Multimodal · NOVEMBER 8, 2026
Gemini Omni 1.2 Flash cuts video costs, but edits still need inspection
Google’s 15-second video model adds conversational editing and cheaper 720p generation. An early hands-on test shows why ad makers should still check continuity and sound.
Google made Gemini Omni 1.2 Flash generally available through the Gemini API and Google Cloud on November 5, adding longer clips and conversational editing while cutting its stated 720p generation price by 30% from its predecessor to $0.07 per second.
For small businesses making product videos or ad variants, the practical change is the ability to revise an existing clip with follow-up instructions rather than generate it afresh. That could make iteration more practical, but an early hands-on test found continuity and audio defects. Lower generation costs don’t establish better campaign results.
Longer clips and a revision workflow
Google says the duration ceiling rises from 10 seconds to 15 seconds, with selectable lengths in one-second increments starting at three seconds. Its launch post describes combining product or character images, short motion references and audio cues, then changing backgrounds, lighting or aspect ratios through conversation.
Operators can draft at 360p or 720p and upscale selected results to 1080p or 4K. That suggests a workflow in which owners review drafts before selecting assets for higher-resolution output, rather than treating every generation as a finished advertisement.
DeepMind’s Gemini Omni Flash - Model Card lists text, image, audio and video inputs, with video-and-audio output. There’s an access qualification: Google Cloud says audio references require allowlisting.
API availability is clearer than app access
The November 5 general-availability announcement covers the Gemini API and Google Cloud. The app picture is less consistent: Google’s main announcement describes availability across Gemini for business, while Google Cloud’s closing instructions call that interface "coming soon".
The Next Web also reports rollouts across Flow, Google Vids and Google Ads. Those reports shouldn’t be read as confirmation that every interface offers identical access. For an owner choosing where to produce a 15-second ad, the distinction between model availability and access through a particular application matters.
The edits still need a human check
DataCamp’s early hands-on account used a 15-second marble-run clip, followed by a color change and a stop-motion restyle. The color edit changed the marble while preserving timing, but the marble briefly disappeared and reappeared behind dominoes. The restyle changed textures while retaining smooth motion; sound drifted slightly across edits.
Google itself says fast sports action, rapid script changes and complex physics remain under development.
Neither the launch materials nor DataCamp’s test establishes improved campaign performance.
Neither the launch materials nor DataCamp’s test establishes improved campaign performance.
For a founder-led business, the defensible test is a small set of product or service video-ad variants made from approved existing photos or clips. Each finished asset should be inspected for continuity and sound, then tested with the intended audience for qualified inquiries or purchases before production is scaled.