Google has launched Gemini Omni 1.2 Flash, the latest model of its video era and enhancing mannequin. Gemini Omni 1.2 Flash generates and edits clips as much as 15 seconds natively. It additionally cuts the value by 30% in contrast with Omni 1.1 Flash. The mannequin is generally available on Google Cloud and the Gemini API. For builders, the takeaways are longer clips, fewer re-renders and a decrease invoice per second.
TL;DR
- Runs on: Google-hosted solely (Gemini API, Google Cloud). No open weights, no native deployment.
- Efficiency: Google studies higher reference adherence and fewer drift. No 1.2 benchmark scores are printed.
- Finest: Native 15-second clips at $0.07 per second of 720p video, about $1.05 per 15-second clip.
- Worst: Google says quick sports activities motion, speedy script adjustments and complicated physics nonetheless want work.
What’s Gemini Omni 1.2 Flash?
Gemini Omni 1.2 Flash is a multimodal video mannequin from Google DeepMind. It takes textual content, pictures, brief video clips and audio as enter. It outputs video with audio. Google’s developer docs describe the Omni line as natively multimodal, with enhancing dealt with by the Interactions API.
What modified from Omni 1.1 Flash?
The launch lists 6 upgrades:
- Longer clips: Era and enhancing now attain 15 seconds, up from 10 seconds in Omni 1.1.
- Length management: You set size in 1-second steps from 3 to fifteen seconds.
- Reference adherence: You possibly can mix character, product or model pictures with movement clips as much as 3 seconds.
- Consistency: Faces, clothes, packaging textual content and model colours maintain throughout digicam angles with much less morphing.
- Speech: Dialogue sounds extra pure, with clearer lip-sync and distinct multi-speaker conditioning.
- Digicam management: FPV flythroughs, zoom-outs and orbits present fewer undesirable soar cuts.
How does the draft, edit, upscale workflow work?
Google frames value management as a workflow, not only a worth. You draft at 360p or 720p. You repair a element with a follow-up edit as a substitute of regenerating the entire clip. Then you definitely upscale one of the best consequence to 1080p or 4K.
Multi-turn conversational enhancing now works throughout full 15-second clips. In observe, you’ll be able to swap a product, background, lighting or facet ratio flip by flip. The docs present this by a previous_interaction_id, so the prior video is just not re-uploaded.
How a lot does Gemini Omni 1.2 Flash value?
Google costs Omni 1.2 Flash at $0.07 per second of 720p video. Financial savings are similar at other resolutions. For context, the pricing page payments Omni 1.1 Flash at roughly $0.10 per second of 720p video. A full 15-second 720p clip on 1.2 works out to about $1.05.
Google Vids customers can generate 15-second clips at no extra value. Provisioned Throughput is out there, and outputs carry a SynthID watermark.
How does it evaluate with different video fashions?
| Function | Gemini Omni 1.2 Flash | Gemini Omni 1.1 Flash | Veo 3.1 | Seedance 2.0 |
|---|---|---|---|---|
| Developer | Google DeepMind | Google DeepMind | ByteDance | |
| Parameters | Not disclosed | Not disclosed | Not disclosed | Not disclosed |
| Max single clip | 15 s | 10 s | 8 s | 15 s |
| Length choices | 3 to fifteen s, 1 s steps | Not disclosed | 4, 6 or 8 s | 4 to fifteen s |
| Output decision | 360p, 720p; 1080p and 4K upscaled | 360p, 720p; 1080p and 4K upscaled | 720p, 1080p, 4K | 480p, 720p native |
| Native audio | Sure | Sure | Sure | Sure |
| Multi-turn enhancing | Sure, as much as 15 s | Sure | Not listed | Not disclosed |
| 720p worth | $0.07/s | About $0.10/s | Not listed on supply web page | Token-based, varies |
| Weights | Closed, API solely | Closed, API solely | Closed, API solely | Closed, API solely |
(perform(){var f=doc.getElementById(‘mtp-omni12-frame’);window.addEventListener(‘message’,perform(e){if(e.knowledge&&e.knowledge.sort===’mtp-omni12-h’&&e.supply===f.contentWindow){f.model.peak=e.knowledge.peak+’px’;}});})();
<!– Fallback: add gemini-omni-1-2-flash-embed.html to /wp-content/uploads/ by way of FTP and use –>
How does it carry out on benchmarks?
Google has not printed benchmark scores for model 1.2. Its claims relaxation on side-by-side demos in opposition to Omni 1.1. The DeepMind model page studies human-rater outcomes for the Omni Flash line, together with 1,003 MovieGenBench prompts. In June, an earlier Omni Flash model took #1 on the Arena text-to-video board. These rankings predate 1.2, so deal with them as household context solely.
Early associate suggestions is constructive however self-selected. In line with Google’s launch web page, invideo’s VP of Engineering stated the mannequin sticks to the immediate and that physics and sound really feel actual. QuickFrame AI says 15 seconds at 1080p lets customers construct a full CTV advert in 1 era.
The place can builders use it?
- Gemini API: Prototype in Google AI Studio and construct with the developer docs.
- Google Cloud: Use the Agent Platform API with enterprise SLAs.
- Google apps: Move, Google Vids, Google Advertisements and the Gemini app (rolling out to AI Extremely).
Key Takeaways
- Gemini Omni 1.2 Flash generates and edits native 15-second clips, up from 10.
- Worth drops 30% to $0.07 per second of 720p video.
- Multi-turn enhancing now spans full 15-second clips with out re-generation.
- No benchmark scores for 1.2 are printed but.
- Audio references stay allowlist-gated on Google Cloud.
Take a look at the launch post, the Google Cloud announcement and the developer docs.. Additionally, be happy to comply with us on Twitter and don’t overlook to hitch our 150k+ML SubReddit and Subscribe to our Newsletter. Wait! are you on telegram? now you can join us on telegram as well.
Have to associate with us for selling your GitHub Repo OR Hugging Face Web page OR Product Launch OR Webinar and so forth.? Connect with us
The publish Google Releases Gemini Omni 1.2 Flash: 15-Second AI Video at $0.07 per Second appeared first on MarkTechPost.

