Gemini Omni Flash 비디오 생성
Gemini Omni Flash로 텍스트, 최대 10개 참조 이미지 또는 최대 3개 입력 비디오를 사용해 비디오 생성 및 편집
중국 본토 사용자 안내: 중국 본토 사용자는 https://toapis.cn 를 API 엔드포인트(Base URL)로 사용해 주세요. 본 문서의 예시에서 https://toapis.com 을 https://toapis.cn 로 바꿔 주세요.
- Use
gemini-omni-flash-preview-officialin themodelfield - Supports text-to-video, image/reference-to-video, and video editing
- Supports up to 10 reference images and up to 3 input videos; images and videos can be combined in one request
- Video duration is limited to 10 seconds; the Playground offers 4 / 6 / 8 / 10 second presets
- Only
720presolution is supported; both16:9and9:16aspect ratios are available - Async task API: submit a task, then query by task ID
This page documents gemini-omni-flash-preview-official. The platform routes it through the Vertex AI Interactions API, not the Vertex AI Publisher Model predictLongRunning endpoint.
Upload local images with the Upload 이미지 API and local videos with the Upload Video API first, then pass the returned URLs. Do not pass base64 media data directly.
Authorizations
Authorizationstring헤더필수Use Bearer Token authentication:
Authorization: Bearer YOUR_API_KEYBody
modelstring본문필수기본값: gemini-omni-flash-preview-official모델 name. Use gemini-omni-flash-preview-official.
promptstring본문필수Text prompt for video generation.
durationinteger본문기본값: 6Video duration in seconds. The maximum is 10; the Playground offers 4, 6, 8, and 10 second presets.
aspect_ratiostring본문기본값: 16:9Video aspect ratio:
16:9landscape9:16portrait
resolutionstring본문기본값: 720pVideo resolution:
720pdefault resolution- Only
720pis supported; Veo-specific parameters are not accepted
image_urlsstring[]본문Optional reference image URL array with up to 10 items. Omit it for text-to-video. Can be combined with video_list, for example to attach reference images when extending a video.
video_listobject[]본문Optional input videos for video editing, with up to 3 items. Each object must contain video_url. Can be combined with image_urls for mixed image-plus-video input.
metadataobject본문Optional generation controls: temperature, topP / top_p, candidateCount / candidate_count, previous_interaction_id, and task (text_to_video, image_to_video, reference_to_video, edit, or extend). extend continues a source video supplied through video_list, cannot be combined with previous_interaction_id, and omits aspect_ratio.
응답
idstringTask ID for status polling.
objectstringObject type, usually generation.task.
modelstring모델 name used for the request.
statusstringTask status: queued, in_progress, completed, or failed.
created_atintegerTask creation timestamp.
video_urlstringGenerated video URL when the task completes.
Examples
요청 예시
curl --request POST \
--url https://toapis.com/v1/videos/generations \
--header 'Authorization: Bearer YOUR_API_KEY' \
--header 'Content-Type: application/json' \
--data '{
"model": "gemini-omni-flash-preview-official",
"prompt": "A blue pump bottle slowly rotates in a clean white studio, commercial product video",
"duration": 6,
"aspect_ratio": "9:16",
"resolution": "720p",
"image_urls": ["https://example.com/reference.jpg"]
}'Video extension example (duration is the continuation length in seconds, about 10 when omitted; the output is the source video plus the continuation):
curl --request POST \
--url https://toapis.com/v1/videos/generations \
--header 'Authorization: Bearer YOUR_API_KEY' \
--header 'Content-Type: application/json' \
--data '{
"model": "gemini-omni-flash-preview-official",
"prompt": "Continue the same scene and camera motion",
"duration": 5,
"video_list": [{"video_url": "https://example.com/input.mp4"}],
"metadata": {"task": "extend"}
}'응답 예시
{
"id": "video_01JZEXAMPLE",
"object": "generation.task",
"model": "gemini-omni-flash-preview-official",
"status": "queued",
"created_at": 1779247407
}Query Task
The generation endpoint returns a task ID. Use the common video task status endpoint to fetch status and results:
curl --request GET \
--url https://toapis.com/v1/videos/generations/{task_id} \
--header "Authorization: Bearer YOUR_API_KEY"