Google’s Veo 3.1 brings sound and smart edits to AI filmmaking

The most frustrating part of AI filmmaking has always been the deafening silence. You get an incredible visual cut, but there is zero audio, forcing you into a tedious round of sound hunting and sync editing.

Google is tackling that workflow pain head-on. Five months after launching Flow, its Veo-powered creative video suite—which Google says has already clocked over 275 million generations—the company just rolled out Veo 3.1 alongside major granular editing updates.

Audio across the entire board

Veo 3.1 builds directly on May's Veo 3 launch, promising tighter prompt adherence, more realistic surface textures, and noticeable audiovisual improvements when turning still images into clips.

What matters most in practice is that audio is no longer treated like an afterthought. Google has brought generated native sound into virtually every existing Flow feature.

That includes Ingredients to Video, where you use reference photos to lock down characters and styles, as well as Frames to Video, which stitches two benchmark frames into a smooth transition. Even the Extend tool—which continues action based on the last second of a previous cut to build continuous scenes lasting a minute or more—now outputs matching audio.

Precision surgery inside your frames

Flow is also borrowing a page from professional image editors by adding direct, in-canvas adjustments.

A new Insert feature lets you drop fresh objects or mythical creatures directly into an existing shot. Instead of generating a weird, floating sticker, Flow calculates surrounding light bounce and cast shadows so the addition actually blends in.

Google also confirmed that a Remove tool is coming soon to Flow, letting creators brush away unwanted bystanders or props while the background automatically reconstructs itself. Outside of Flow, Veo 3.1 is landing across the Gemini app, Vertex AI, and the Gemini API.

Why it matters

We think Google is finally shifting the AI video race from casino-style prompt rolling to actual timeline control. If a director cannot alter an element in a take without starting from scratch, AI remains a gimmick for real productions. Adding localized insertions and frame-to-frame bridging gives creators genuine artistic leverage.

Natively attaching audio to core generation steps is a smart move that saves hours in post-production. But there is a catch: generating past a minute using chained extensions still risks character drift and visual hallucinations over time. Still, while competitors focus primarily on raw 5-second eye candy, Google looks determined to build an actual editing suite.