Video generation.

Key request parameters

Key response parameters

Integration examples

Replace the API key and model ID in these examples with your values.

curl https://hashbidx.ai/v1/videos \
  -H "Authorization: Bearer <YOUR_API_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "example-video-5s",
    "prompt": "a cat dancing on a piano, slow dolly in",
    "input_reference": "https://example.com/first-frame.png",
    "metadata": {"ratio": "16:9"}
  }'

import httpx

resp = httpx.post(
    "https://hashbidx.ai/v1/videos",
    headers={"Authorization": "Bearer <YOUR_API_KEY>"},
    json={
        "model": "example-video-5s",
        "prompt": "a cat dancing on a piano, slow dolly in",
        "input_reference": "https://example.com/first-frame.png",
        "metadata": {"ratio": "16:9"},
    },
)
resp.raise_for_status()
task = resp.json()
print("task id:", task["id"], "status:", task["status"])

const resp = await fetch("https://hashbidx.ai/v1/videos", {
  method: "POST",
  headers: {
    "Authorization": "Bearer <YOUR_API_KEY>",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    model: "example-video-5s",
    prompt: "a cat dancing on a piano, slow dolly in",
    input_reference: "https://example.com/first-frame.png",
    metadata: { ratio: "16:9" },
  }),
});
const task = await resp.json();
console.log("task id:", task.id, "status:", task.status);

body, _ := json.Marshal(map[string]any{
  "model":           "example-video-5s",
  "prompt":          "a cat dancing on a piano, slow dolly in",
  "input_reference": "https://example.com/first-frame.png",
  "metadata":        map[string]any{"ratio": "16:9"},
})
req, _ := http.NewRequest("POST",
  "https://hashbidx.ai/v1/videos", bytes.NewReader(body))
req.Header.Set("Authorization", "Bearer <YOUR_API_KEY>")
req.Header.Set("Content-Type", "application/json")
resp, err := http.DefaultClient.Do(req)

let body = serde_json::json!({
    "model": "example-video-5s",
    "prompt": "a cat dancing on a piano, slow dolly in",
    "input_reference": "https://example.com/first-frame.png",
    "metadata": { "ratio": "16:9" },
});
let resp = reqwest::Client::new()
    .post("https://hashbidx.ai/v1/videos")
    .bearer_auth("<YOUR_API_KEY>")
    .json(&body)
    .send().await?;

Video generation (/v1/videos)

• Create: POST /v1/videos with model and prompt in the body; optionally input_reference as the first frame (an image URL or base64) and metadata.ratio for the aspect ratio (e.g. 16:9).
• Retrieve: GET /v1/videos/{id} returns the task status (queued, in_progress, completed, failed); once completed, download the video from GET /v1/videos/{id}/content.
• Billing: most video models charge per second: seconds × the per-second price of the chosen resolution and audio; some charge by the video tokens the upstream reports: tokens × the price per million tokens ÷ 1,000,000, which may differ when a reference video is given; a model that makes one video only charges one per-request price. Some models also charge per input image (first and last frames and reference images all count), for example after a few free ones. Failed generations are not charged. See the Models page for each model's prices.
• Duration: seconds (a string such as "8") or metadata.duration (an integer); if you send both they must agree. Omitted, the model's default duration is used. A duration the chosen resolution doesn't offer returns 400.
• Resolution: metadata.resolution, e.g. 720p or 1080P (case-insensitive); omitted, the model's default is used. Don't send size; any size returns 400. Use metadata.resolution with metadata.ratio instead.
• Audio: metadata.generate_audio (true or false); omitted, the model's default is used. Each model lists its resolutions and, for each, the durations and audio it offers on the Models page; a value outside them returns 400 unsupported_video_spec before anything is charged, and the message lists what the model offers.
• metadata with frames, draft: true, tools or callback_url, or a service_tier other than default, returns 400.
• metadata.content takes only text and image_url items; video, audio or any other kind returns 400.
• input_reference accepts three forms only; anything else returns 400: an http(s) image address whose last path segment ends in png, jpg, jpeg, webp, gif or bmp, or has no extension; a data URL starting with data:image/; or a PNG, JPEG, GIF, WEBP or BMP image in base64.
• Reference images go in metadata.content: items without the first_frame or last_frame role, of type image_url. They can't be mixed with first or last frames.
• input_reference is sent as the first frame: giving a first frame in metadata.content as well, or sending reference images with it, returns 400.

Example request body (the model name and image URL are placeholders):

{
  "model": "example-video-5s",
  "prompt": "a cat dancing on a piano, slow dolly in",
  "input_reference": "https://example.com/first-frame.png",
  "metadata": { "ratio": "16:9" }
}