Gemini Omni 1.2 Flash Adds 15-Second Video and Conversational Editing

Introduction

Google has introduced Gemini Omni 1.2 Flash, a video-generation model aimed at more continuous and repeatable production workflows. The update is not limited to making clips longer. Its central goal is to reduce the amount of re-generation required when users need to correct a detail, change a scene, or create multiple versions of the same asset. The model is becoming generally available through the Gemini API and business-oriented Gemini services, with a rollout across products such as Google Vids.

What changed

• Up to 15 seconds per clip: Native generation and editing now support clips from 3 to 15 seconds, selectable in one-second increments. That gives creators more room for product explainers, social ads, and short narrative sequences.
• Stronger reference adherence: The model can combine character, product, and style images with motion references of up to three seconds and audio cues. Google says it is better at retaining faces, clothing, packaging text, brand colors, and material details throughout a shot.
• Improved temporal consistency: Characters, objects, props, and environments are less likely to morph or drift during camera transitions, first-to-last-frame interpolation, and clip extension. Demonstrations include virtual try-ons, layered poster animation, and architectural fly-throughs.
• Conversational editing: Users can revise an existing clip over multiple turns, changing the background, product, weather, lighting, or aspect ratio without restarting from scratch. This could make it easier to produce variants for different audiences and placements.
• More natural speech and camera motion: Google highlights clearer lip synchronization, more distinct multi-speaker conditioning, and smoother dialogue. The model is also designed to handle continuous FPV flights, zooms, and orbits with fewer unwanted cuts.
• Lower listed cost: 720p generation is priced at $0.07 per second, a 30% reduction from Gemini Omni 1.1. Google recommends drafting at 360p or 720p, then upscaling selected results to 1080p or 4K.

Why it matters

The update illustrates a broader shift in generative video: the important question is increasingly whether a model can be corrected and reused, not merely whether it can produce an impressive first take. Consistent packaging, wardrobe, typography, and character identity are especially valuable for commercial teams, where a clip may need to be adapted to several formats or campaign concepts.

There are still clear boundaries. Google says rapid sports action, frequent script changes, and complex physics remain areas for ongoing improvement. A 15-second limit therefore does not guarantee a production-ready long take in every scenario; users may still need multiple generations, selection, and post-production. Even so, Omni 1.2 Flash positions itself as a workflow-oriented upgrade for developers, creators, and businesses rather than simply a larger video model.

Source: Hacker News