Gemini Notebook voice mode is rolling out this week, and only Google AI Ultra subscribers get it first. Google announced the feature on 15 September 2026 as part of a study-focused update to Gemini Notebook, the research tool that was called NotebookLM until July. The company describes it plainly: you can “have a real-time conversation with your notebooks using our mobile app (Android/iOS) in nearly 100 languages”, with every answer grounded in the sources you uploaded.
That one sentence carries four separate restrictions, and most coverage has repeated only the headline. The conversation runs on mobile, not the web. The rollout starts with Ultra and reaches Google AI Pro and AI Plus only “soon”. The feature is limited to users aged 18 and over. And while you can speak to it in nearly 100 languages, the output language is English.
Below we set out exactly what Gemini Notebook voice mode does, who can use it today, which model powers it, and how it compares with the Audio Overview interactive mode that already existed in the product. We also cover the four other study features announced alongside it, the plan ladder that decides your usage limits, and the practical question at the end of all this: whether the feature is worth an Ultra subscription, or whether waiting costs you nothing.
Table of contents
- What Gemini Notebook Voice Mode Actually Does
- Who Gets Gemini Notebook Voice Mode First
- Nearly 100 Languages In, English Out
- The Model Behind Gemini Notebook Voice Mode
- The Other Four Study Features Announced With It
- What Each Plan Tier Buys
- Where Gemini Notebook Voice Mode Falls Short
- How to Decide Whether Gemini Notebook Voice Mode Is Worth Ultra
- Frequently Asked Questions
- References
What Gemini Notebook Voice Mode Actually Does
Google’s framing is deliberately narrow. This is not a general assistant you talk to. It is a spoken interface onto one notebook and the sources inside it.
Grounded in your own sources, not the open web
The defining property is grounding. Gemini Notebook voice mode answers from the documents, slides, PDFs, videos and pasted notes you added to that notebook, and cites them, exactly as the typed chat does. Ask it something your sources do not cover and the honest answer is that it does not know. That constraint is the whole point of the product, and it is what separates Gemini Notebook voice mode from Gemini Live.
You can interrupt it mid-sentence
Google specifically says the conversation supports interruption and step-by-step guidance. In practice that means you can cut in when the explanation goes the wrong way, ask it to slow down, or redirect it to a different source without restarting the exchange. Full-duplex behaviour of this kind is the current bar for voice products, and it is the feature that makes Gemini Notebook voice mode feel like a conversation rather than like dictating into a form.
Mobile only, at least for now
Gemini Notebook voice mode runs in the Gemini Notebook mobile app on Android and iOS. There is no web equivalent in this release. That is a meaningful constraint for the audience Google is aiming at, because the same announcement positions the tool around lectures, revision and study sessions — activity that often happens on a laptop with the notebook open on screen.
Where Gemini Notebook voice mode sits next to Audio Overviews
Gemini Notebook already had a spoken feature. Audio Overviews generate a podcast-style discussion between two AI hosts, and interactive mode lets you tap Join to interrupt those hosts with a spoken question. That is a different shape of interaction: you are joining a pre-generated programme. Gemini Notebook voice mode is a direct conversation with the notebook itself, with no hosts and no generated show to interrupt.
| Aspect | Audio Overview interactive mode | Gemini Notebook voice mode |
|---|---|---|
| Shape of the interaction | Join a generated two-host discussion | Direct conversation with the notebook |
| Where it runs | Web and mobile | Mobile app only (Android, iOS) |
| Spoken input languages | English only | Nearly 100 |
| Spoken output language | English | English |
| Who can use it | All plans, subject to daily caps | Google AI Ultra first, 18+ |
| Works on a shared notebook link | No | Not stated |
Who Gets Gemini Notebook Voice Mode First
The tier gate is the most consequential detail in the announcement, and it is the part a headline cannot carry.
Ultra this week, Pro and Plus soon
Google AI Ultra subscribers get Gemini Notebook voice mode this week. Google says the feature is “coming soon” to Google AI Pro and Google AI Plus, without naming a date. Nothing in the announcement promises Gemini Notebook voice mode to the free tier at all, which is a notable omission given that Gemini Notebook itself became free for all users earlier in September.
The 18+ restriction is a hard gate, not a default
Access to Gemini Notebook voice mode requires an account holder aged 18 or over. This is the same restriction Google applied to Cinematic Video Overviews earlier in 2026, and it sits awkwardly against the announcement’s framing, which is aimed squarely at students. A sixth-form or first-year undergraduate under 18 is exactly the user the update describes and exactly the user who cannot have this particular feature.
The audio recorder is the opposite case
Worth contrasting: the audio recorder announced in the same post is supported on web and mobile, for all users including those under 18. So Google is not applying the Gemini Notebook voice mode age policy across the whole update. The gate is specific to the live spoken conversation, which suggests the reasoning is about real-time generative audio rather than about the notebook product generally.
What this means if you are not on Ultra
If you are on Pro or Plus, the practical advice is to wait rather than upgrade. Google has stated the feature is coming to both, and every recent Gemini Notebook capability that started on Ultra — code execution, cinematic video, short video overviews — reached the lower tiers within weeks. Upgrading to Ultra purely for Gemini Notebook voice mode buys you a head start, not exclusive access.
Nearly 100 Languages In, English Out
The language figures in this launch look generous until you separate input from output, and the split is where the practical limit sits.
The asymmetry, in one chart
Google says Gemini Notebook voice mode works “in nearly 100 languages”. The output language, in the same breath, is English. The short video overviews announced alongside it run to 80+ languages. The underlying model handles 97. The number that actually governs what you hear back is one.
Who this actually serves
A student in Jakarta or Sao Paulo can ask Gemini Notebook voice mode a question in their own language and will be answered in English. That is genuinely useful if you are studying English-language source material, which describes a large share of university reading lists worldwide. It is much less useful if the notebook holds Portuguese or Indonesian sources and you wanted a spoken explanation in the same language.
Expect the output list to grow
Google has expanded output languages steadily on every other audio feature in this product. Audio Overviews went from English-only to dozens of languages; short video overviews added 70+ languages and three English variants at the start of September 2026. The English-only output on Gemini Notebook voice mode reads as a launch constraint rather than a design decision, but Google has not committed to a timeline.
The Model Behind Gemini Notebook Voice Mode
Google shipped a new speech model on exactly the same day, and it is almost certainly what this runs on.
Gemini 3.8 Live, announced 15 September 2026
Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking were announced on 15 September 2026, the same day as the Gemini Notebook study update. The model behind Gemini Notebook voice mode is built for natural voice conversation with background processing, so it can execute a task while the conversation continues. Google is shipping it into Search Live, the Gemini Live app, and Workspace apps including Docs, Gmail and Keep, as well as the Gemini API and AI Studio. See our coverage of the Gemini 3.8 Live rollout for the wider picture.
The published benchmark numbers
Google quotes four figures for the model. They are worth reading as a set, because the spread between them is large and each measures something different.
| Benchmark | Score | What it measures |
|---|---|---|
| Artificial Analysis Speech to Speech Quality Index | 82.6 (ranked first) | Overall spoken conversation quality |
| Big Bench Audio | 97.7% | Reasoning over spoken input |
| Tau-Voice | 68.6% | Completing agentic tasks by voice |
| Sierra Tau-Voice-banking | 35.1% | Agentic tasks in a banking scenario |
Reading the 97.7% against the 35.1%
The gap between those two scores is the honest picture of where voice agents are. Answering a reasoning question from spoken input is close to solved at 97.7%. Completing a multi-step task in a realistic banking workflow sits at 35.1% — roughly one in three. Gemini Notebook voice mode lives at the easier end of that range, because explaining a source you already supplied is a reasoning task, not an agentic one.
Google has not confirmed the model pairing
One caution: Google’s announcement does not name the model powering Gemini Notebook voice mode. The same-day timing, the interruption support and the 97-language figure all point to Gemini 3.8 Live, but that is an inference from two announcements rather than a stated fact. Treat the pairing as very likely rather than confirmed.
The Other Four Study Features Announced With It
Gemini Notebook voice mode was the headline of a five-part update, and two of the other four are not available yet either.
Audio recording in the mobile app
Starting the week after the announcement, the mobile app gains an audio recorder for lectures and spoken thoughts. Recordings sit alongside your other sources, so you can cite them, question them and edit around them. This is available on web and mobile to all users, including under-18s, with English as the output language — a materially wider rollout than Gemini Notebook voice mode itself.
Interactive Learning Overviews
Coming “in the next few weeks”, Interactive Learning Overviews build on the existing Reports feature, combining a written summary with infographics, quizzes and flashcards drawn from the same source collection. TestingCatalog had spotted an early version of this in testing on 14 September 2026, describing a report container with placeholders that readers generate on demand by tapping Add.
Expanded quiz formats
Quizzes gain short answer, multiple select and fill-in-the-blank question types, on top of the multiple choice that already existed. You can also add or edit questions by hand, and ask the assistant about your own performance. This is the least glamorous item in the update and probably the most immediately useful one for revision.
Short Video Overviews
Roughly 60-second videos combining narrative overviews with educational animations, in 80+ languages. These are aimed at concepts that are hard to read and easy to watch — a formula being rearranged, a process with stages. Short Video Overviews launched earlier in 2026 and reached 70+ additional languages and three English variants at the start of September.
| Feature | When | Who | Where |
|---|---|---|---|
| Gemini Notebook voice mode | This week | Google AI Ultra, 18+ | Mobile app |
| Audio recording | Next week | All users, incl. under 18 | Web and mobile |
| Interactive Learning Overviews | Next few weeks | Not stated | Reports |
| Expanded quiz formats | Announced | Not stated | Quizzes |
| Short Video Overviews | Available | Plan-dependent caps | Web and mobile |
What Each Plan Tier Buys
Since Gemini Notebook voice mode is gated on Ultra, the plan ladder that decides who can reach it is part of the story rather than a footnote.
The published limits, with one large caveat
Google’s help documentation sets out per-plan limits for notebooks, sources, daily chats and generated artefacts. The caveat is significant: Gemini Notebook moved to compute-based usage limits on 2 September 2026, dropping fixed daily caps in favour of a pooled allowance. The table below is the published ladder and is still the clearest guide to relative generosity between tiers, but treat the individual daily numbers as indicative rather than literal.
| Limit | Standard | AI Plus | AI Pro | AI Ultra 20TB | AI Ultra 30TB |
|---|---|---|---|---|---|
| Notebooks | 100 | 200 | 500 | 500 | 500 |
| Sources per notebook | 50 | 100 | 300 | 500 | 600 |
| Chats per day | 50 | 200 | 500 | 2,500 | 5,000 |
| Audio Overviews per day | 3 | 6 | 20 | 100 | 200 |
| Cinematic videos per day | — | — | 2 | 10 | 20 |
| Reports, quizzes, flashcards per day | 10 | 20 | 100 | 500 | 1,000 |
The ratio that matters
Read down the chat row and the shape of the ladder is obvious. Each tier is roughly an order of magnitude apart until Ultra, where the jump is five-fold again. This is the commercial logic behind putting Gemini Notebook voice mode on Ultra first: the tier is already priced for heavy generative use, and a live spoken conversation is the most compute-hungry thing in the product.
The student offers change the arithmetic
Alongside the update, Google put a year of paid access in front of students. In the United States, college students get one year of Google AI Pro free, a plan Google values at $19.99 per month, with 4x higher usage limits. In 140+ other markets, students get one year of Google AI Plus free, valued at $4.99 per month, with 2x higher limits.
| Offer | Plan | Stated value | Usage uplift | Deadline |
|---|---|---|---|---|
| US college students | Google AI Pro, 1 year | $19.99 per month | 4x | 31 December 2026 |
| 140+ other markets | Google AI Plus, 1 year | $4.99 per month | 2x | 31 December 2026 |
The markets Google excluded
The international offer excludes the United States, Bolivia, Albania, Canada, Macau, Hong Kong and Tunisia. The United States exclusion is simply because US students get the better Pro offer instead. The others are not explained in the announcement. Note that neither student offer includes Ultra, so no student promotion currently unlocks Gemini Notebook voice mode.
Where Gemini Notebook Voice Mode Falls Short
Four limitations are worth stating plainly before anyone changes a subscription over this.
Gemini Live still cannot reach your notebooks
The most persistent complaint about Google’s voice products is that Gemini Live — the general assistant voice mode in the main Gemini app — cannot access Gems or Notebooks. That has not changed. Gemini Notebook voice mode lives in a separate app, talking to one notebook at a time. If you wanted a single voice assistant with access to your research, this is not it.
Mobile-only splits the workflow
Because the conversation runs only in the mobile app, the natural study setup is awkward: sources and typed chat on the laptop, spoken questions on the phone. Google has given no indication of when, or whether, the web app gets the same capability.
Grounding is a guarantee about scope, not accuracy
Source grounding reduces the chance of a fabricated answer, because the model is answering from documents you supplied. It does not eliminate misreading, over-confident summarising, or a citation pointing at the wrong passage. Spoken answers make this harder to audit than written ones, because there is no text on screen to scan against the source. Check anything that matters against the citation.
The interactive audio precedent is not encouraging
Google’s own documentation for Audio Overview interactive mode warns of “audio glitches like random speaker switches, third voice, or voice glitches”, and restricts interaction to newly generated overviews and to notebooks you own rather than shared links. Real-time generated audio in this product has a track record of rough edges, and Gemini Notebook voice mode is built on the same foundation. Expect Gemini Notebook voice mode to arrive with some of its own.
How to Decide Whether Gemini Notebook Voice Mode Is Worth Ultra
A short decision framework, because the honest answer for most readers is to wait.
Upgrade now only if three things are true
Upgrade to Ultra for Gemini Notebook voice mode only if you are already close to your current plan’s limits, you study or work primarily on a phone or tablet, and English output suits your sources. If any one of those is false, the head start is not worth the price difference.
Wait if you are on Pro or Plus
Google has said Gemini Notebook voice mode is coming to both paid tiers. Every comparable Gemini Notebook feature has followed that path within weeks, including the code execution capability that launched Ultra-first in July 2026 and the short video overviews that went global at the start of September.
If you are a student, take the free year first
US college students should redeem the free year of Google AI Pro, and students in the 140+ eligible markets should redeem AI Plus, before 31 December 2026. Neither unlocks Gemini Notebook voice mode today, but both raise your usage limits immediately and both put you on a paid tier that Google has said will receive the feature.
What to watch next
Three signals will tell you the rollout is maturing: an output-language expansion beyond English, a web version of the conversation, and an announcement that Pro and Plus have it. Until at least two of those land, Gemini Notebook voice mode is best understood as an Ultra preview with a confident name. For context on how quickly Google is moving across the rest of the line-up, see our write-up of the Gemini desktop app for Windows.
Frequently Asked Questions
Is Gemini Notebook voice mode available on the web?
No. Gemini Notebook voice mode is limited to the Gemini Notebook mobile app on Android and iOS. The audio recorder announced in the same update does work on web and mobile.
Which languages can I speak to it in?
Google says Gemini Notebook voice mode accepts nearly 100 languages. The spoken output is English. The underlying Gemini 3.8 Live model supports 97 languages and can switch between them mid-conversation.
Do I need Google AI Ultra?
Today, yes. Google says Gemini Notebook voice mode is rolling out to Google AI Ultra subscribers this week and is coming soon to Google AI Pro and Google AI Plus. No free-tier availability has been announced.
Is there an age restriction?
Yes. Gemini Notebook voice mode requires the account holder to be 18 or over. The audio recording feature in the same update has no such restriction.
How is this different from an Audio Overview?
An Audio Overview is a generated discussion between two AI hosts that you can join with a spoken question. Gemini Notebook voice mode is a direct, two-way conversation with the notebook, grounded in your sources, with no hosts involved.
Will it answer from the open web?
No. Gemini Notebook voice mode answers only from the sources you added to that notebook, which is the core design constraint of the product and the reason its citations are checkable. For a wider view of how the assistant market is splitting across subscription tiers, see our coverage of Meta’s new subscription plans.
References
Sharpen your study routine with new Gemini Notebook tools
Introducing Gemini 3.8 Live and 3.8 Live Extended Thinking
Build real-time voice applications with Gemini 3.8 Live and 3.5 Transcribe
Generate Audio Overview in Gemini Notebook
NotebookLM is now Gemini Notebook
Early look at Interactive Reports on Gemini Notebook
Gemini Notebook switching to compute-based usage limits, like Gemini app
More AI coverage: explore Progressive Robot's AI Models, Tools & Releases hub — hands-on reviews, setup guides and benchmarks in one place.