Product Launch

Gemini Omni 1.2 Flash: 15-Second Video, Output Pricing and Editing

Q: Where can I try Gemini Omni 1.2 Flash through an API?

For EvoLink text-to-video requests, use gemini-omni-1.2-flash-text-to-video; other tasks use their own model IDs in the API reference.

What changes with Gemini Omni 1.2 Flash: longer clips, official video pricing, reference consistency, editing, and production evaluation.

EvoLink TeamProduct Team

November 6, 20268 min read

On this page

Google released Gemini Omni 1.2 Flash on November 5, 2026, bringing video generation and editing of up to 15 seconds and a lower advertised price of approximately $0.07 per second for 720p video output. It is generally available through the Gemini API and Google Cloud. Google also reports improvements in reference consistency, conversational editing, speech, and camera control. Google release announcement

For teams building video products, the useful question is how these changes affect a complete job: getting a usable product shot, revising it, and producing the final version. The update gives teams more room within a single clip and a lower video output rate, while keeping the choice of references, resolution, and revision strategy central to the cost.

What changes with Gemini Omni 1.2 Flash

The release extends the maximum generation and editing duration from 10 to 15 seconds. Google's advertised 720p price falls by 30% compared with Omni 1.1 Flash. On EvoLink, a 360p clip can be upscaled to 720p or 1080p; 4K upscaling requires a 720p or 1080p source. Google also highlights closer reference following and more natural dialogue. Google Cloud launch overview

Evaluate each change against a production requirement: does the extra duration fit the scene, do product or character details survive generation and editing, and does the complete job cost less? Compare finished clips and revision counts alongside the published output rates.

Which scenes can use a 15-second clip?

A product demonstration can need time for the viewer to recognize the item, see an action, and understand the result. A longer continuous shot may fit that sequence without stitching together separate generations.

For example, a team could brief a short film in which a camera approaches a coffee machine, follows the cup as coffee pours, and settles on the finished drink. The useful acceptance test is whether the product, action, and camera movement remain coherent through the whole shot. Filling all 15 seconds is optional; a simpler scene may work better at a shorter duration.

The maximum also depends on the operation. Google's Cloud documentation lists 3–15 seconds for generation and editing, while first-and-last-frame interpolation remains 3–10 seconds. Extension adds 3–10 seconds rather than applying the same 15-second rule to every request. Google Cloud model specifications On EvoLink, using more than one reference video limits the output to 10 seconds.

What should you check after editing the first draft?

For a catalog team, the practical test is specific: does the same bag retain its handle shape, material, and branding as the camera moves? For an ad variant, can the background change while the approved product remains recognizable?

A useful review compares the original reference, the first generated clip, and the edited version. Check the details the edit was supposed to change, then inspect the details that were supposed to stay fixed. This makes the decision concrete: keep the revision, try a narrower instruction, or generate another take.

Google says Omni 1.2 follows visual references more closely and supports refining clips through successive conversational edits. One launch demonstration combines a model photo with front-and-back product shots of three outfits to create a 15-second virtual try-on. Another changes backgrounds over multiple turns while retaining the characters and framing. These vendor demonstrations illustrate the intended control over a scene. Google release demonstrations

Does Omni 1.2 Flash cost less than 1.1?

For 720p video output at the Gemini API Standard rate, yes. Whether the complete job costs less also depends on input usage, text or thinking output, revisions and retries.

Google's Gemini API Standard pricing lists $12 per million video output tokens, down from $17.50 for Omni 1.1 Flash. At 5,792 tokens per second for 720p, that works out to approximately $0.0695 per second, rounded to $0.07 in the announcement. Input is priced separately at $1.50 per million tokens, and text output, including thinking, at $9 per million tokens. Official Gemini API pricing

A 15-second 720p clip therefore represents about $1.04 in video output charges at that Standard rate. That is a calculation from the published token rate, before input and text or thinking charges.

For budgeting, count revisions as well as initial generations. A lower rate helps, but a project with several discarded takes can still cost more than one completed with fewer attempts. Compare total spend against the number of clips that meet the brief.

When should you draft first and then upscale?

When exploring several concepts, start with lower-resolution drafts and upscale the selected take. If most first drafts are already acceptable, compare the combined draft, edit and upscale cost against producing the required resolution directly. Judge the complete workflow cost per accepted clip.

The official Gemini API guide identifies 1080p and 4K as upscaled outputs. Google video generation guide

This suggests a practical sequence: judge composition and motion in a lower resolution draft, resolve the main creative changes, and then upscale the selected take. Inspect the final file again, particularly where the shot includes packaging text, fine patterns, or small product details.

Availability depends on the product and API channel

For EvoLink text-to-video requests, use gemini-omni-1.2-flash-text-to-video; other tasks use their own model IDs in the API reference. Developers can start with the stable upstream model ID gemini-omni-1.2-flash. For existing Gemini API integrations, Google says requests to gemini-omni-1.1-flash will automatically route to 1.2 from December 3, 2026, at the new rates. That transition applies to the Gemini API; teams using another channel should check its own migration policy. Gemini API model documentation

The model generates video with synchronized audio, including dialogue. Google Cloud offers audio references through allowlisting; this feature is unavailable on EvoLink, which does not accept audio files as input. Google Cloud

For teams evaluating access through a unified gateway, EvoLink's Gemini Omni 1.2 Flash page lists six task routes covering generation from text, images, or references, plus editing, extension, and upscaling. It provides a Playground and token cost estimates. Use the task-specific documentation and current account pricing when moving from evaluation to integration.

What to test before switching a production workflow

Google acknowledges that fast sports action, rapid script changes, and complex physics remain areas for improvement. Those scenes deserve particular attention in an evaluation. Google release limitations

Start with a small set of briefs your team already understands: one product shot, one controlled edit, and one longer camera move. Compare product fidelity, motion, audio, revisions, and total cost against the same acceptance criteria used for the current model.

Switch when the finished clips meet your acceptance criteria at a suitable total cost. Base that decision on accepted results, correction work and cost per usable clip.

Gemini Omni 1.2 Flash FAQ

What changed from Omni 1.1 Flash to Omni 1.2 Flash?

The maximum generation and editing length rises from 10 to 15 seconds, and Google’s advertised 720p video output price falls by 30%. Google also highlights closer reference following and multi-turn conversational editing. Google

How long can a Gemini Omni 1.2 Flash video be?

API generation and editing support 3–15 seconds. First-and-last-frame interpolation supports 3–10 seconds, and extension adds 3–10 seconds per operation. Choose the duration for the operation you are using. Google documentation On EvoLink, using more than one reference video limits the output to 10 seconds.

How much does a 15-second 720p clip cost?

At Google’s Standard video output rate, about $1.04: 15 seconds × 5,792 tokens per second × $12 per million tokens. Input and text or thinking usage are charged separately. For EvoLink route estimates, use the existing calculator. Google pricing

Can Omni 1.2 Flash produce 1080p or 4K video?

On EvoLink, a 360p clip can be upscaled to 720p or 1080p; 4K upscaling requires a 720p or 1080p source. EvoLink

Can I edit an existing video with Omni 1.2 Flash?

Yes. Provide the input clip and describe the change, such as a new background, lighting, or product detail. Conversational edits let you refine the clip over successive turns. For product work, name the visual details to preserve alongside the requested change. Google

Where can I try Gemini Omni 1.2 Flash through an API?

For EvoLink text-to-video requests, use gemini-omni-1.2-flash-text-to-video; other tasks use their own model IDs in the API reference.

Can Omni 1.2 Flash generate sound, and can I supply an audio reference?

All Posts

#Gemini Omni 1.2 Flash #Video Generation