Google Releases Gemini Omni 1.1 Flash Video Generator: Scene Stitching Up to 40 Seconds, Draft Mode, and 4K Output

Google has moved its flagship multimodal video generation model, Gemini Omni 1.1 Flash, from testing to stable production release (General Availability). The updated tool can now seamlessly stitch and extend video sequences up to 40 seconds while maintaining character and environment context. It also adds interpolation between keyframes and introduces an ultra-cheap draft mode priced at $0.03 per second.

Scene Extension to 40 Seconds and 10 Seconds of Context Memory

Gemini Omni 1.1 Flash’s base generation still produces short clips ranging from 3 to 10 seconds. However, version 1.1 introduces a full Scene Extension feature that lets you sequentially extend a scene’s runtime in 10-second increments, up to 40 seconds.

According to Google’s official blog, the main engineering breakthrough is a tenfold increase in the context window:

• the experimental version only looked at the last second of the source clip;
• version 1.1 analyzes up to 10 seconds of previous video, accurately preserving characters’ appearance, clothing, camera movement, lighting, and synchronized audio (including voices and background noise);
• seams between iterations are smoothed by softly redrawing boundary frames.

Cheap 360p Mode and Pricing Tiers

To help developers save money, Google has added a draft mode Draft Mode at 360p resolution. These clips generate 60% faster than standard 720p videos and cost about three times less, allowing quick testing of prompts, angles, and composition before final rendering.

According to Google AI for Developers’ pricing page, inference costs break down as follows:

The technical documentation explicitly notes an important detail: 1080p and 4K are upscaled from the base 720p stream, not native 4K renders.

Frame Interpolation and Conversational Editing

Gemini Omni 1.1 Flash now includes first-and-last-frame interpolation: you can upload a start and end image, provide a text description of the transition, and the neural network will calculate smooth camera movement, object morphing, or create a seamless looping clip. For style control, you can attach up to three seconds of reference video.

According to the Gemini API documentation, the tool retains native multimodality:

• the model simultaneously accepts text instructions, static images, audio files, and video streams;
• via the Interactions API, you can conversationally edit a finished scene (replace a specific object, change the weather, or adjust the angle via text request without destroying the rest of the frame);
• all generated content is protected by an invisible cryptographic watermark, SynthID.

Benchmark Leadership and Old Version Shutdown

In the authoritative blind user test Text-to-Video Arena, the Gemini Omni 1.1 Flash model took first place with 1515 points, beating FLUX 3 Video and Seedance 2.5. In the image-to-video category, the service holds second place, trailing only MiniMax H3.

introducing Gemini Omni 1.1 Flash

this model brings a new suite of creative controls and generative video capabilities to developers

– extend scenes for longer storytelling
– specify first and last frames
– draft videos more efficiently in 360p
– upscale up to 4K resolution
-… pic.twitter.com/24Bmqz2k4Z

— Google AI Studio (@GoogleAIStudio) August 27, 2026

The stable model is now available under the identifier gemini-omni-1.1-flash in Google AI Studio, on the Flow platform, and in the Gemini interface for subscribers of Google AI Plus, Pro, and Ultra plans. The experimental branch gemini-omni-flash-preview has been deprecated and will be fully shut down on September 30, 2026.