- Blog
- Gemini Omni 1.2 Flash Explained: Specs, Pricing and Video Workflows
Gemini Omni 1.2 Flash Explained: Specs, Pricing and Video Workflows
on 10 days ago

Gemini Omni 1.2 Flash is Google's model for creating and editing video through prompts and references. For creators, the useful question is whether its controls fit a real brief: a product reveal, a short explanation, or a variation of an existing shot. This guide separates documented specifications from practical production advice, with sources checked on November 10, 2026.
The illustrations in this article are editorial concepts, not samples generated by Gemini Omni or screenshots of its interface.
What is Gemini Omni 1.2 Flash?
The Google DeepMind model card describes a model based on Gemini Omni Flash that accepts text, images, audio, and video, and produces video with audio. Its central use is video creation and conversational editing. That broad model description does not mean every service exposes every input type.
Google announced the release on November 5, 2026. Its launch overview highlights longer clips, closer adherence to references, and revisions made through follow-up instructions. Those are vendor claims about improvements, not results from a Shotyard benchmark. Evaluate them against the details your own project needs to preserve.
For a product video, that might mean checking whether a bottle cap stays attached during an orbit. For a character scene, it might mean watching whether clothing and facial features remain stable after a lighting change.
Specifications that affect your video brief
Use the limits for the actual service and operation you select. Google's Cloud model documentation lists the following:
| Setting | Documented Cloud specification |
|---|---|
| Model identifier | gemini-omni-1.2-flash |
| Generation and editing duration | 3–15 seconds |
| First-and-last-frame generation | 3–10 seconds |
| Image references | Up to 10 images per prompt |
| Frame shape | 16:9 or 9:16 |
| Resolution options | 360p, 720p, 1080p, and 4K |
| Video operations | Generation, editing, extension, and upscaling |
The shorter limit for first-and-last-frame generation matters. If your brief requires an exact opening and closing composition, plan within that mode's limit rather than assuming the headline 15-second duration applies. Extension is a separate operation, so check its constraints before planning a longer sequence.
Audio also needs care. The model card describes audio inputs, while the Cloud listing marks standalone audio unsupported. The Google Cloud announcement says audio references require allowlisting. Confirm access before building a brief around an uploaded voice reference.
Pricing: what does a short video cost?
Google Cloud's announcement gives a price of $0.07 per second for 720p video, described as 30% lower than Omni 1.1 Flash. At that advertised rate, a 10-second output corresponds to $0.70 and a 15-second output to $1.05. These estimates cover the advertised video-output rate; your project may incur other charges.
Budget for attempts as well as final clips. In a hypothetical project with four 10-second outputs, the same calculation gives $2.80 before any other applicable charges. If you keep only one, the cost of producing that usable shot includes the discarded takes. Record actual charges while evaluating a service instead of estimating only from the final running time.
Check the current billing terms for your chosen service, resolution, inputs, edits, and upscaling. An API price is not a subscription price, and third-party credits need their own calculation. On Shotyard, consult the pricing page and the generation cost shown for your selected settings. Do not assume Google's dollar rate converts directly into Shotyard credits.

Plan a controlled product shot before generating
Start with one visual event. A bottle turning slowly on a pedestal is easier to review than a sequence combining a room change, a speaking character, and several camera moves. Write down what must stay constant: bottle shape, cap, label placement, background color, and the position of the light.
Give each reference a job. A front view can establish the label; a side view can clarify the silhouette. Extra near-duplicates add little to that brief. Check that the images show the same packaging version before asking the model to preserve its details.
For a first-and-last-frame workflow, make both frames plausible parts of the same shot. Keep the object and environment consistent, then describe the movement between them. An original example brief is:
Create an eight-second product shot of a blue glass perfume bottle on an ivory pedestal. Begin with the supplied wide composition and finish on the supplied close-up. Move the camera slowly forward. Preserve the cap, bottle silhouette, and label position. Use warm side lighting and a quiet room ambience. Keep the shot continuous.
This is a planning example, not a tested prompt or a promise of exact adherence. Change its duration and controls to match the interface you actually use.
Review drafts, revise one detail, then export
Google's launch overview recommends drafting at 360p or 720p, correcting details, and upscaling selected results. It also describes improvements to dialogue and camera control while acknowledging ongoing work on fast sports action and complex physics. Treat an attractive first preview as a candidate for review.
Watch the full clip once at normal speed to judge pacing. Then inspect the beginning, middle, and end for identity changes. Pause when the product turns, passes behind something, or touches another object. A single attractive frame cannot tell you whether the whole shot is usable.
When something fails, change one detail in the next instruction. Ask for warmer lighting while preserving camera motion, for example, then compare both versions. Changing the subject, camera, setting, and sound together makes it harder to tell which revision helped.

For a practical review sheet, record the prompt, references, selected model, duration, resolution, charge, and reason for accepting or rejecting each take. Keep the original reference beside the video while checking product details. Approve the motion before spending time on the final export, then inspect that export again for framing and sound.
Check which controls are available on Shotyard
The Shotyard Gemini Omni page currently describes 4–10 second clips and up to seven references. Those published controls differ from Google's 1.2 Flash documentation. A shared family name alone does not confirm that an interface exposes the newest version or every Cloud capability.
Use that page to inspect Shotyard's available workflow, then verify the model and settings before generating. If your brief depends on a 15-second native clip, first-and-last-frame input, or conversational editing, confirm that the selected route explicitly supports it. Adapt the brief to the available controls rather than relying on a feature described for another service.
FAQ
Can Gemini Omni 1.2 Flash generate a 15-second video?
Yes. Google documents generation and editing from 3 to 15 seconds. Its Cloud specifications give first-and-last-frame generation a narrower 3–10 second range, so check the operation as well as the model name.
Does the $0.07-per-second price apply to every output?
No. The announced figure is for 720p video. Verify the current rate for other resolutions and operations, and distinguish direct API billing from a platform's subscription or credits.
Can I upload an audio reference on any service?
Do not assume so. Google Cloud says audio references require allowlisting. Audio generated within a video and permission to upload an audio reference are different capabilities.
Are the illustrations examples of this model's results?
No. They illustrate shot planning and draft review. This article is based on documented capabilities and editorial workflow advice, not a hands-on comparison of generated videos.
Conclusion
Gemini Omni 1.2 Flash offers documented video creation and editing controls worth evaluating against a specific brief. Start with one short shot, confirm the limits and price of your selected service, and judge continuity across the entire clip. Keep a record of accepted takes and their costs before expanding the workflow.
