": "image_to_video",
    "cost": { "amount": 0.575, "currency": "USD", "value_type": "per_second" }
  }
}

202Async mode: `data.job_id` and `data.poll_url` — rendering continues on the server

400Incompatible model with the endpoint (`INVALID_MODEL_CATEGORY`) or invalid body

502/504Provider failed or timed out in sync mode — use `async: true`

First and last frame

The mode that renders the best short scenes: generate (or upload) the image of how the scene begins and the image of how it ends, and the model creates the movement between them.

cena com narrativa (medido: ~1m45s, US$ 0,575 por 5s no seedance-2.0-mini)bash

curl -X POST https://api.hinow.ai/v1/videos \
  -H "Authorization: Bearer hi_SUA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "bytedance/seedance-2.0-mini",
    "prompt": "the puppy slowly wakes up and looks at the camera, gentle motion, static camera",
    "images": ["https://.../quadro-inicial.jpg", "https://.../quadro-final.jpg"],
    "duration": 5,
    "async": true
  }'

The two frames should look like the same place

Generate both with the same scenery, lighting and framing — only what the narrative changes should change. The ready-made flow Scene Director does exactly that from a natural language request.

Price per second varies by model

Models per_second do not expose the unit value in the catalog — the actual cost comes in the cost field of the response (measured: 5s on seedance-2.0-mini = US$ 0.575). The lightricks/ltx-23-* models have fixed price per execution (US$ 0.19–0.25).

Was this page helpful?