Omni Flash vs Omni 1.1 Flash: what changed and what Sume exposes
Original Omni Flash was 3-10 s at 720p; 1.1 adds 360p draft, 1080p, 4K, scene extension and frame control. What Sume's id exposes, read 2026-10-05.
Omni 1.1 Flash adds a 360p draft tier, 1080p and 4K, scene extension up to 40 seconds, first and last frame control and video references of up to 3 seconds, on top of the original Omni Flash, which Google described as 3 to 10 second clips at 720p. On Sume, gemini-omni-flash-1.1 exposes the resolution tiers, frame control, references and an edit mode, but not scene extension.
Both vendor pages were read on 2026-10-05: the August 27, 2026 launch post for 1.1 (Google blog) and the earlier Google Cloud post that describes the original (Google Cloud).
The two side by side
| Feature | Original Omni Flash | Omni 1.1 Flash (vendor) | Sume `gemini-omni-flash-1.1` |
|---|---|---|---|
| Clip length | 3-10 s | Scene extension to 40 s in 10 s steps | 3-10 s per clip |
| Resolution | 720p | 360p draft, 720p, 1080p, 4K | 360p, 720p, 1080p, 4K |
| Image references | Up to 7 | Not restated | Up to 10 |
| Video references | Up to 3 clips of up to 3 s | Up to 3 s | Up to 3 clips, each up to 3 s |
| First and last frame | Not listed | Yes | image_url + end_image_url |
| Edit existing video | Not covered here | Not covered here | video_url |
Reading the differences
Two entries need care. The image reference count differs, 7 on the vendor page for the original and 10 on the Sume row. Sume's number is from its own docs and is the one to build against for Sume. And scene extension is a vendor feature that the Sume docs say is not exposed: the API does not offer previous_interaction_id or extend, so a long scene is a chain of clips.
The 360p draft tier is the change with the biggest effect on cost. Google says it is up to 60 percent faster and a third of the cost of 720p. Sume's list price is $0.03 against $0.10 per second, so a ten-second draft is $0.38 against $1.25 after the 1.25 multiple and rounding.
Prices on Sume
| Resolution | Per second | 5 s | 10 s |
|---|---|---|---|
| 360p | $0.0375 | $0.19 | $0.38 |
| 720p | $0.1250 | $0.63 | $1.25 |
| 1080p | $0.1875 | $0.94 | $1.88 |
| 4K | $0.3750 | $1.88 | $3.75 |
What to re-test when you move
- Prompts that used 720p only: re-run at 360p to see which still read in draft.
- Pipelines that stopped at 10 seconds: plan chains of clips; there is no extend field.
- Anything with an image list: stay at or under 10 images and 3 clips per request.
- Anything that needs silence: native audio is always on, and the API rejects
generate_audio: false.
Why the id looks different
Google writes the model as Omni 1.1 Flash; Sume's catalog id is gemini-omni-flash-1.1. They name the same model. Use the Sume id in requests to Sume and never guess the id from the vendor's name. If a request names a model that is not in the catalog, the Video Router answers 404 model_not_found.
What the upgrade means for a budget
The upgrade changes the shape of a budget more than the price of a clip. A pipeline that ran every 720p clip at $1.25 per ten seconds can now draft at 360p for $0.38 and finish at 1080p for $1.88. If you ran three attempts per shot at 720p ($3.75), a three-draft plan with a 1080p final is $3.02. You get a higher-resolution deliverable for less money, as long as drafts predict finals well.
Premium tiers add cost without extending length. 4K is $3.75 for ten seconds on Sume, twice the 1080p price. Use them for finals only.
Where to read the live row
The catalog is the source of truth for Sume's limits: GET /v1/video-router/models/gemini-omni-flash-1.1 returns its capabilities and rates. Use that, not a blog table, in code.
curl https://api.sume.com/v1/video-router/models/gemini-omni-flash-1.1 \
-H "Authorization: Bearer $SUME_API_KEY"Sources
Related posts
More in Models
- Omni reference limits: 10 images and 3 clips versus Veo's 3 images
On Sume, Gemini Omni takes up to 10 reference images and 3 clips of 3 seconds each; Veo 3.1 takes three images and Lite none. Build the request in Python.
- GPT Image 2.5 can take 2 minutes: submit async, not sync, on Sume
OpenAI says complex GPT Image 2.5 prompts can run up to 2 minutes. Sume's image route blocks only 30 seconds, so send mode async and poll the job.
- GPT Image 2.5 edits through Sume: mask_url and 16 references
Sume's GPT Image 2.5 takes up to 16 references, an optional mask_url and background auto, transparent or opaque. OpenAI: transparent needs png or webp.
- Pocket TTS license: the repo says MIT, not Apache-2.0
Some roundups call Kyutai's Pocket TTS Apache-2.0. Its GitHub page and LICENSE file read MIT-style. How to check a TTS license before you build on it.
Written by Sume