Keep a Shorts series consistent with Omni Flash video references
Gemini Omni Flash 1.1 on Sume accepts image and video references for 3 to 10 second clips. Feed the last episode's clip in as the next one's reference.

Pass the previous episode's clip, or a still from it, as an input_references entry on the next Gemini Omni Flash 1.1 request. Sume's catalog lists that model at 3 to 10 seconds, 360p to 4K, 16:9 or 9:16, with image and video references and no audio references.
A reference is guidance, not an exact frame, so expect the look and wardrobe to carry over and the pose to change. Use it to hold a series together, not to freeze a shot.
What the model takes
The docs describe gemini-omni-flash-1.1 as accepting image and video input_references and generating native synced audio. input_references is the reference-to-video mode: the model uses the images as visual guidance. If you also send frame_images, those control the mode and the request is processed as image-to-video.
So the choice is between two continuity tools. Reference mode keeps the cast and set. First-frame mode keeps the exact opening shot. A series usually wants reference mode for every episode and first-frame mode only for a deliberate match cut.
| Field | Value |
|---|---|
| Duration | 3 to 10 seconds |
| Resolutions | 360p, 720p, 1080p, 4K |
| Aspect ratios | 16:9, 9:16 |
| References | Image and video (no audio) |
| Audio | Native synced audio |
| 720p price | $0.125 per second |
A practical loop
Keep a small reference sheet per recurring character, and add the last good clip of the previous episode on top. Ten seconds at 720p is $1.25, so a six-beat, one-minute episode is $7.50 before re-rolls.
Each request is still its own job. Nothing in the catalog promises that two requests return the same face, and the no-seed post explains why you cannot pin one.
- Name the character the same way in every prompt.
- Reuse the same two or three reference images across a season.
- Re-roll one 10-second beat instead of the whole episode.
Related checks
If the shot is right and one detail is wrong, the edit route may be cheaper than a re-roll; see edit or re-roll. Confirm the numbers on the Videos page first.
Cost of a continuity pass
A reference does not change the price: Omni Flash 1.1 bills per output second, at $0.0375 for 360p, $0.125 for 720p, $0.1875 for 1080p and $0.375 for 4K in Sume prices. A 10-second beat at 360p is $0.375, so a draft pass of six beats costs $2.25 before you spend on the finals.
That makes a two-pass routine cheap. Draft the whole episode at 360p with your references, read it for continuity, then re-run only the beats that held together at 720p.
Sources
Related posts
More in Models
- Gemini Omni 'no dialogue' prompt: steer the audio on Sume
Omni clips always carry sound. Google suggests 'no dialogue' and 'no extra sound effects' in the prompt. How to use that on Sume and what a test costs.
- Omni replaces Veo in the Gemini app: does your API code change?
Google's page says Omni replaces Veo in the Gemini app. What that means for API code, and the one Omni row Sume lists, with its limits and per-second price.
- Gemini Omni sound effects prompt: name each sound in 8 seconds
How to prompt footsteps, a door and breaking glass in a Gemini Omni clip: Google says describe audio explicitly. Sume request, timing and cost per take.
- Open-weights video news does not change your Sume bill: LTX, H3, Wan
LTX-2.5 and MiniMax H3 are reported open weights, but Sume still bills hosted per-second rows. Which ids to call and what each costs.
Written by Sume