Google on July 16, 2026, added two capabilities to Google Vids, its Workspace video creation tool: a generative model called Gemini Omni that produces and edits clips from written prompts, and personal avatars that let a user build a digital stand-in from a single selfie and a short voice recording. The features are restricted to paying subscribers, and avatar access carries regional and age limits that narrow who can use it.

The update was published on Google's Keyword blog and attributed to Justin Luk, a Product Manager at the company. It arrives three and a half months after Google opened basic video generation inside Vids to every account holder, and it moves the tool further from timeline-based editing toward a workflow driven entirely by natural language.

What Gemini Omni does inside Vids

Gemini Omni handles two jobs that previously required separate effort: generating a video from scratch and refining one after the fact. According to Google, a user starts with a plain-language text prompt describing the desired footage and can attach image references, such as a photograph or a rough sketch, to guide the output. The model combines those inputs and returns a clip intended to match the description.

The editing side is where the shift is most visible. Getting a first draft exactly right usually takes adjustments, and Google states those adjustments can now be made by describing them in ordinary language. A user can prompt Vids to swap a background, correct lighting, or add effects, and the instruction applies whether the source is a clip generated by Omni or footage shot on a phone. Because the model supports step-by-step edits, changes accumulate without forcing a restart from the beginning.

That step-by-step handling addresses a specific limitation in earlier AI video tools, where a single re-prompt often regenerated an entire clip and discarded prior work. The described behavior keeps the base video intact while layering discrete modifications on top.

The model behind the feature

Gemini Omni is not new to Google's product line, and its arrival in Vids follows a pattern of the same model appearing across consumer and advertising surfaces on a compressed timeline. The model was introduced at Google I/O 2026 in May, where the company described it as combining Gemini reasoning with generative media across text, images, and video simultaneously. Days later at Google Marketing Live 2026, Google confirmed the same model would be added to Asset Studio, the creative hub inside Google Ads, over the summer, positioning it as the engine for multimodal video production in campaigns.

The deployment logic is horizontal. YouTube brought Gemini Omni Flash remixing and personal avatars to Shorts at the same I/O event, and the model's presence in Vids, Shorts, and Asset Studio within roughly two months reflects a single capability being distributed across Google's video products rather than three unrelated launches.

Vids itself has a short but active history. The application launched on April 9, 2024 as part of Workspace and reached general users in a basic form during September 2025. In February 2026, Google rolled out Veo 3.1 to all Vids users, and on April 2, 2026, the company expanded the tool with free Veo 3.1 generation, Lyria 3 music composition, and directable AI avatars for all account holders. The July 16 update sits on top of that foundation.

The underlying video technology traces back through Google's earlier filmmaking products. Veo 3.1 succeeded the Veo 3 Fast and Veo 3 Quality variants Google introduced through its Flow platform and the Gemini API during 2025. In July 2025, Google added speech generation to Flow, and later that month it launched Veo 3 Fast for developers at 0.40 dollars per second with audio, a reduced rate against the 0.75 dollars per second of the standard Veo 3 model. According to Google, Veo 3.1 improved on those iterations by maintaining character appearance across multiple generations even when settings change, a consistency problem that had persisted in earlier AI video models. Gemini Omni now sits above that video layer as the model coordinating text, image, and video generation in a single workflow.

Personal avatars and the selfie-to-video path

The second feature targets a narrower use case: sending a quick video message without appearing on camera. According to Google, a user creates a personal avatar by uploading a selfie and a short voice recording. After the avatar is built, the user types what they want said, and the avatar delivers the message with the person's likeness and voice, with no live recording required.

Google placed several restrictions on the feature. Personal avatars are linked to a user's Google Account and are limited to the account holder's own likeness, meaning the tool is not designed to replicate other people. Access is currently available only to users in certain regions who are 18 or older. Google did not specify in the announcement which regions are included or excluded.

Those constraints matter for a capability that reconstructs a person's face and voice from minimal input. A selfie and a brief audio clip are low barriers to entry, and the account-linking and likeness restrictions function as the guardrails limiting how the reconstruction can be used. Concern about synthetic likeness in Google's video models is not hypothetical: an antisemitism monitoring group documented users adapting Google's Veo3 model to generate hateful video content beginning in July 2025, illustrating how generative video tools can be turned toward abuse once they reach scale. Restricting avatars to a verified account holder's own likeness is the design response to that class of risk.

Who can access the update

Both Gemini Omni and personal avatars are gated behind paid tiers. According to Google, the features are available to Google AI Pro and Ultra subscribers and to Google Workspace business customers. Free Google account holders, who received Veo 3.1 generation in the April rollout, do not get Gemini Omni editing or avatars under this release.

That paywall marks a divergence from the April expansion, which extended core generation to everyone. The most capable editing model and the avatar feature are reserved for subscription revenue, drawing a line between free generation and premium production inside the same product.

Watermarking and content transparency

Every clip generated through the new tools carries an embedded marker. According to Google, each generated clip includes an invisible SynthID digital watermark, which the company frames as a way for people to verify that a video was created with AI.

SynthID is Google DeepMind's watermarking system. DeepMind introduced it in 2024 as a suite of tools for watermarking and identifying AI-generated content across text, audio, images, and video. The marker is imperceptible to human perception but machine-readable, and it is designed to persist through subsequent editing and reformatting.

Google has been building consumer-facing verification around the same technology. On December 18, 2025, the company added SynthID verification to the Gemini app, letting users upload a video and check whether it was created or edited with Google AI. That system covers both audio and visual tracks, supports files up to 100 MB and 90 seconds, and returns timestamp-specific feedback distinguishing fully synthetic content from partially edited media.

The automatic watermarking in Vids is consistent with how Google has treated its other Gemini Omni deployments. Content produced through YouTube's avatar and Omni remixing tools carries automatic SynthID and C2PA labels with no opt-out, meaning creators who use the platform's own AI tools always have a machine-readable disclosure signal attached.

Why this matters for the marketing community

The relevance to advertisers and marketers runs along two lines: production economics and regulatory disclosure.

On production, the same Gemini Omni model reaching Vids is the model Google is threading into its advertising stack. At Google Marketing Live 2026, the company said Asset Studio now understands a marketer's brief, brand guidelines, website, and campaign goals, then generates assets across creative themes through a natural language interface, with Gemini Omni bringing video creation into the same workspace used for image and text. Google's Josh Moser presented creative as driving nearly half of all incremental ad sales as the rationale for the investment. A marketer testing the natural-language editing behavior in Vids is exercising the same interaction model that will govern campaign creative production.

On disclosure, the automatic watermarking lands inside a tightening compliance environment. The EU AI Act, formally Regulation (EU) 2024/1689, makes its Article 50 transparency obligations applicable on August 2, 2026, and advertising content is explicitly excluded from the lighter disclosure regime available to genuinely artistic or satirical work, meaning ordinary full labeling applies to persuasive material. Google has already moved on the ads side: on July 9, 2026, the company introduced an AI label setting rolling out across five products including Google Ads, Display and Video 360, and Merchant Center, shifting disclosure from something advertisers might volunteer into a control enforced at the interface level.

Consumer sentiment supplies the backdrop. Research the IAB published in January 2026 found Gen Z consumers nearly twice as likely as Millennials to feel negatively toward AI ads, with 39 percent reporting negative sentiment against 20 percent of Millennials, and nearly half of surveyed consumers wanting disclosure for AI voices or AI avatars in advertising. A tool that manufactures a synthetic version of a person's face and voice from a selfie sits directly inside the category of content those consumers said they want flagged.

The infrastructure and the concern advance together. Google is distributing a model that lowers the cost of producing synthetic video to a text prompt, and it is attaching a persistent, machine-readable watermark to everything that model makes. Whether buyers and platforms filter on that signal remains a separate question, but the July 16 update adds another surface where the watermark is applied by default.

Timeline

  • April 9, 2024 - Google Vids launches as part of Google Workspace
  • May 16, 2024 - DeepMind introduces SynthID for watermarking AI-generated content
  • July 10, 2025 - Google adds speech generation to its Flow AI filmmaking platform
  • July 31, 2025 - Google launches Veo 3 Fast for developers at 0.40 dollars per second
  • July 2025 - Users begin adapting Google's Veo3 model to generate antisemitic video content
  • September 2025 - Google begins rolling out a basic version of Vids to Workspace users
  • December 18, 2025 - Google adds SynthID verification to the Gemini app for video content
  • February 2026 - Google rolls out Veo 3.1 to all Google Vids users
  • April 2, 2026 - Google expands Vids with free Veo 3.1 generation, Lyria 3 music, and directable AI avatars for all account holders
  • May 19, 2026 - Google introduces Gemini Omni and personal avatars for YouTube Shorts at Google I/O 2026
  • May 20, 2026 - Google confirms Gemini Omni integration into Asset Studio at Google Marketing Live 2026, planned for summer
  • July 9, 2026 - Google introduces an AI label setting across Google Ads, Display and Video 360, Campaign Manager 360, Merchant Center, and Ads Editor
  • July 16, 2026 - Google adds Gemini Omni editing and personal avatars to Google Vids for paid subscribers
  • August 2, 2026 - EU AI Act Article 50 transparency obligations become applicable

Summary

Who: Google, with the announcement attributed to Justin Luk, a Product Manager at the company, publishing on the Keyword blog.

What: Two additions to Google Vids: Gemini Omni, a model that generates and edits video from text prompts and image references using step-by-step natural-language editing, and personal avatars built from a selfie and a short voice recording. Both features are limited to Google AI Pro and Ultra subscribers and Google Workspace business customers. Avatars are restricted to the account holder's likeness and available only to users 18 or older in certain regions. Every generated clip carries an invisible SynthID watermark.

When: July 16, 2026, following the April 2, 2026 rollout of free Veo 3.1 generation and avatars to all account holders.

Where: Inside Google Vids, the video creation tool within Google Workspace, with avatar access geographically limited.

Why: The update distributes Gemini Omni, the same model Google is integrating into Asset Studio for advertising creative, across another product surface, while the default SynthID watermarking aligns the tool with tightening AI disclosure requirements including the EU AI Act's Article 50 obligations that become applicable on August 2, 2026.