Google just turned its workplace video tool into an AI video studio — with you as the star. On July 16, the company rolled out Gemini Omni, its latest media generation model, inside Google Vids, alongside personal avatars that let users cast a verified digital likeness of themselves in AI-generated clips. Google likened the integration of Gemini Omni Flash to "Nano Banana — but for video," signaling that the viral energy of its image-editing model is now aimed squarely at the enterprise video market.
From Prompt to Finished Clip
The headline capability is text- and image-to-video generation woven directly into a mainstream productivity app. Users start with a written prompt and can add reference images — a product photo, even a rough sketch — which Omni blends into a matching video. But the deeper shift is in how edits work:
- Conversational, step-by-step editing lets users refine a clip in plain language — swapping backgrounds, fixing lighting or adding effects — without regenerating from scratch.
- The editing works on existing footage, not just AI output, so a shaky phone recording can be cleaned up or restaged with the same commands.
- Because changes are incremental, Vids sidesteps the frustrating slot-machine loop that still defines many AI video generators.
That iterative editing model addresses one of generative video's most persistent complaints: promising first drafts that collapse the moment you ask for a small change.
Your Face, Verified and Watermarked
The second pillar of the update is personal avatars. From a single selfie and roughly ten seconds of recorded voice, Vids builds a digital avatar that looks and sounds like the user, which can then be cast as a character in Omni generations — turning a script into a talking-head video without cameras, lighting or reshoots.
Given the obvious deepfake risks, Google has wrapped the feature in unusually tight guardrails. Avatars are restricted to the account holder's own likeness, created through a secure verification process and managed directly in the user's Google Account. Every generated clip is invisibly watermarked with SynthID, the provenance technology developed by Google DeepMind. At launch, avatars are English-only, limited to adults 18 and over, and unavailable in the European Economic Area, Switzerland and the UK — a geographic carve-out that hints at regulatory caution as much as engineering rollout.
Availability and the Competitive Chessboard
The features are live for Google AI Pro and Ultra subscribers and Workspace business customers, with Rapid Release domains beginning a gradual rollout on July 16 and Scheduled Release domains following from August 5. Admins can toggle the capability at the domain level, though it ships enabled by default.
Strategically, the update pushes Vids far beyond its origins as an AI-assisted presentation tool and into territory occupied by avatar-video specialists like HeyGen, Synthesia, Captions and D-ID. TechCrunch noted the timing is pointed: with OpenAI's Sora reportedly wound down, Google is betting there is durable demand for AI video where people star in their own content — especially inside the corporate workflows it already dominates. Training videos, sales outreach, internal comms and onboarding content are the obvious first targets.
Why It Matters
This launch marks the moment consumer-grade generative video became a checkbox feature of office software. For years, AI video lived in standalone tools with separate subscriptions and steep learning curves; now it ships inside the same suite as Docs and Slides, distributed instantly to millions of Workspace seats. That distribution advantage is exactly how Google previously commoditized grammar checking and image generation — and it puts immediate pricing pressure on venture-backed avatar startups whose core product just became a bundled feature.
The avatar guardrails matter just as much as the generation quality. By binding likenesses to verified accounts and watermarking output with SynthID, Google is effectively proposing an industry template for consent-based synthetic media — one regulators in Brussels and London will scrutinize closely, given the feature's conspicuous absence from those markets at launch. Whether that framework holds up against misuse will shape how quickly personal avatars spread from corporate training videos to the open social web.
For the multimodal AI race, the message is clear: the frontier is no longer just generating video, but editing it conversationally and personalizing it safely. On both fronts, Google just raised the bar.
