Gemini Omni 1.1 Flash: AI Video Generator

How to Use Gemini Omni 1.1 Flash

1. Write a multimodal prompt Describe the visuals, camera direction, dialogue, and ambient sound together. Example: 'A street musician plays a violin on a rainy Paris evening, raindrops tap the cobblestones, a slow melancholic melody, distant café chatter.'

2. Choose one input mode

   • Reference images (image_urls, up to 7) for image-to-video.
   • A first/last frame pair (first_frame_url + last_frame_url) for keyframe-controlled generation — mutually exclusive with images, video, audio, and character IDs.
   • A reference video (video_url + trim_start/trim_end, max 10-second window) for continuity-driven extension.
   • Voice profiles (audio_ids) and/or character references (character_ids) layered on top of images or video.

3. Respect the shared quota Images count 1 unit each, the reference video counts 2 units, and each character ID counts 1 unit — the total must not exceed 7.

4. Pick duration, resolution, and aspect ratio Duration (4/6/8/10s) is ignored when a reference video is supplied — the model determines output length. Resolution is billed per second of output: $0.10/s at 720p, $0.15/s at 1080p, $0.30/s at 4K.

5. Submit and poll

   curl -X POST https://api.muapi.ai/api/v1/gemini-omni-flash-1-1 \
     -H "Authorization: Bearer YOUR_API_KEY" \
     -H "Content-Type: application/json" \
     -d '{
       "prompt": "A street musician plays a violin on a rainy Paris evening",
       "image_urls": ["https://example.com/scene.jpg"],
       "duration": 8,
       "resolution": "1080p",
       "aspect_ratio": "16:9"
     }'
   

   Then poll GET /api/v1/predictions/{request_id}/result until status is completed.

6. Prompt tips

   • Describe sound explicitly: 'soft piano underscore', 'wind through trees'.
   • For dialogue, quote the spoken line and describe the speaker.
   • Specify camera moves — 'slow dolly in', 'handheld tracking shot' — for cinematic results.