Master the NEW Gemini Notebook

**TL;DR:** Gemini Notebook (formerly Notebook LM) has evolved into a full research and learning platform: ground AI in your own sources, analyze gaps, and turn everything into podcasts, videos, slides, quizzes, infographics, and more. ## What Is Gemini Notebook? The core reason to use Gemini Notebook hasn't changed. It lets you ask an AI to dive into a specific topic without worrying about hallucinations, utilize information across different file types, and transform that information into interesting and useful formats. No other tool does it quite like Gemini Notebook. To get started, log in with your existing Google account at notebook.google.com. All your notebooks are listed on the main screen. When you open a notebook, you see three panels: - **Left panel — Knowledge base:** Your sources. Gemini is grounded in these, so all answers pull only from them, which cuts hallucinations. - **Middle panel — Chat:** A familiar chat interface that only pulls from your sources, with inline citations so you can verify exactly where information came from. - **Right panel — Studio:** Transform your data and conversations into infographics, podcasts, short-form videos, slide decks, reports, and more. ## Building a New Notebook from Scratch Create a new notebook and start by adding sources. Click **Add sources** and upload files you already have. Gemini Notebook supports a wide variety of file types across text, image, audio, and video. In the video, the author builds a notebook about the environmental impact of AI. He uploads three PDFs, then adds websites by pasting links (press space to add another), YouTube videos, and more websites. Sometimes a source will turn red, meaning it failed to load. In the video, three failed sources turned out to be downloadable PDFs. If that's not possible, you can copy the full text and paste it using the text option. ### Let Gemini Search the Web If you don't already have enough sources, you can let Gemini search for you. Type a query like "environmental impact of AI" and let it search. Gemini returns a list of sources. The author recommends opening each one and verifying it comes from a reputable website. You can deselect any you don't want — he removed Forbes and kept the rest. ### Finding Gaps and New Perspectives with Prompts For nuanced topics with many differing perspectives, the author runs a few prompts to analyze his sources and fill gaps. This is done in the chat panel. **Prompt 1:** > Are there any contrarian, alternative, or lesser known viewpoints on AI's environmental impact that are not represented in these sources? Describe those viewpoints at a high level and find reputable sources that represent those perspectives. This is a new capability. Normally Gemini Notebook works only from existing sources, but a prompt like this lets it go out and find new sources. Once you ask a typical question again, it switches back to its more confined, source-citing behavior. The response listed perspectives already covered, plus missing ones like EAC data centers as grid stabilizing assets and radical degrowth. The author then asked: > Find reputable sources that represent those three viewpoints. After a minute or two, Gemini returned a research report and seven new sources to add. **Prompt 2:** > Based on these sources, what important questions or subtopics about AI's environmental impact are missing or barely covered? List the biggest gaps that would need to be filled to really understand this topic well. This lays out the unknowns and missing data in your current source set. **Prompt 3:** > Looking at all the sources in this notebook, identify any areas where the sources disagree with each other and any clear contradictions or conflicting claims. After the first prompt, there are often many disagreements to surface. One surprising finding from the example notebook: > "Utilizing AI systems for writing and illustration resulted in 130 to 2,900 times lower carbon emissions than employing human writers and illustrators." The author notes it's a weird one to think about, but plausible. These prompts can also surface bad sources you accidentally included. ### Saving and Cleaning Up If you want to preserve a useful chat response: 1. Click **Save to note** in the response. 2. It appears in the Studio panel. 3. Click **Convert to source** — now Gemini understands the nuance and conflicts in your sources as it responds. You can also delete chat history via the three-dot menu if you've saved everything important. This can't be undone, so make sure any important responses are saved first. ## Customizing Your Notebook Under the **Customized notebook** settings, you can: - Change the title — the author prefers concise, literal titles for easy scanning. - Replace the autogenerated summary with a simpler custom one (often just the title again). - Add a background image if desired. - Configure chat style, including a custom option to guide the conversation direction. - Set response length — the default is usually long, but switching to shorter is common. Another useful source feature is **Auto label sources by topic**. It sorts your sources into different tabs with labels, and a source can appear in multiple tabs. You can also manually reorganize them. ### Google Drive Sources You can also use sources from Google Drive — documents, spreadsheets, slides, etc. Docs and sheets act as living documents: if you update the file in Drive, it updates in your notebook too. ## Using the Chat Panel The chat panel works like any AI chat, but with some important differences: - **Inline citations** appear throughout responses. Hovering shows the source, and clicking lets you read surrounding context on the source panel. - The prompt bar shows how many sources are being used. - By default, answers come from all your sources. You can manually deselect sources to narrow the focus. - Auto-labeling makes it easy to select or deselect entire topic groups. For example, by deselecting some sources, the author narrowed a question to 14 sources, leaving out philosophy and policy-focused papers. ## Exploring the Studio Panel The Studio panel makes Gemini Notebook stand out. You can transform sources and research into many formats: audio overview, slide deck, video overview, mind map, reports, flashcards, quiz, infographic, and data table. ### Audio Overview This is the podcast feature. The author mostly uses the **Deep Dive** option; the Critique and Debate options sound interesting but he hasn't found them very useful. You can choose language, length, select/deselect sources, and guide the conversation. The deep dive is great for reviewing information while driving or doing dishes on the mobile app. All generations sync there. If you click **Interactive**, you can jump into the conversation yourself. The author notes it's a unique feature — you can comment, ask a question, or steer the conversation — though he usually just listens. ### Slide Deck There are two options: - **Detailed:** Includes all relevant information read directly on the slides. - **Presenter:** Meant to have a speaker adding context and explaining things. You can describe the deck you want, and that guides both content and style. If the result needs changes, use the **Revise** button. You can click around to different slides, add multiple edits at once, send it, and the whole deck regenerates with those edits. ### Video Overview Three options are available: **Cinematic**, **Explainer**, and **Short**. - **Explainer** is basically slides with a generated voice speaking over them. - **Cinematic** is more engaging — it mixes voiceover with data visualizations. However, the author rarely uses it because it takes a long time to generate and isn't good enough all the way through. - **Short** is a new option: bite-sized overviews designed to feel like educational doom scrolling. A sample "Short" video about transformers and self-attention: > "How does a modern neural network actually understand the text you type? ... The strongest mathematical relationship, the link with cat, gets the highest score. By looking across the sequence for context, the network performs self attention." The author notes some Shorts turn out really well. ### Mind Maps Mind maps provide a high-level overview of your topic and subtopics. The most useful part: if you notice a subtopic you hadn't considered, click it to jump right into the chat. It autofills a prompt and sends it for you. ### Reports Reports open with a few generic presets plus four customized suggestions generated from your specific sources. Usually the customized suggestions are much better than the defaults. The **Conceptual Primer** option, for example, generates a full written report on the topic. ### Flashcards Select the number of cards, difficulty level, and topic. Then generate and test your knowledge. The example included: - "Process of breaking raw input text into units like words, subwords or characters is called tokenization." - "Primary role of the embedding layer: creating the initial vectors." If you want to dive deeper, hit **Explain** and it opens a chat about that card. ### Quiz Same options as flashcards. Generates a solid quiz to test knowledge of whatever is in your notebook. ### Data Table If any information in your sources would be useful as a spreadsheet, this option builds it. You can export directly to Google Sheets. ### Infographic Infographics are one of the more unique options. You can choose orientation, visual style, and level of detail. - **Concise** and **Standard** work great. - **Detailed** often tries to cram too much in and can be full of errors. The author showed examples of errors from a Detailed infographic: "This should be quick. This one brown," garbled text like "inject the injection or calculate chunk prevent 10 scening," and sideways labels that should have read "translate synthesize." ### Fixing Infographic Errors with ChatGPT The easiest fix is to drop the infographic into ChatGPT and ask it to fix all the errors. ChatGPT doesn't just fix misspellings — it analyzes logic and grammar, catches wrong equations, and can even correct layout issues. In the example, ChatGPT: - Fixed "brown" and "quick." - Rewrote a nonsense sentence to "calculate cross entropy loss at each position in parallel to update the model." - Removed redundant sideways labels and replaced them with a proper axis label like "Model Performance." - Reorganized the layout in a way that made more sense. The author notes that GPT Image 2 is just better at this level of detail and usually comes back with every error fixed. ## Generating Other Formats via Chat The Studio panel doesn't list every format. If you need a Word doc, Markdown, PowerPoint, JSON, or PNG, you can ask for it directly in chat. It will appear in the Studio panel, but it won't be viewable inside Notebook — you'll need to click to download it first. You can also ask the chat to generate any of the standard Studio options. This is useful when you have relevant information in the chat and don't want to manually copy it into the customization options. ## Gemini Notebook + Gemini App Integration One of Gemini Notebook's greatest strengths — accuracy — can also become a weakness when you want creativity. That's where Gemini itself shines. All your notebooks show up in the Gemini app. You can open them, and your chats work from the same knowledge base you built in Gemini Notebook. This allows: - Multiple separate conversations against the same notebook, keeping them organized. - More creative conversations than the grounded, source-only mode in Notebook. Those Gemini conversations then show up in Notebook under your sources. You may want to deselect them depending on what you discussed, but the integration is seamless. You get the accuracy of Gemini Notebook and the creativity of Gemini. ## Final Thoughts The new Gemini Notebook is an incredibly useful tool for research, learning, and much more. Whether you're building a deep research base, generating study materials, or creating polished deliverables, the full platform now covers the entire workflow. Source: [Master the NEW Gemini Notebook - Futurepedia](https://www.youtube.com/watch?v=TGlPNGaHo6I)