Omni scene extension to 40 s: Gemini app plans versus Sume per-second
Google offers Omni scene extension to 40 s in Flow and the Gemini app for Plus, Pro and Ultra. On Sume you pay per second instead; here is what carries over.

Google makes Omni 1.1 Flash scene extension available in Flow and the Gemini app to Plus, Pro and Ultra subscribers, up to 40 seconds in 10-second steps; developers get the model in AI Studio and the Gemini API (Google blog, read 2026-10-05). On Sume you pay per output second with no plan, but there is no extend button, so you chain clips yourself.
The question is who needs the plan. If you make clips by hand in a UI, the plan route is simpler. If your code makes clips, the per-second route is the one with an API.
Surfaces and who they are for
| Surface | Access | Scene extension | On Sume |
|---|---|---|---|
| Google Flow | Plus, Pro, Ultra subscribers | Per the vendor post | Not a surface; use the API |
| Gemini app | Plus, Pro, Ultra subscribers (scene extension) | Yes | Not a surface |
| Google AI Studio and Gemini API | Developers | Per the vendor post | gemini-omni-flash-1.1 on the Video Router |
| Sume Video Router | API key and balance | No extend field | Chain clips: last frame, reference clip, or edit |
What carries over
Everything the vendor post lists except extension: 360p, 720p, 1080p and 4K tiers, first and last frame control through image_url and end_image_url, and video references. The Sume catalog row says 3 to 10 seconds, 16:9 or 9:16, native audio always on, references up to 10 images and 3 clips of up to 3 seconds each (Video Router page).
Where it differs is billing. A plan covers use inside the Google product. Sume bills the provider list times 1.25 for each output second, rounded up to the cent per clip: $0.0375 a second at 360p, $0.125 at 720p, $0.1875 at 1080p and $0.375 at 4K.
A 40-second sequence in dollars
To approximate a 40-second extension, run four 10-second clips and join them. The prices are exact; the continuity is your job.
| Resolution | Four clips | Per second of film |
|---|---|---|
| 360p | $1.52 | $0.0380 |
| 720p | $5.00 | $0.1250 |
| 1080p | $7.52 | $0.1880 |
| 4K | $15.00 | $0.3750 |
How to decide
- One-off, hands-on: use the product with a plan, since you may already have one.
- Repeatable, many clips, code-driven: the API, because jobs, retries and idempotency keys are part of it.
- Need true extension with 10 s of context: that exists in Google's own surfaces. Do not expect it from a chain on Sume.
- Need a hand-off between the two: generate in either place, then bring the clips onto a Sume Timeline.
Resolution and cost, side by side
Google's post sets 360p as a draft at a third of the cost of standard 720p, and 1080p and 4K as premium. The Sume ladder has the same four tiers, and the draft tier is 30 percent of the 720p rate on the provider list ($0.03 against $0.10). If you plan a 40-second sequence, draft it at 360p first: four clips are $1.52 against $5.00 at 720p, before you spend on finals.
Not the same product
A subscription gives you a product with a history panel, credits and a timeline. The API gives you jobs, ids and a bill. Sume does not claim to replace the Google surfaces. It is a way to run the same model from code, with its own billing and limits, and you should compare the pieces that matter to you, such as concurrency, balance holds and result URLs.
A short decision guide
Ask three questions. First, does a person or a program start the clip? A person with a Plus, Pro or Ultra plan can use Google's own surfaces; a program needs an API. Second, do you need true extension with 10 seconds of prior context? Only the vendor surfaces list it. Third, do you need a bill per clip that you can attribute to a client or a job? Per-second API billing with job ids gives you that.
Many teams will use both: explore in the product, then move the winning prompt into the API for a batch.
Checks before committing
Read the plan terms on Google's side for your own region and account, since the vendor blog names the plan tiers but the pages change. On the Sume side, read the catalog row for gemini-omni-flash-1.1 from GET /v1/video-router/models rather than relying on a blog table.
Sources
Related posts
More in Comparisons
- OpenAI tts-1-hd at $30 per million characters vs Cartesia Sonic
tts-1-hd is $30 per million characters and tts-1 is $15. Cartesia Sonic is $38 at list on Scale overage, and Sume's Sonic-3.6 route is $47.50 per million.
- OpenAI TTS: 100+ languages but English-optimized voices; test yours
OpenAI's guide lists 100+ languages yet says its voices are optimized for English. Run a 1-cent test in your language on Sume's TTS router before choosing.
- A 'Pareto frontier' speech-to-text claim: plot your own cost and delay
Microsoft says MAI-Transcribe-2-Streaming sits on the Pareto frontier. A frontier needs two axes and your data. Plot Sume STT cost against measured job time.
- File-size limits as average bitrate for a 60-second clip
A 60-second clip can average 3.3 Mbps on free ArtStation, 6.7 on a Linktree background, 25 on Reels and 60 on Truth Social. Full table and script.
Written by Sume