# Grok, Omni and Veo video

https://video-doc.1route.dev/en/docs/standard-video



These five models use the independent **视频** group, separate from Midjourney and 慢速. Use your site key. Prices are in CNY and billed per second.

| Model ID                 | Retail price | Default seconds |
| ------------------------ | -----------: | --------------: |
| `grok-imagine-video`     | ¥0.07/second |               8 |
| `grok-imagine-video-1.5` | ¥0.07/second |               8 |
| `gemini-omni-1.1-flash`  | ¥0.07/second |               8 |
| `gemini-veo-3.1-fast`    | ¥0.05/second |               8 |
| `gemini-veo-3.1-lite`    | ¥0.05/second |               8 |

## Create with a first-frame image [#create-with-a-first-frame-image]

```bash
curl https://video.1route.dev/v1/videos \
  -H "Authorization: Bearer <YOUR_API_KEY>" \
  -H "Content-Type: application/json" \
  -d '{"model":"grok-imagine-video-1.5","prompt":"The red kite sways gently in the wind","seconds":1,"resolution":"480p","images":["https://video-doc.1route.dev/video/kite-poster.jpg"],"generate_audio":false}'
```

Omit images for text-only requests. Grok accepts one image without a prompt; Gemini models require prompt. seconds accepts an integer or integer string; duration is an alias and must match if both are supplied. Choose one of images, reference_images, input_reference or image, with at most one first frame. HTTP(S) URLs and PNG/JPEG/WebP Base64 data URLs are accepted.

This group also accepts multipart: upload one PNG/JPEG/WebP as input_reference, up to 10 MiB. Form generate_audio values are true or false. This does not add multipart support to older model families.

## Enabled capabilities [#enabled-capabilities]

* Grok: 1–15 whole seconds, 480p/720p; 1.5 additionally accepts 1080p. generate_audio is a boolean.
* Omni / Veo: 8-second default. The 1–30 second range is a gateway validation limit; other durations and resolutions remain unverified.
* Text and one first-frame image are enabled. Reference audio, video, multiple images, tail frames and unknown parameters return 400 to prevent silently ignored paid inputs.

## Status, downloads and billing [#status-downloads-and-billing]

```bash
curl "https://video.1route.dev/v1/videos/<TASK_ID>" \
  -H "Authorization: Bearer <YOUR_API_KEY>"
curl "https://video.1route.dev/v1/tasks/<TASK_ID>/artifacts" \
  -H "Authorization: Bearer <YOUR_API_KEY>"
```

After completed, use the returned site url / metadata.url or artifact content_url. Alternatively call `/v1/videos/<TASK_ID>/content` with your site key; Range is supported. Treat access-bearing download links as credentials.

The requested duration is reserved at submission. Success settles using valid usage or a uniquely matched billing record. Missing reliable usage retains the reservation for review. Failure follows the refund path. Do not automatically repeat a generation POST after waiting, timeout or an ambiguous outcome.

## Verification and deferred model [#verification-and-deferred-model]

Customer-side checks cover single-image first-frame generation, status, full downloads on site URLs and Range downloads for these five models: 1 second at the 480p tier for both Grok variants, and 8 seconds at 720p for Omni / Veo. Other durations, resolutions and parameter combinations have not each been paid-tested.

The separate gemini-omni-1.1-flash-10s model did not return a task ID within a 180-second submission wait and is not currently published. That submission was not automatically repeated. This is not a confirmed generation failure.