Google has added three new creative controls to Flow, its AI video and image studio: keyframe control for start and end frames, 4K video export, and cheaper 360p test renders. The update, which rolled out worldwide on August 27, 2026, is based on the new Gemini Omni 1.1 Flash model and is available to all paying Flow subscribers.
Flow, which Google unveiled at I/O in May 2025, brings together models like Veo, Nano Banana, and Gemini under a single interface. Until now, guiding a clip toward a specific final shot was tedious: the model only looked at the last second of existing footage to decide how to continue, which often caused breaks in motion, lighting, or camera angle when multiple scenes were strung together.
Start and end frame control
With Gemini Omni 1.1, Flow now analyzes up to ten seconds of existing footage before continuing its calculations. Users can specify an exact start frame and end frame in increments of ten seconds, up to a total of 40 seconds at a time. You design a scene that begins with a character at the door and ends inside the room, and Flow fills in the movement in between.
Up to three seconds of separate video footage can be imported as a reference to keep characters and environments consistent across multiple clips. For commercials, short films, or social media content with recurring characters, this replaces the old one-second logic, where you could never be entirely sure if a character would still be wearing the same jacket in the next clip.
4K export and cheaper drafts
Finished clips can now be exported directly in 1080p or 4K, making Flow suitable for broadcast and professional social media workflows that require fixed resolution specifications. The third feature targets cost: if you just want to test whether a prompt works, you can render a 360p preview. According to Google, these drafts are generated up to 60 percent faster and cost about one-third of a regular 720p rendering. Only once the result is satisfactory is it upscaled to full resolution.
All three features run on Gemini Omni 1.1 Flash, a faster version of Google's multimodal Omni model first introduced in May 2026. Omni models process text, images, audio, and video within a single system, which lets the model understand longer video segments simultaneously and maintain movement, lighting, and perspective more consistently across multiple frames than the previous version.
Availability
The update is live worldwide in Flow itself and in the Gemini app. Access is available to subscribers of the Google AI Plus, Pro, and Ultra tiers. Developers can also access the features via Google AI Studio and the Gemini Enterprise Agent Platform. Users without a paid subscription remain limited to Flow's basic features. The new tools are available at labs.google/fx/tools/flow.
For those working with AI video professionally, the update addresses a core problem: each new prompt run turns out slightly different. Start and end frame control fixes the beginning and end of a clip, leaving only the path in between to be generated. Combined with more affordable test renders, this moves Flow from a gimmick toward a tool that enables planned production. Anyone publishing clips created with Flow should also check current guidelines for labeling AI-generated content, especially for commercial use in Germany.
Why this matters for creatives
For video editors, motion designers, and content creators, the practical gain is predictability. You can now lock in a character's appearance and a scene's lighting across multiple clips, export at broadcast quality, and test prompt variations at a fraction of the previous cost. That combination - fixed endpoints, professional output resolution, and cheap iteration - is what makes Flow usable for client work and scheduled production rather than just experimentation. If you're already working with generative video tools, this update is worth testing against your current workflow. If you're new to the space, the 360p draft mode lowers the barrier to learning how prompts behave before committing credits to full renders.
Your membership also unlocks:
