Google releases Gemini Omni 1.2 Flash with 15-second video generation

Extended limits and lower costs

Google officially released Gemini Omni 1.2 Flash on November 5, 2026. This update focuses on practical improvements for video creators, most notably by increasing the maximum generation length from 10 seconds to 15 seconds. Users can now select any duration between 3 and 15 seconds in one-second increments.

The model also introduces a more affordable pricing structure for developers and businesses. A 720p video now costs $0.07 per second via the API, which is roughly 30% cheaper than the previous version, Omni 1.1. This makes it easier for teams to produce high volumes of content like social ads or product explainers.

While the 15-second limit applies to standard generation, some specific workflows still have limitations. For instance, extending an existing clip or performing interpolation between start and end frames can still take between 3 to 10 seconds per step. These updates are designed to be a maintenance release rather than a total overhaul of the underlying technology.

Multi-step chat editing

One of the most significant functional changes in version 1.2 is the ability to perform multi-step editing directly in a chat interface. Instead of having to regenerate an entire video because of a minor error, users can now request specific changes through text prompts. You can ask the model to swap out a product, change the weather, or adjust the lighting.

This targeted approach is intended to save significant time and money for professional workflows. Because the API bills per second, making small corrections to a 15-second clip is much more efficient than starting from scratch every time a detail is off. It allows for a more iterative process where a draft can be refined round by round until it meets the desired specifications.

The model also supports changing the aspect ratio of a clip without needing to re-run the full generation process. This flexibility is particularly useful for marketing teams who need to adapt the same piece of content for different platforms. It moves the technology closer to a standard video editing suite where specific elements can be manipulated independently.

Visual and audio consistency

Google aims to solve the common problem of 'morphing' where objects or characters change shape during camera movements. Version 1.2 is designed to keep faces, clothing, and even text on packaging consistent from the first frame to the last. This is a notable improvement over version 1.1, which sometimes caused items like soup ladles to turn into chopsticks during pans.

The model also handles complex camera movements with fewer unwanted cuts. It can produce long, continuous shots such as drone flights or dramatic zoom-outs without breaking the scene. For example, Google demonstrated a shot flying through the Pantheon without a single cut, showcasing the model's ability to maintain spatial logic over longer durations.

Audio improvements also accompany these visual updates. The model is expected to provide more natural-sounding dialogue and cleaner lip-syncing. It also handles multi-speaker scenes more reliably by assigning the correct voices to the correct characters. These refinements make the tool more viable for creating short stories or realistic product demonstrations.

Availability and workflow

Gemini Omni 1.2 Flash is now generally available in the Gemini API and on Google Cloud for business users. It is also being rolled out across various Google products, including Google Vids and Google Ads. For users of Google Vids, the 15-second clips are included at no extra cost within the platform's monthly AI limits.

For those using the Gemini app, the new 15-second video capabilities are initially restricted to Google AI Ultra subscribers. It is currently unclear when these features will become available to Pro users. However, the model is already being integrated into Flow, Google's video editing tool, which now supports scene direction and multi-step edits.

Google suggests a specific workflow to maximize efficiency and minimize costs. They recommend creating drafts in lower resolutions like 360p or 720p first to correct errors through the editing tool. Once the sequence is perfect, users can then upscale the final version to 1080p or 4K. This strategy helps avoid paying full price for failed high-resolution generations.