YouTube brings DeepMind's Veo 3 and song remixing to Shorts

Staring at a blank editing timeline is the ultimate creator buzzkill. YouTube knows it, and it is deploying DeepMind’s heavy artillery so you never have to start from scratch again.

At its annual Made on YouTube event on September 16, the video giant rolled out an entire suite of generative AI tools tailored for Shorts. Leading the charge is a custom, speed-optimized video model called Veo 3 Fast, built directly into the mobile creation flow.

Instant drafts and turning spoken words into beats

The most practical addition is a feature called Edit with AI. Instead of manually combing through dozens of clips in your camera roll, you feed your raw footage into the app, and the model assembles a cohesive first draft. It picks the best moments, adds transitions, syncs background music, and can even generate a contextual voiceover in English or Hindi that reacts to what is happening on screen.

Then there is Speech to Song. If you stumble upon a hilarious quote or memorable dialogue in a Shorts clip, this feature lets you sample the spoken audio and flip it into a catchy soundtrack for your own video. Powered by DeepMind’s Lyria 2 music model, creators can pick vibes like chill, fun, or danceable, while original creators receive automatic attribution.

Fast 480p over cinema specs

For generative video, YouTube is prioritizing mobile speed over prestige fidelity. Veo 3 Fast renders at 480p, but it delivers low latency and, for the first time, generates audio right alongside the video clips. Creators in the US, UK, Canada, Australia, and New Zealand can already access it via the sparkle icon in the Shorts camera.

TechCrunch reported that more Veo features are queuing up for testing over the coming months. Those include motion transfer, which lets creators take dance choreography from a video and animate a still photo with it, as well as text-based object insertion and artistic restyling like pop art or origami. Everything generated through the suite carries DeepMind’s SynthID watermark and clear user-facing labels.

Why it matters

We think YouTube played this hand smartly by refusing to chase 4K benchmark vanity. On a smartphone screen, churning out a quick 480p gag clip with synced audio in seconds is infinitely more valuable than waiting three minutes for a high-res masterpiece.

The real win here is lowering the barrier between having raw phone footage and publishing a finished short. Edit with AI solves the blank-canvas paralysis that stops casual users from posting, effectively turning anyone's mundane camera roll into a passable draft.

For the broader industry, this raises the table stakes in the rivalry with TikTok. YouTube is not just handing creators a standalone generative sandbox; it is wiring AI directly into remix culture, letting creators sample existing audio into fresh tracks with attribution built in. The catch, however, is clear: as making videos becomes practically effortless, YouTube will need to prove its feed won't drown in homogenized, AI-narrated slop.