gemini-omni-flash-preview

视频Googlegemini-omni-videogemini-omni-flash-preview

Open Playground

gemini-omni-video

Authentication

Base URL: https://host.docker.internal:8000
Authorization: Authorization: Bearer YOUR_API_KEY
Content-Type: application/json
Model name: gemini-omni-flash-preview

Integration protocols

Google Interactions官方

Why this protocol

适合使用 Google Interactions 原生接口提交多模态视频任务并轮询状态的用户。

Endpoint

POST/v1beta/interactions

Authorization: Bearer YOUR_API_KEYContent-Type: application/json

Request / Response JSON

// REQUEST

{
  "model": "gemini-omni-flash-preview",
  "background": true,
  "input": [
    {
      "type": "image",
      "data": "BASE64_IMAGE_DATA",
      "mime_type": "image/jpeg"
    },
    {
      "type": "text",
      "text": "让参考图中的主体自然运动，保持外观一致，使用单一连续镜头。"
    }
  ],
  "generation_config": {
    "video_config": {
      "task": "image_to_video"
    }
  },
  "response_format": {
    "type": "video",
    "aspect_ratio": "16:9",
    "delivery": "uri"
  }
}

// RESPONSE

{
  "id": "interaction_example",
  "object": "interaction",
  "status": "completed",
  "model": "gemini-omni-flash-preview",
  "steps": [
    {
      "type": "model_output",
      "content": [
        {
          "type": "video",
          "mime_type": "video/mp4",
          "uri": "https://host.docker.internal:8000/v1beta/files/file_example:download"
        }
      ]
    }
  ],
  "usage": {
    "total_input_tokens": 22,
    "total_output_tokens": 58805,
    "total_tokens": 59183
  }
}

Request / Response schema

// INPUT SCHEMA

Field	Type	Required
model	string	Y
background	boolean	Y
input	array	Y
input[]	object	Y
input[].type	string	Y
input[].data	string	Y
input[].mime_type	string	Y
generation_config	object	Y
generation_config.video_config	object	Y
generation_config.video_config.task	string	Y
response_format	object	Y
response_format.type	string	Y
response_format.aspect_ratio	string	Y
response_format.delivery	string	Y

// RESPONSE SCHEMA

Field	Type	Required
id	string	Y
object	string	Y
status	string	Y
model	string	Y
steps	array	Y
steps[]	object	Y
steps[].type	string	Y
steps[].content	array	Y
steps[].content[]	object	Y
steps[].content[].type	string	Y
steps[].content[].mime_type	string	Y
steps[].content[].uri	string	Y
usage	object	Y
usage.total_input_tokens	integer	Y
usage.total_output_tokens	integer	Y
usage.total_tokens	integer	Y

Code examples

cURLrequest.sh

curl -X POST "https://host.docker.internal:8000/v1beta/interactions" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
  "model": "gemini-omni-flash-preview",
  "background": true,
  "input": [
    {
      "type": "image",
      "data": "BASE64_IMAGE_DATA",
      "mime_type": "image/jpeg"
    },
    {
      "type": "text",
      "text": "让参考图中的主体自然运动，保持外观一致，使用单一连续镜头。"
    }
  ],
  "generation_config": {
    "video_config": {
      "task": "image_to_video"
    }
  },
  "response_format": {
    "type": "video",
    "aspect_ratio": "16:9",
    "delivery": "uri"
  }
}'

# Poll status cURL
# Replace interaction_example with the interaction_id returned by the create response
curl -X GET "https://host.docker.internal:8000/v1beta/interactions/interaction_example" \
  -H "Authorization: Bearer YOUR_API_KEY"