Cold open with a recurring cast: Gemini Omni Flash with 10 image refs
Gemini Omni Flash 1.1 on Sume takes up to 10 image refs and 3 video refs of 3 s each for a 3 to 10 s vertical cold open with synced audio. Request and limits.

Gemini Omni Flash 1.1 on Sume makes a 3 to 10 second cold open from up to 10 reference images and up to 3 reference clips of at most 3 seconds each, in 16:9 or 9:16, with synced audio always on. Put your recurring cast in the references and address them as <IMAGE_REF_0> in the prompt. It is a short-form tool; for 30 second shots use Seedance 2.5.
A cold open is the few seconds before the title, and episodic micro-dramas, which Metricool reports as gaining traction on TikTok, lean on it to hook the viewer.
Limits from the docs
Sume routes this model by the shape of the request, so you do not pick an endpoint. Read capabilities from GET /v1/video-router/models instead of assuming.
| Item | Value |
|---|---|
| Model id | gemini-omni-flash-1.1 |
| Duration | 3 to 10 s |
| Resolution | 360p, 720p, 1080p, 4K |
| Aspect ratio | 16:9 or 9:16 |
| Image references | Up to 10 |
| Video references | Up to 3, each up to 3 s |
| Audio | Native, always on |
| Billing | Provider list x 1.25 per output second by resolution |
The request
Send reference_image_urls and, if you have motion to carry over, reference_video_urls. Reference media is addressed by its zero-based position in the list. Media must be on media.sume.com. The prompt below uses two cast images and one clip.
curl -X POST https://api.sume.com/v1/video-router/generate \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: s1e5-cold-open-v1" \
-d '{"model":"gemini-omni-flash-1.1","aspect_ratio":"9:16","resolution":"720p","duration":8,
"reference_image_urls":["https://media.sume.com/artifacts/artf_a/lead.png","https://media.sume.com/artifacts/artf_a/rival.png"],
"prompt":"<IMAGE_REF_0> runs through a night market, <IMAGE_REF_1> watches from a rooftop, tense hush"}'Rules you must respect
Native audio cannot be switched off: generate_audio set to false is rejected. There is no reference_audio_urls. A video_url edit source cannot be combined with image or reference lists. So a cold open with your own voice track needs the sound replaced afterwards in Timeline 1.0.
Keeping the cast stable
Reference images are the lever. Use a clean, front-facing image of each character on a plain background, name them in the prompt by ref, and reuse the same files in every episode. Our reference image guide covers building the set.
Sume does not promise identical faces from one render to the next. Check the first frame against your approved image before posting.
Sources
Related posts
More in Models
- Reference audio for AI video: which models accept a voice clip
Seedance 2.5, Wan 3.0 and MiniMax H3 take reference audio on Sume; Kling 3, Gemini Omni Flash and the swap rows do not. Limits and the one-reference rule.
- Reference videos in AI video on Sume: Wan, H3, H3 Max and Omni limits
How many reference images, videos and audio files each Sume video model accepts: Wan 10/5/5, H3 9/3/3, Omni 10/3 with 3-second clips, Kling none.
- Omni Flash vs H3 Max for reference-to-video: limits side by side
Gemini Omni Flash 1.1 takes 10 images and 3 short videos; MiniMax H3 Max adds audio refs, 12 files total. Limits, tags and a sample request on Sume.
- Runway Veo negativePrompt: 1,000 characters, and Sume's options
Runway's API takes an optional negativePrompt of up to 1,000 characters on three Veo models. Sume's video request lists no such field, so use positive wording.
Written by Sume