Use the Veo 3.1 Fast API for 8-second horizontal or vertical video

xplaai exposes the Veo 3.1 Fast API through POST /v1/videos using the public model name veo-3.1-fast. The contract creates fixed eight-second video tasks in 720p or 1080p, supports 16:9 and 9:16, and accepts frame or ingredient reference images. Frame mode accepts up to two references, ingredient mode up to three, while reference-video input is unsupported.

Create an API key or review the current video documentation.

Veo 3.1 Fast API contract at a glance

The model name is a contract boundary. Send only fields documented for veo-3.1-fast; controls accepted by another video model are not automatically valid here.

Create a 1080p reference-image task

export XPLA_API_KEY="your_api_key"

curl --request POST \
  --url https://xplaai.com/v1/videos \
  --header "Authorization: Bearer $XPLA_API_KEY" \
  --header "Content-Type: application/json" \
  --data '{
    "model": "veo-3.1-fast",
    "prompt": "animate this product reference image into a clean cinematic shot with subtle camera movement and realistic lighting",
    "seconds": "8",
    "images": [
      "https://example.com/reference.png"
    ],
    "metadata": {
      "durationSeconds": 8,
      "resolution": "1080p",
      "aspectRatio": "16:9",
      "mode_image": "frame"
    }
  }'

The duration aliases in this example agree. If you send more than one duration field, every value must resolve to eight seconds. A conflicting or unsupported value should be treated as an invalid request.

Choose frame mode or ingredient mode

Reference images play different roles depending on metadata.mode_image.

Frame mode

Use frame when references define the visual sequence:

• one image can act as the starting frame;
• two ordered images can define first and last frames;
• all reference-image entry points count toward the two-image limit.

For first-and-last-frame generation, preserve the intended order in the images array.

Ingredient mode

Use ingredient when references describe subjects or visual materials that should influence the generated scene:

• pass up to three reference images;
• do not assume their ordering defines a first and last frame;
• keep the prompt explicit about the role of each supplied subject or asset.

Reference inputs can use supported public URLs or base64 values. A multipart request can upload reference-image files. Do not mix remote reference strings and local uploads of the same media type in one request.

Generate a 9:16 vertical video

Veo 3.1 Fast supports 9:16 directly:

{
  "model": "veo-3.1-fast",
  "prompt": "create a natural vertical product reveal with a slow upward camera move",
  "seconds": "8",
  "resolution": "1080p",
  "ratio": "9:16"
}

You can express the aspect ratio through a supported top-level alias or the documented metadata field. If multiple aliases are present, keep them consistent.

Poll the asynchronous task

The initial generation call creates a video task. Store its returned ID and use the standard xplaai video lifecycle:

GET https://xplaai.com/v1/videos/{task_id}
Authorization: Bearer $XPLA_API_KEY

When the task is complete:

GET https://xplaai.com/v1/videos/{task_id}/content
Authorization: Bearer $XPLA_API_KEY

Persist the task ID so a page refresh or worker restart does not lose the render. Your interface should distinguish pending, completed and failed states.

Validate requests before submission

Reject invalid combinations in your application before calling the API:

• a duration other than eight seconds;
• a resolution other than 720p or 1080p;
• an aspect ratio other than 16:9 or 9:16;
• more than two frame references;
• more than three ingredient references;
• metadata.ref_videos;
• top-level mode or input_reference;
• controls such as negative_prompt, camera_control, style or seed that belong to a different contract.

This validation gives users immediate feedback and prevents a generic UI from passing unrelated model settings into the request.

When to use Veo 3.1 Fast

The fixed eight-second contract is a useful fit when your product needs a predictable clip unit:

• a short horizontal product reveal;
• a vertical mobile creative;
• animation between supplied first and last frames;
• a reference-guided subject or product shot;
• a repeatable component inside a longer editing workflow.

For flexible 3–15 second text-to-video output, compare the Kling Video 3.0 API. For 6, 10 or 15 second 720p output with a broader set of aspect ratios, review the Grok Imagine Video API.

If the workflow also needs product research, copy, narration or editing, an API call is only one layer. Explore the Batch Commerce Video Skill for a packaged workflow.

Frequently asked questions

How long is a Veo 3.1 Fast video?

The public veo-3.1-fast contract is fixed at eight seconds. Other duration values are not part of this model's xplaai contract.

Does Veo 3.1 Fast support 1080p?

Yes. The supported resolution values are 720p and 1080p. The default is 720p, so send 1080p explicitly when required.

Can I create a 9:16 vertical video?

Yes. Use the supported 9:16 aspect ratio through a documented top-level or metadata field.

What is the difference between frame and ingredient mode?

Frame mode accepts up to two ordered images and is suited to start or first-and-last-frame control. Ingredient mode accepts up to three references that guide subjects or visual materials.

Can I pass a reference video?

No. Reference-video input is not supported by veo-3.1-fast. Use reference images or choose a separately documented editing model.

How do I retrieve the completed video?

Poll GET /v1/videos/{task_id}. After completion, request GET /v1/videos/{task_id}/content with Bearer authentication.

Run a Veo 3.1 Fast test

Create an xplaai API key, submit a small eight-second request and verify the complete task lifecycle before integrating it into production. Return to the AI video API overview to compare other video contracts.