> ## Documentation Index
> Fetch the complete documentation index at: https://docs.qingbo.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Gemini Omni Flash

> The model-chosen-length Gemini Omni video models: text, reference images or a reference video, with multi-turn editing

This page covers the two Omni models whose length is decided by the model: `gemini-omni-1.1-flash` and `gemini-omni-flash-preview`. They share one request shape and one billing model (resolution tier × seconds); they differ in resolution tiers, reference-material limits, and which image parameters they accept — switching between them means changing `model` and nothing else. When the clip has to land on exactly 4, 6, 8 or 10 seconds, use [Omni-Flash-Ext](/en/api-reference/video/omni/omni-flash-ext).

## Quick start

<CodeGroup>
  ```bash Text to video theme={"system"}
  curl -X POST https://www.qingbo.dev/v1/tasks \
    -H "Authorization: Bearer $WAVE_API_KEY" \
    -H "Content-Type: application/json" \
    -H "Idempotency-Key: omni-flash-001" \
    -d '{
      "model": "gemini-omni-flash-preview",
      "action": "generate",
      "prompt": "a blue butterfly landing on a flower, macro, soft light",
      "resolution": "720p",
      "aspect_ratio": "9:16"
    }'
  ```

  ```bash Reference images theme={"system"}
  curl -X POST https://www.qingbo.dev/v1/tasks \
    -H "Authorization: Bearer $WAVE_API_KEY" \
    -H "Content-Type: application/json" \
    -d '{
      "model": "gemini-omni-1.1-flash",
      "action": "generate",
      "prompt": "the cat playfully bats at the ball of yarn",
      "image_urls": [
        "https://cdn.example.com/cat.png",
        "https://cdn.example.com/yarn.png"
      ],
      "resolution": "1080p",
      "aspect_ratio": "16:9"
    }'
  ```

  ```bash First and last frame theme={"system"}
  curl -X POST https://www.qingbo.dev/v1/tasks \
    -H "Authorization: Bearer $WAVE_API_KEY" \
    -H "Content-Type: application/json" \
    -d '{
      "model": "gemini-omni-1.1-flash",
      "action": "generate",
      "prompt": "smooth transition, camera slowly pushes in",
      "first_frame_image": "https://cdn.example.com/start.jpg",
      "last_frame_image": "https://cdn.example.com/end.jpg",
      "resolution": "720p",
      "aspect_ratio": "16:9"
    }'
  ```

  ```bash Rework a reference video theme={"system"}
  curl -X POST https://www.qingbo.dev/v1/tasks \
    -H "Authorization: Bearer $WAVE_API_KEY" \
    -H "Content-Type: application/json" \
    -d '{
      "model": "gemini-omni-flash-preview",
      "action": "generate",
      "prompt": "when the person touches the mirror, make it ripple like liquid",
      "video_urls": ["https://cdn.example.com/clip.mp4"]
    }'
  ```

  ```bash Continue from the previous result theme={"system"}
  curl -X POST https://www.qingbo.dev/v1/tasks \
    -H "Authorization: Bearer $WAVE_API_KEY" \
    -H "Content-Type: application/json" \
    -d '{
      "model": "gemini-omni-flash-preview",
      "action": "generate",
      "prompt": "keep going: the camera pans across the mountains in the distance",
      "ref_task_id": "task-wave1788783109b950713584"
    }'
  ```
</CodeGroup>

A successful submission returns a `task_id`. Retrieve the result with [`GET /v1/tasks/{task_id}`](/en/api-reference/task/status), wait inline with [`Prefer: wait`](/en/api-reference/task/submit#in-request-waiting-prefer-wait), or configure a webhook.

**Working in rounds**: submit one generation, take its `task_id`, and pass it as `ref_task_id` on the next request to extend or edit that result without re-uploading anything. `ref_task_id` and `video_urls` are mutually exclusive — send one or the other.

## Request parameters

<ParamField body="model" type="string" required>
  Either `gemini-omni-1.1-flash` or `gemini-omni-flash-preview`.
</ParamField>

<ParamField body="action" type="string" default="generate">
  Value `generate`; it may be omitted. The mode is not selected with `action` but by what you send: `prompt` alone is text-to-video, images make it image-to-video, and `video_urls` or `ref_task_id` reworks an existing video.
</ParamField>

<ParamField body="prompt" type="string" required>
  A scene description for text-to-video; an action, style or editing instruction for image-to-video and video rework. Keep edit instructions short, and add "keep everything else the same" when changing only one thing.
</ParamField>

<ParamField body="resolution" type="string" default="720p">
  Output resolution, case-insensitive.

  `gemini-omni-1.1-flash`: `360p`, `720p`, `1080p`, `4k`.

  `gemini-omni-flash-preview`: `720p` only.
</ParamField>

<ParamField body="aspect_ratio" type="string" default="16:9">
  Frame ratio: `16:9` landscape or `9:16` portrait. With `video_urls`, the output follows the reference video's ratio.
</ParamField>

<ParamField body="image_urls" type="string[]">
  Reference image URLs, publicly accessible. One image is used as the starting frame by default; several are used as multi-subject or style references — describe each image's role and how they interact in `prompt`.

  Limits: 10 images in total for `gemini-omni-1.1-flash` (first and last frames included), 16 for `gemini-omni-flash-preview`.
</ParamField>

<ParamField body="first_frame_image" type="string">
  First-frame image URL. On its own it becomes the opening frame; together with `last_frame_image` it produces a transition between the two.

  `gemini-omni-1.1-flash` only.
</ParamField>

<ParamField body="last_frame_image" type="string">
  Last-frame image URL, which must be sent together with `first_frame_image`. Use the same frame ratio for both images and match it to `aspect_ratio`.

  `gemini-omni-1.1-flash` only.
</ParamField>

<ParamField body="video_urls" type="string[]">
  URLs of the video to edit or extend, as publicly downloadable direct links. At most one clip.

  Length limit: 10 seconds for `gemini-omni-1.1-flash`, 24 seconds for `gemini-omni-flash-preview`.
</ParamField>

<ParamField body="ref_task_id" type="string">
  Our task id to extend or conversationally edit (translated to the upstream task id); mutually exclusive with `video_urls`.
</ParamField>

<ParamField body="image_with_roles" type="array">
  Role-tagged images `{url, role}` with role `first_frame` / `last_frame` / `reference`; use either this or `first_frame_image` / `last_frame_image`.

  `gemini-omni-1.1-flash` only.
</ParamField>

<ParamField body="metadata" type="object">
  Generation-intent object; currently only `task`: `text_to_video` | `image_to_video` | `reference_to_video` | `edit` | `extend` (inferred when omitted).

  `gemini-omni-1.1-flash` only.
</ParamField>

See [Submit Task](/en/api-reference/task/submit) for `callback_url`, `callback_events`, `Prefer: wait`, `Idempotency-Key`, and the maximum-cost header.

Neither model takes `duration`: the model decides the length from the content, typically 3–10 seconds, and pacing is steered with natural language in `prompt`. `seed` is not supported either. Any other parameter returns `400` and is not billed.

## Limits

* One video per task; at most one reference video.
* A reference video and `ref_task_id` are mutually exclusive.

| Item             | `gemini-omni-1.1-flash`          | `gemini-omni-flash-preview` |
| ---------------- | -------------------------------- | --------------------------- |
| Resolution       | `360p` / `720p` / `1080p` / `4k` | `720p` only                 |
| Reference images | 10 in total (frames included)    | 16                          |
| Reference video  | At most 1, up to 10 seconds      | At most 1, up to 24 seconds |
| Duration         | Model-chosen, 3–10 seconds       | Model-chosen, 3–10 seconds  |
| Aspect ratio     | `16:9` / `9:16`                  | `16:9` / `9:16`             |

## Pricing

Billing runs on **resolution tier × output seconds**: each resolution tier has its own per-second rate, and the seconds are those of the clip the model actually produced. Requests carrying a reference video or `ref_task_id` are billed the same way, on output seconds; reference videos and reference images cost nothing extra.

Because the length is the model's choice and is unknown at submission, the hold is placed at the model's maximum length (10 seconds); the task then settles on the actual generated result and the difference is released.

You are only billed for a successfully generated video. Failed and cancelled tasks, and tasks that return no usable video, are refunded in full. What a call actually cost is the `cost` field on the task response — an integer quota at **500,000 quota = 1 USD**, not dollars.

<Note>
  Per-model rates are in `price_config` on `GET /v1/models` and in the console's Model Market.

  This section describes the billing dimensions only; it carries no amounts.
</Note>

## Response

```json Completed task theme={"system"}
{
  "task_id": "taski-9e2b0d22d4b9faeabaad8215fe1585fc-task-wave1788783072b950204622",
  "model": "gemini-omni-1.1-flash",
  "action": "generate",
  "status": "completed",
  "progress": "100%",
  "created_at": 1788783072,
  "completed_at": 1788783101,
  "result": {
    "videos": [
      {
        "expires_at": 1789089508,
        "url": ["https://cdn.example.com/result.mp4"]
      }
    ]
  },
  "billing_status": "settled",
  "cost": 152081,
  "urls": {
    "get": "https://www.qingbo.dev/v1/tasks/{task_id}",
    "cancel": "https://www.qingbo.dev/v1/tasks/{task_id}/cancel"
  }
}
```

The generated video is returned in `result.videos`, where `url` is an array and `expires_at` is the link's expiry as a Unix timestamp in seconds. See [Query Task Status](/en/api-reference/task/status) for every field.

## Available models

| Model ID                    | Resolution                       | Reference images | Reference video | Model-only parameters                                                   |
| --------------------------- | -------------------------------- | ---------------- | --------------- | ----------------------------------------------------------------------- |
| `gemini-omni-1.1-flash`     | `360p` / `720p` / `1080p` / `4k` | 10               | ≤ 10 s          | `first_frame_image`, `last_frame_image`, `image_with_roles`, `metadata` |
| `gemini-omni-flash-preview` | `720p` only                      | 16               | ≤ 24 s          | —                                                                       |

Actions, limits and billing are identical on both models, and both accept `ref_task_id`.

## Related

* [Omni Video Series](/en/api-reference/video/omni/overview)
* [Omni-Flash-Ext](/en/api-reference/video/omni/omni-flash-ext)
* [Video Generation Overview](/en/api-reference/video/overview)
* [Submit Task](/en/api-reference/task/submit)
* [Query Task Status](/en/api-reference/task/status)
* [Task System](/en/docs/task-system)