# Gemini Omni 1.2 Flash stretches clips to 15 seconds

_DeepMind's Gemini Omni 1.2 Flash adds native 15-second generation, tighter reference adherence and multi-turn editing at $0.07/sec for 720p._

**Published:** 2026-11-06
**Source:** https://www.startuphub.ai/ai-news/artificial-intelligence/2026/gemini-omni-1-2-flash-stretches-clips-to-15-seconds

---

[DeepMind](https://deepmind.google/blog/gemini-omni-12-flash-fast-and-cost-effective-with-longer-clips/) is rolling out Gemini Omni 1.2 Flash with native 15-second video generation, stronger reference adherence and multi-turn conversational editing, generally available across the Gemini API, Google Cloud, Google Vids and Flow.

Can you make a finished ad or explainer in one generation without stitching clips? For many product and character shots, yes, now that you can generate 3 to 15 seconds in one-second increments.

Gemini Omni is a multimodal model optimized for video generation, editing, extension and upscaling alongside text responses, and 1.2 Flash is the fast, cost-effective member of that family. The update moves the ceiling from 10 seconds in Omni 1.1 to 15 seconds, and reference adherence, which means the model holds onto a provided face, product or style image, is tighter from first frame to last. DeepMind shows it combining one model photo with front-and-back shots of three outfits to produce a 15-second virtual try-on with 360-degree turns, and rebuilding a dusty-pink cyclorama poster for HALO, GLOW and LUMEN layer by layer without drifting materials, shadows or typography.

So what does cheaper and longer actually change for production? Price falls 30% from 1.1 to $0.07 per second for 720p video according to DeepMind, and the workflow is designed to avoid full re-renders. You can draft at 360p or 720p, fix a detail with a follow-up prompt such as swapping a background or changing aspect ratio, and then upscale only the winner to 1080p or 4K. In one demo a single clip is iterated turn by turn through five backgrounds from a striped pavilion to a stone castle while keeping performance, armor reflections and framing intact, and another routes four distinct voice samples to the right characters with clearer lip-sync.

Consistency and camera control are the other practical gains. DeepMind says the model holds character, object and scene identity with less morphing across motion and transitions, whether you interpolate between first and last keyframes or extend a sequence, and it steers continuous moves like an FPV flight through the Pantheon to the oculus with fewer jump cuts and steadier scale. The limits are still stated plainly. Development is ongoing for fast sports action, rapid script changes and complex physics, so those shots may still need extra takes.

Developers can prototype in [Google](https://www.startuphub.ai/ai-news/ai-research/2026/deepmind-s-scale-how-agents-run-at-google) AI Studio and ship via the Gemini API with Provisioned Throughput and an imperceptible SynthID watermark, while teams using Google Vids or the Gemini app can direct scenes and multi-turn edits without extra cost on eligible Workspace plans. Watch whether the 15-second, edit-in-place pattern replaces the cut-and-stitch workaround for catalog spins and social spots, and whether the promised 30% saving holds once you count upscales and throughput.

---

Original analysis from [startuphub.ai](https://www.startuphub.ai), the #1 AI startup directory.

© StartupHub.ai. All rights reserved. You may not republish this article in full without a license. Search engines and AI research tools may crawl and summarize for reference. Bulk reproduction or model training on this content requires a license. See https://www.startuphub.ai/terms.