Verdict: Google's July 2026 Gemini rollout is a four-part upgrade that turns AI from a separate app you open into something baked directly into the tools you already spend your day in. The biggest is Gemini Omni in Google Vids — you can now capture your real face once, use it as a reusable personal avatar, generate video clips from text prompts, and edit those clips conversationally. NotebookLM has been renamed Gemini Notebook and gained direct integration with Google Search and the Gemini app plus secure cloud-side code execution. Gmail's Help me write now accepts free-text refinement prompts in its own bar so you can iteratively fix a draft instead of starting over each time. And Google Docs picked up eleven new supported languages and a reimagined in-doc Gemini experience. Useful for: any business writing scripts, training videos, multilingual docs, or routine email. Skip it if: you have no paid Google AI Pro, AI Ultra, or Workspace plan that qualifies for Gemini in Workspace, because nearly every feature below is gated behind a subscription tier.
- Best all-rounder: Gemini Omni in Vids for AI-generated video with your own face and voice.
- Best for research: Gemini Notebook (formerly NotebookLM) now deeply tied into Search and the Gemini app — 30M+ users, code execution natively.
- Best for email: Gmail's Help me write refinement bar — iterate on a draft instead of throwing it away.
- Best for multilingual docs: Google Docs Gemini adds 11 languages including Mandarin, Dutch, Polish, and Turkish.
- Best free-ish surface: Personal avatars and Omni in Vids are available to Google AI Pro/Ultra subscribers and Workspace business plans — not free-tier users.
What are the new Google Gemini updates rolled out in July 2026?
Google shipped four Workspace-adjacent Gemini updates between July 14 and July 17, 2026, all of which integrate the model deeper into products people already use rather than launching a standalone flagship (blog.google, July 2026; Google Workspace Updates blog). They are: Gemini Omni + Personal Avatars in Google Vids, the NotebookLM → Gemini Notebook rename, a free-text refinement bar in Gmail's Help me write, and an 11-language expansion for Gemini in Google Docs. Each release is gated by plan and region, and every generated clip includes a SynthID invisible watermark for AI-content transparency.
The four updates at a glance
| Update | Surface | What changed | Who can use it |
|---|---|---|---|
| Gemini Omni + Personal Avatars | Google Vids | Generate video from prompts/image refs; chat-to-edit; use your face-and-voice avatar as a character | Google AI Pro/Ultra, Workspace Business/Enterprise (18+, English-only, not in EEA/UK/CH at launch) |
| Gemini Notebook (NotebookLM rename) | notebooklm.google | Same product, new name; now appears inside Gemini app + Google Search AI mode; secure per-notebook cloud code execution expands to Pro | Free personal accounts, Workspace Business/Enterprise, AI Pro/Ultra |
| Help me write refinement bar | Gmail | Accept any free-text refinement prompt on an existing draft (not just Formalize/Shorten presets); undo/redo | Gemini-enabled Workspace plans; personal Google One AI Premium |
| Gemini in Docs — 11 new languages | Google Docs | Help me create, Help me write, Match writing style, Match doc format in Mandarin, Dutch, Malay, Hebrew, Polish, Turkish, Czech, Indonesian, Swedish, Danish, Norwegian | Workspace plans + Google AI Pro/Ultra |
How does Gemini Omni in Google Vids use your own face?
Gemini Omni is Google DeepMind's mixed-input video model that combines Gemini's core intelligence with image-to-video and video-to-video generation, and it is now available directly inside Google Vids to generate and edit high-quality videos with text prompts, image references, or both (Google DeepMind; blog.google, July 16 2026). Working alongside it are Personal Avatars — a feature completely separate from Gemini Omni's preset character gallery — which lets you capture your actual face and voice once and then direct that avatar in any future video without setting up a camera or recording live.
What personal avatars do
The avatar workflow has four steps, per Google's official support docs and the Workspace Updates announcement:
- Upload a selfie and a short voice recording from within your Google Account's
myaccount.google.com/video-verificationpage or directly inside Vids. The likeness is bound to your Google Account and restricted to the account holder. - Verify via the secure verification step Google built to confirm it's really you. This is the anti-impersonation control — the avatar cannot be made of a public figure or anyone other than the verified account owner.
- Type the script you want the avatar to deliver inside Vids. The avatar then generates the video delivery without any further filming.
- Direct with stage-direction syntax in square brackets. Typing
[excitedly]mid-script, for example, shifts the avatar's tone and energy, letting you set pace, emotion, and emphasis with natural-language direction rather than time-coded faces. This bracketed-emotion control for voiceover and AI avatars in Vids was announced in June 2026.
Eligibility and constraints you should know before banking on it
Google explicitly limits Personal Avatars at launch: the feature is available in English only, restricted to users 18 and older, and not available in the European Economic Area, Switzerland, or the United Kingdom as of the rapid-release rollout that started July 16, 2026 (Google Workspace Updates blog, July 16 2026). Workspace admins can disable or enable the feature at the domain level. Every AI-generated clip carries an invisible SynthID digital watermark developed by Google DeepMind so downstream viewers can verify a video was made with AI.
How Gemini Omni editing actually works
Beyond avatar generation, the model itself handles edits conversationally: you prompt "swap the background to a dark studio," or "fix the lighting on the subject's face," and Omni modifies the existing clip step by step. Crucially, you do not lose prior good edits — Omni revises one thing at a time, preserving the parts you wanted to keep. You can refine Omni-generated clips or clips you shot yourself; the chat-to-edit interface works on either.
Output is always watermarked. Every Gemini Omni and Personal Avatar clip SynthID-tags so AI provenance is verifiable downstream.
A real-world handicap worth naming
Personal Avatars do not produce all the B-roll a finished video needs. They handle the on-camera presenter line — the announcement, the walkthrough, the personalized shout-out — not the cutaway screenshots, music, or animated charts. Omni complements that by generating the visual layer around the avatar, but the suite still assumes you're building a Vids project with your own assets where helpful. If you've already worked through our broader comparison of free options, the 5 best free AI video generators in 2026 maps out where Vids/Omni sits relative to Meta AI Vibes, Arena Video Arena, Qwen, and Synthesia-style alternatives.
What changed when NotebookLM became Gemini Notebook?
On July 16, 2026 Google renamed NotebookLM to Gemini Notebook. It remains the same standalone research product, but the new name signals that its research features will surface in more places across the Google ecosystem — including directly inside the Gemini app and Google Search AI mode rather than living as an isolated app (blog.google, July 16 2026; Google Workspace Updates blog).
Scale of the tool
Google reported that more than 30 million people and over 600,000 organizations actively use NotebookLM/Gemini Notebook, using it for everything from business owners creating interactive onboarding materials to students generating audio and video summaries of class notes.
The bigger actual upgrade: secure cloud code execution
Renaming gets the press, but the real capability jump is secure cloud code execution, first announced June 8, 2026 and now expanding from AI Ultra to AI Pro users. Each notebook gets a secure cloud computer that lets Gemini Notebook write and execute code natively, enabling grounded data analysis on the documents you uploaded. You do this without exporting to Colab, importing a repo, or copy-pasting between tools.
Code execution in Gemini Notebook can:
- Aggregate and summarize numerical data from multiple imported documents.
- Format source data across different file types (CSV, PDF, plain text).
- Perform statistical processing on numbers already in your notebook.
- Generate graphs from your uploaded sources, output PDFs/Word/Excel/PowerPoint/Markdown/CSV/JSON.
If you maintain a library of coaching transcripts, internal policies, customer call notes, or research PDFs, you can now answer "What did we tell customers about lead generation last quarter?" with grounded citations derived from your own corpus rather than from re-reading each source.
Why this matters for the small business / builder
The integrated-Supremacy hypothesis: when the research assistant stops being a separate app and starts showing up inside Search and the Gemini app, the friction between "ask a question" and "ask grounded in my own stuff" disappears. Competitive NotebookLM coverage existed — but Google's bet is that being the default inside Search outweighs feature parity. If you organize research at scale, see our piece on Gemini Notebook Collections: how to organize your AI research without losing your mind for a maintenance model that scales past a few dozen notebooks.
How does the new Gmail "Help me write" refinement actually work?
The Gmail upgrade is the quietest of the four, but for users who draft dozens of emails a day it's the most immediately useful. Previously, Help me write offered fixed presets: Polish (smooth the prose), Formalize (raise the register), Elaborate (add detail), Shorten (cut it down). After this update, the side bar plus a new refine bar accepts free-text instructions like "make this warmer, keep the close, add a clear next step" — and you can keep iterating on a single draft instead of throwing the whole thing out when one element doesn't land (Google Workspace Updates blog, July 14 2026).
How to refine one email
- Open the reply draft in Gmail. With Gemini-enabled Workspace features turned on (admin: smart features and personalization), Help me write is selectable from the compose toolbar.
- Select Help me write to generate a baseline draft with full thread context.
- Type a follow-up refinement in the prompt bar. Examples per Google's published docs: "shorten it", "make this more direct", "add a clear next step at the end", "wait — make it warmer".
- Use Undo/Redo between edits. The bar keeps history of refinements within the session so you can back out a refinement and try a different one without losing the version you liked.
A practical example to anchor it
Say you reply to a customer who asked how your coaching calls work and what's included. Draft one is generated by Gemini. You type "make this warmer" — the tone shifts in place. You type "add a clear next step at the end" — a book-onto-the-next-call CTA appears. Two follow-up prompts produces a final email without you rewriting a single sentence. Compare this to the previous workflow: regenerate from scratch, lose what was good in the first draft, fight the model in your head.
Pricing/context: who gets it
Help me write in Gmail is available across Google Workspace Business Starter/Standard/Plus, Enterprise plans, Google One AI Premium consumer accounts, and (now) Workspace Individual subscribers since Google folded most Gemini Workspace features into base plans in January 2025.
Why did Google Gemini in Google Docs gain 11 new languages?
The Docs update is the most enterprise-relevant if your customers or team speak anything other than English. Effective July 15-16, 2026 rapid release, Gemini in Docs now supports eleven additional languages: Mandarin, Dutch, Malay, Hebrew, Polish, Turkish, Czech, Indonesian, Swedish, Danish, and Norwegian. These join the previously supported English, Spanish, Portuguese, Japanese, French, Korean, German, and Italian (Google Workspace Updates blog, July 16 2026).
That's not just a language count bump — it's a capability jump matched to a redesigned Gemini in Docs experience
Gemini in Docs is now centralized with a bottom-bar plus side-panel spark that supports four interrelated writing capabilities, per Google's own announcement:
- Help me create — generates fully formatted first drafts by synthesizing context across your Drive, Gmail, Chat, and the web. A blank doc stops being "blank".
- Help me write (refined) — prompt Gemini inline from the bottom bar or side panel to edit your doc or to focus on highlighted text. Suggestions stay private (only visible to you) until you accept.
- Match writing style — keeps a consistent tone and voice across a multi-author document.
- Match doc format — mirrors structural elements (headings, table columns) and typography (fonts, colors) from a source doc.
Why this combination matters for cross-border teams
If your customer support team writes English, Mandarin, Dutch, and Turkish docs from the same brand playbook, you can now standardize the tone and formatting across languages inside one workspace rather than doing tone-matching by hand. For builders who care about model routing per language, our comparison of Gemini Flash models — 3.5 Flash-Lite vs 3.6 Flash walks through latency/price/task routing decisions that apply equally to which language/workspace combos to optimize.
How are these four updates connected?
The unifying hypothesis behind the July 2026 batch is that the AI that wins is the one you never open as a separate app. Google didn't ship one flagship Gemini model announcement this round. It shipped four surface integrations, each proof-of-the-same-claim: omnichannel, embedded, multimodal Gemini making the boring parts of writing/video/research/email faster without forcing anyone to leave the tools they're already in.
That's why the avatar feature lives inside Vids (not as a separate recording studio product), why NotebookLM's new name exists so it can appear inside Search and the Gemini app, why Help me write's refinement bar lives inside Gmail's compose UI instead of a separate Gemini chat page, and why Docs Gemini now bundles itself into the doc editor rather than being a one-off sidebar.
For a fuller picture of how these tie back into Google's broader generative-Workspace pricing-and-feature story across the year — including what's already baked into basic Workspace tiers vs. paid add-ons — the resource we maintain on Google's Gemini models: Flash, Flash-Lite, and Flash-Cyber is the canonical reference at the model layer.
Who can actually use each update right now?
| Use case | Feature | Minimum plan | Region/age limits |
|---|---|---|---|
| "Generate a video update with my own face" | Gemini Omni + Personal Avatars | Google AI Pro / AI Ultra, Workspace Business or Enterprise | English only; 18+; excludes EEA, UK, Switzerland |
| "Generate a video clip from a text prompt/image edit" | Gemini Omni in Vids | Google AI Pro / AI Ultra, Workspace Business or Enterprise | English at launch |
"Direct my avatar's tone with [excitedly]" |
Vids avatar emotional control | Same as above | June 2026 release; broader rollout July 16, 2026 |
| "Run grounded code over my research notes" | Gemini Notebook secure code execution | Google AI Ultra → now expanding to AI Pro (with Workspace AI Expanded Access add-on also eligible) | Universal where NotebookLM is available |
| "Use Help me write with free-text refinement prompts" | Gmail prompt bar update | Workspace Business/Enterprise, Google One AI Premium, Workspace Individual | GA rollout since July 14, 2026 |
| "Run Help me create/write in Mandarin, Dutch, Polish, Turkish, etc." | Gemini in Docs 11-language expansion | Most Workspace Business/Enterprise + AI Pro/Ultra | Rapid Release July 15, 2026; Scheduled Release August 1, 2026 |
How should a small business or builder actually use these updates?
For most small businesses running Google Workspace, two of the four updates are the highest-leverage this quarter.
Start with Gmail's refinement bar — it's the zero-friction win. Every customer-service reply, every outbound pitch, every internal email can now be iteratively refined in seconds, and the workflow fits inside the compose toolbar you already use.
Then evaluate Personal Avatars in Vids if you publish any kind of recurring video: weekly updates, customer onboarding walkthroughs, internal announcements. The benefit — recording once and then generating unlimited iterations from a script — is meaningful for any team currently paying studio time or an outsourced editor.
Gemini Notebook matters most for any organization with a corpus of text it routinely re-references: meeting transcripts, SOPs, customer call logs, sales collateral. The cloud code execution layer turns those sources into a queryable data asset.
The Docs language expansion is operationally critical only if you actually publish in one of the eleven newly supported languages. For the rest of us, the side-benefit is that Help me write as a whole gets more capable in the same update, since Workspace Intelligence is writing across Drive/Gmail/Chat context.
What this means for you
If you already pay for Gemini in Workspace: all four updates are either already in your account or rolling out within the next two weeks (Rapid Release vs. Scheduled Release gate).
If you're on a free Google account only: Personal Avatars and Omni are gated behind Google AI Pro/Ultra; Gemini Notebook is the only one of the four broadly available without a Workspace plan.
If you administer Workspace: the Personal Avatar feature is ON by default as of the July 16, 2026 rollout — you'll want to communicate the consent implications of using your employees' face-and-voice likeness explicitly. Admins can disable it via the Admin Console.
If you operate in the EEA, UK, or Switzerland: you cannot use Personal Avatars at all yet. Gemini Omni's text-to-video editing is region-eligible, just not the likeness feature stacked on top.
FAQ
Q: What is Gemini Omni and how is it different from Veo?
A: Gemini Omni is Google DeepMind's mixed-input video generation and editing model that replaced Veo inside the Gemini app. It combines Gemini's core intelligence with image-to-video and video-to-video generation, lets you reference and combine multiple inputs (text, image, video, audio) into a single generated clip, and edits iterative via natural-language chat prompts.
Q: Can I use my own face as a personal avatar in Google Vids?
A: Yes — as of the July 16, 2026 rollout, you can capture your face and voice once through Google's secure verification process, then direct that avatar in any future generated video. The feature is restricted to the verified account holder, available in English only, limited to users 18 and older, and not available in the EEA, UK, or Switzerland at launch. Workspace admins can disable this at the domain level.
Q: Is NotebookLM gone, or just renamed?
A: NotebookLM is the same product, renamed to Gemini Notebook on July 16, 2026. All notebooks, sources, and data transferred automatically. The bigger functional upgrade is secure cloud-side code execution — a per-notebook sandbox that lets Gemini Notebook write and run code on your uploaded documents natively, previously AI Ultra only, now expanding to AI Pro.
Q: Can I edit AI-generated email drafts in Gmail without rewriting them?
A: Yes — Gmail's Help me write now accepts free-text refinement prompts in a dedicated prompt bar (rolled out July 14, 2026). You can ask it to "make it warmer", "add a clear next step", "shorten it", or any other tweak, and the draft updates in place. The bar supports undo/redo so you can backtrack refinements.
Q: Which languages does Gemini in Google Docs now support?
A: The July 16, 2026 expansion added Mandarin, Dutch, Malay, Hebrew, Polish, Turkish, Czech, Indonesian, Swedish, Danish, and Norwegian to the previously supported English, Spanish, Portuguese, Japanese, French, Korean, German, and Italian — 19 languages total, covering Help me create, Help me write, Match writing style, and Match doc format.
Q: Are the videos generated by Gemini Omni watermarked?
A: Yes — every AI-generated clip produced in Vids via Gemini Omni or Personal Avatars carries an invisible SynthID digital watermark developed by Google DeepMind, so the content's AI provenance is verifiable downstream.
Discussion
0 comments