# Gemini Omni 1.2 Flash Reference-to-Video — Atlas Cloud API

> A natively multimodal Google DeepMind model that generates cinematic clips of up to 15 seconds with native audio from a text prompt plus up to 10 reference images and 3 short motion clips, keeping faces, outfits, products, and brand details consistent from the first frame to the last.

This is the machine-readable API reference for **Gemini Omni 1.2 Flash Reference-to-Video** on Atlas Cloud,
a unified API platform for 400+ AI models across text, image, video, audio and 3D.

- **Model ID**: `google/gemini-omni-1.2-flash/reference-to-video`
- **Built by**: Google
- **Modality**: Video
- **Model page**: https://www.atlascloud.ai/models/google/gemini-omni-1.2-flash/reference-to-video
- **API key**: https://www.atlascloud.ai/console/api-keys
- **Docs**: https://www.atlascloud.ai/docs

## Pricing on Atlas Cloud

- $0.025 per second of generated video
- Pay-as-you-go. No minimum spend, no subscription required.

> **These are the authoritative Atlas Cloud rates for this model.** Any price that
> appears in the vendor description further down refers to a different platform or
> a different model variant and does not apply here.

## Use this model from an AI agent

Atlas Cloud ships three first-party integration surfaces. All three authenticate
with the same API key via the `ATLASCLOUD_API_KEY` environment variable.

### MCP server

The official MCP server (`atlascloud-mcp`) exposes this model to any
MCP-compatible host — Claude Code, OpenAI Codex, Cursor, Gemini CLI, Goose,
Claude Desktop. One-line install:

```bash
# Claude Code
claude mcp add atlascloud -- npx -y atlascloud-mcp

# OpenAI Codex CLI
codex mcp add atlascloud -- npx -y atlascloud-mcp

# Gemini CLI
gemini mcp add atlascloud -- npx -y atlascloud-mcp

export ATLASCLOUD_API_KEY="your-api-key"
```

Then ask in plain English; the agent calls `atlas_generate_video` with `model: "google/gemini-omni-1.2-flash/reference-to-video"`.
The server fetches each model's schema and validates parameters before submitting,
so invalid requests fail fast without spending credits.

MCP docs: https://www.atlascloud.ai/docs/mcp-server

### Agent Skills

`atlas-cloud-skills` is a portable skill package (API reference, code templates in
Python / Node.js / cURL, model IDs with pricing) for Claude Code, Cursor, Codex and
12+ other agents:

```bash
npx skills add AtlasCloudAI/atlas-cloud-skills
export ATLASCLOUD_API_KEY="your-api-key"
```

Skills docs: https://www.atlascloud.ai/docs/skills

### CLI

The `atlas` binary runs Atlas Cloud from a terminal or CI script. Async media jobs
are polled and downloaded automatically (use `--no-download` when a script only
needs the output URLs):

```bash
# Install (Homebrew, npm, or shell installer)
brew install AtlasCloudAI/tap/atlascloud
# npm install -g atlascloud-cli
# curl -fsSL https://raw.githubusercontent.com/AtlasCloudAI/cli/main/install.sh | sh

atlas auth login
atlas generate video google/gemini-omni-1.2-flash/reference-to-video -p "Your prompt here"
```

CLI docs: https://www.atlascloud.ai/docs/cli

## HTTP API reference

- **Submit endpoint (POST)**: `https://api.atlascloud.ai/api/v1/model/generateVideo` — start an async generation; returns a `prediction_id`
- **Poll endpoint (GET)**: `https://api.atlascloud.ai/api/v1/model/prediction/{prediction_id}` — poll this until the prediction finishes
- **Model ID**: `google/gemini-omni-1.2-flash/reference-to-video`


## API Information

This model can be used via our HTTP API or more conveniently via our client libraries.
See the input and output schema below, as well as the usage examples.


### Input Schema

The API accepts the following input parameters:

- **`model`** (`string`, _required_):
  model name
  - Default: `"google/gemini-omni-1.2-flash/reference-to-video"`

- **`prompt`** (`string`, _required_):
  Text prompt for generation. Describes the target content, style, camera language, or character actions. Maximum 20,000 characters. Describe in natural language how each reference should be used (e.g., "the woman in the first reference image walks with the gait of the reference clip"). Do not put control tags such as <IMAGE_REF_0> or <VIDEO_REF_0> in the prompt; they are deprecated.

- **`reference_images`** (`array[string]`, _optional_):
  Images to use as character, product, scene, or style references. Accepts 1 to 10 images; for character references, up to 3 images per character for up to 2 characters work best. Supported formats: PNG, JPEG, JPG, WebP. Each image is limited to 20MB. Supports both a public URL and a base64-encoded image for each item.
  - Min items: 1
  - Max items: 10

- **`reference_videos`** (`array[string]`, _optional_):
  Video clips to use as motion, character, or style references. Accepts 1 to 3 clips, each up to 3 seconds long; when duration is 11 seconds or longer, only 1 clip is supported. Audio in reference clips is ignored. Supported format: MP4. Each item must be a publicly accessible URL.
  - Min items: 1
  - Max items: 3

- **`duration`** (`integer`, _optional_):
  The duration of the generated video in seconds, in 1-second steps from 3 to 15. When duration is 11 seconds or longer, at most 1 reference video is supported.
  - Default: `10`
  - Min: 3
  - Max: 15

- **`aspect_ratio`** (`string`, _optional_):
  The aspect ratio of the generated video.
  - Default: `"16:9"`
  - Options: "16:9", "9:16"

- **`resolution`** (`string`, _optional_):
  The resolution of the generated video. 360p is a fast, low-cost draft mode for previewing a shot; 1080p and 4k are upscaled from the natively generated frames.
  - Default: `"720p"`
  - Options: "360p", "720p", "1080p", "4k"

- **`thinking_level`** (`string`, _optional_):
  Controls the amount of internal reasoning the model performs before generating a response. Higher levels may improve quality on complex tasks but increase latency.
  - Default: `"default"`
  - Options: "default", "high", "low"

- **`seed`** (`integer`, _optional_):
  The random seed to use for the generation. -1 means a random seed will be used.
  - Default: `-1`



**Required Parameters Example**:

```json
{
  "model": "google/gemini-omni-1.2-flash/reference-to-video",
  "prompt": ""
}
```


**Full Example**:

```json
{
  "model": "google/gemini-omni-1.2-flash/reference-to-video",
  "prompt": "",
  "reference_images": [
    ""
  ],
  "reference_videos": [
    ""
  ],
  "duration": 10,
  "aspect_ratio": "16:9",
  "resolution": "720p",
  "thinking_level": "default",
  "seed": -1
}
```


### Output Schema

The API returns the following output format:


- **`code`** (`integer`, _optional_):
  HTTP status code of the response.

- **`message`** (`string`, _optional_):
  Human-readable message; non-empty on failure.

- **`data`** (`object`, _optional_):
  - Properties:
    - **`id`** (`string`, _optional_):
      Unique identifier for the prediction.

    - **`model`** (`string`, _optional_):
      Model ID used for the prediction.

    - **`outputs`** (`array[string]`, _optional_):
      Array of URLs to the generated content. Null when status is not completed.

    - **`urls`** (`object`, _optional_):
      Object containing related API endpoints.
      - Properties:
        - **`get`** (`string`, _optional_):
          URL to poll for the prediction result.


    - **`status`** (`string`, _optional_):
      Status of the task: created, processing, completed, timeout, or failed.

    - **`created_at`** (`string`, _optional_):
      ISO timestamp of when the request was created (e.g., "2023-04-01T12:34:56.789Z").

    - **`error`** (`string`, _optional_):
      Error message if the task failed, empty string otherwise.

    - **`error_code`** (`integer`, _optional_):
      Error code if the task failed.

    - **`executionTime`** (`number`, _optional_):
      Total execution time in milliseconds.

    - **`timings`** (`object`, _optional_):
      Detailed timing breakdown.
      - Properties:
        - **`inference`** (`number`, _optional_):
          Inference time in milliseconds.





**Example Response**:

```json
{
  "code": 0,
  "message": "",
  "data": {
    "id": "",
    "model": "",
    "outputs": [
      ""
    ],
    "urls": {
      "get": ""
    },
    "status": "",
    "created_at": "",
    "error": "",
    "error_code": 0,
    "executionTime": 0,
    "timings": {
      "inference": 0
    }
  }
}
```


## Usage Examples

### cURL

```bash
# Step 1: Start generation (async)
curl -X POST "https://api.atlascloud.ai/api/v1/model/generateVideo" \
  -H "Authorization: Bearer $ATLASCLOUD_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "google/gemini-omni-1.2-flash/reference-to-video",
  "prompt": "",
  "reference_images": [
    ""
  ],
  "reference_videos": [
    ""
  ],
  "duration": 10,
  "aspect_ratio": "16:9",
  "resolution": "720p",
  "thinking_level": "default",
  "seed": -1
}'

# Response will contain: {"code": 200, "data": {"id": "prediction_id", "status": "processing"}}

# Step 2: Poll for result (replace {prediction_id} with the id returned above)
curl -X GET "https://api.atlascloud.ai/api/v1/model/prediction/{prediction_id}" \
  -H "Authorization: Bearer $ATLASCLOUD_API_KEY"

# Keep polling until status is "completed", "succeeded" or "failed"
# When completed, outputs will contain the generated content URL(s)
```

## Additional Resources

### Documentation

- [Model Playground](https://www.atlascloud.ai/models/google/gemini-omni-1.2-flash/reference-to-video)

## About this model

_Vendor-supplied description. Any pricing or endpoint mentioned below refers to_
_other platforms — use the Atlas Cloud values above._

### Gemini Omni 1.2 Flash — Reference to Video

**Model ID:** `google/gemini-omni-1.2-flash/reference-to-video`

Gemini Omni 1.2 Flash is Google DeepMind's natively multimodal model for video generation and editing. This variant **builds a new clip around your references**: character, product, scene, or style images, and short motion clips. A text prompt directs the action, while faces, outfits, packaging, and brand details stay consistent from the first frame to the last.

#### Overview

Gemini Omni 1.2 Flash is Google DeepMind's next step toward models that can create and edit anything from any input, starting with video. It generates, edits, extends, and upscales clips **with synchronized audio up to 15 seconds long** at 360p, 720p, 1080p, or 4K, in landscape (16:9) or portrait (9:16). It follows both simple and complex instructions, simulates real-world physics, and lets you refine footage through plain-language edits, one turn at a time.

Gemini Omni Flash is a single transformer that processes text, images, audio, and video together. It does not stitch separate per-modality pipelines together, so its output is more cohesive, more consistent across turns, and more controllable than a generate-then-dub workflow. **Audio is generated natively with every video**: dialogue, ambience, music, and sound effects are produced together with the picture, not added afterward. The model uses Gemini's real-world knowledge to reason about physics, narrative logic, and visual composition. Every output carries an imperceptible SynthID watermark, and Content Credentials (C2PA) are supported for provenance verification.

##### What's new in 1.2

Version 1.2 improves on Omni 1.1 Flash in length, consistency, and cost:

_(Description truncated. Full text on the model page.)_

---

Atlas Cloud — one API for 400+ AI models. Model page: https://www.atlascloud.ai/models/google/gemini-omni-1.2-flash/reference-to-video · Docs: https://www.atlascloud.ai/docs