Notes from lectures and long texts

📋 30-second summary

• Course material isn’t a single PDF: it’s a mix of the professor’s voice, slides, textbook, your own notes. The challenge isn’t summarizing one piece, it’s composing them.
• For the professor’s voice: automatic transcription (Whisper, Otter, Tactiq, Apple Voice Memos). For slides and handwritten notes: photos. For the textbook: a paragraph pasted in or pages photographed.
• The output you want isn’t a summary, it’s a study map that holds the pieces together and tells you where each thing comes from.
• When a course accumulates material over time, move everything into a Project (Claude Project, Custom GPT, Gemini Gem) so you don’t reload it every time.
• Recording a professor is legally delicate in many countries. Good practice: ask at the start of the course.

You have the material, but it’s everywhere. The professor’s voice is in an audio file on your phone, the slides are in a PDF downloaded from the portal, the chapter is in the paper textbook, your own notes are in a notebook. None of the three sources says exactly the same thing: the professor in class emphasizes points that the slides cut and the textbook expands, the slides are bare bones, the textbook is rich but generic.

The challenge when you sit down to study isn’t summarizing any one of them. It’s composing them: holding the pieces together, understanding where they talk about the same thing and where they diverge. In Summarize a long document you saw how to get a reliable summary from a single text. Here the problem is different: not a summary, a study map that works as your compass.

Case 1: the professor’s voice

Automatic audio transcription is one of the areas where AI is mature and free. Practical tools:

• OpenAI Whisper, free. You use it via desktop app, or if you have ChatGPT Plus you can upload the audio file directly into the chat.
• Otter.ai, Tactiq, Read.ai: integrated with Zoom, Google Meet, Microsoft Teams. If the lecture is online, they transcribe live and give you the text at the end of the call.
• Apple Voice Memos: on iPhone/iPad (recent models, iOS 18 and later) the voice recorder transcribes automatically and gives you a button to send the transcript to ChatGPT/Claude.

Once you have the text, you treat it like any other material: paste it or upload it into the chat with everything else. The transcript of an hour of lecture is typically 8000-12000 words. To give you a practical sense: ChatGPT, Claude and Gemini in their current versions handle without trouble a one-hour transcript plus the slides of the lecture plus 40-50 pages of textbook chapter, all in the same prompt. If you add a second hour of audio and another full chapter, you start getting close to the limit (the AI has more trouble connecting things that sit far apart in the prompt). For more recurring material, that’s when you reach for the Project in Case 3.

Recording consent and privacy

Recording a university lecture is legally delicate. In many jurisdictions, recording a conversation you’re part of for strictly personal use is generally allowed; broadcasting or uploading the recording is a different matter, and in some cases unlawful. Check your local rules if you’re unsure. Good practice: ask the professor at the start of the course if they’re fine with recording for personal study use. Most say yes; those who say no have their reasons. Don’t upload audio of your professor to public, shareable chats. For the bigger picture, see What you share when you use AI.

Audio transcription: what works, what doesn’t

Whisper transcribes standard English well, also with regional inflections. It struggles with thick dialects, strong accents, dirty audio from noisy rooms. If the professor speaks into a microphone (even a cheap lavalier), the quality is excellent. If you record on your phone from the back of the classroom, expect to correct some terms by hand. For proper names, foreign surnames, formulas, acronyms, always check: Whisper tends to “normalize” toward more common words that sound similar to it (the philosopher “Husserl” becomes “Usserl”, the author “Schmitt” becomes “Smith”). On a philosophy or law lecture with constant citations of foreign authors, plan on a manual pass to fix the names before using the text for study.

Case 2: composing multiple sources

Three inputs from the same course, one study session. The slides of the lecture, the textbook chapter, your handwritten notes. You take photos with your phone: keep the photo in focus, well lit, and one page per shot. AI is good with legible handwriting but makes mistakes on doctor-style scrawl, and in that case it tells you (“I can’t read these words clearly”) instead of guessing. On using photos in chat see Photos, images and files. The core prompt is this:

“I’m uploading three things: the slides of lecture 7 of Modern History, my handwritten notes from class (photos), and chapter 12 of the Sabbatucci/Vidotto textbook. I want a study map that holds the three together, highlighting where the professor in class emphasized things that the textbook treats differently or doesn’t treat at all. For each entry in the map, tell me which source it comes from, with tags [Prof], [Textbook], [Slides].”

What you’re asking the AI to do is different from summarizing. Not a compression of each source, but a unified structure of the topic, in which each entry makes clear where it comes from. It’s the difference between three separate summaries and a map with the threads that connect them.

The tags in square brackets do one precise thing: you look at the map and see at a glance where each source weighs in. An entry with all three tags [Prof] [Textbook] [Slides] is the heart of the topic, aligned across the sources. An entry with only [Prof] is what the professor underlined out loud and the others skip: probably something that will come up at the oral exam. An entry with only [Textbook] is detail from the book that the professor cut.

The natural output is a hierarchical outline of four or five sections with two or three sub-entries each. Below is an example.

Verify the tags before trusting them

The tags [Prof] [Textbook] [Slides] are written by the AI. They’re useful but they get things wrong: the AI can attribute to the textbook something that was only in the slides, or the other way around. The first quick check is on entries that surprise you. If you see a lone [Prof] on a point you remember reading in the textbook, open the chapter again: the AI probably misattributed. A more systematic check: take two or three lone [Prof] entries (the most important ones for the oral exam) and check in the audio transcript or in your notes whether the professor actually said them. You don’t need to verify every entry, but the ones you’d build your exam answer on, yes.

Case 3: organizing an entire course

When the material from a single course piles up over time (ten sets of slides, ten batches of notes, two textbooks, a few lecture recordings) reloading everything in every chat is a pain. The right tool is the Project: you’ll find it on Claude as “Projects”, on ChatGPT as “Custom GPTs”, on Gemini as “Gems”. We covered it in Reuse context with projects. Practical note for the student: Claude Projects and ChatGPT Custom GPTs require a paid plan (Plus or Pro). Gemini Gems are also available on the free tier with some context limits. If the budget is zero, work with single chats and reload the material each time: less convenient but workable.

The idea applied to study is simple. You create one project per course (“Modern History 2026”, “Private Law I”) and you load all the material in once. From that point on, every chat opened inside the project searches and links across the whole corpus without you having to reload anything. When lecture 11 comes in, you add it to the project and that’s it. When the exam approaches, you open a chat and ask questions that cut across the whole semester: “in which lectures did the professor talk about topic X?”, “in which chapters of the textbook does author Y appear?”.

An example: Italy 1943-1946

A modern history student, lecture from a course on “Italy 1943-1946”. She has three inputs ready: the audio transcript of the lecture (45 minutes, from Whisper), the professor’s slides (10 pages), and the textbook chapter (Sabbatucci/Vidotto, 40 pages). She opens the chat and asks for the study map.

👤

You

I'm uploading three things on the modern history lecture "Italy 1943-1946": the audio transcript of the lecture (45 min), the professor's slides (10 pages), chapter 4 of the Sabbatucci/Vidotto textbook. I want a study map that holds the three together, highlighting where the professor emphasized things that the textbook treats differently or doesn't treat. For each entry, add tags [Prof], [Textbook], [Slides] to indicate which source it comes from. I'm uploading three things on the modern history lecture "Italy 1943-1946": the audio transcript of the lecture (45 min), the professor's slides (10 pages), chapter 4 of the Sabbatucci/Vidotto textbook. I want a study map that holds the three together, highlighting where the professor emphasized things that the textbook treats differently or doesn't treat. For each entry, add tags [Prof], [Textbook], [Slides] to indicate which source it comes from.

🤖

AI

I've read the three materials. The map that comes out has five top-level sections. 1. The fall of Fascism (25 July 1943) - Grandi's motion to the Grand Council and the arrest of Mussolini. [Prof] [Textbook] [Slides] - Role of the monarchy and Vittorio Emanuele III in the decision: the professor in class insists at length on the King's premeditation, the textbook is more cautious and speaks of a "combination of circumstances". [Prof] [Textbook] - Popular reaction in the squares and the illusion of immediate peace. [Textbook] 2. The armistice and 8 September - Castellano-Smith negotiations in Algiers, Badoglio's announcement. [Prof] [Textbook] [Slides] - Flight of the King and the government to Brindisi, dissolution of the army. [Prof] [Textbook] - The point the professor in class insisted on most: 8 September as the "death of the homeland" in Galli della Loggia's reading, explicitly cited. This reading isn't in the textbook. [Prof] 3. The Italian Social Republic and German occupation - Mussolini freed at Gran Sasso, founding of the RSI at Salò. [Prof] [Textbook] [Slides] - Verona Manifesto of November 1943, socialization of firms. [Textbook] - Nazi massacres on Italian soil (Marzabotto, Sant'Anna di Stazzema): the textbook devotes a paragraph, the professor cited two specific testimonies. [Prof] [Textbook] 4. The Resistance - CLN and CLNAI (the National Liberation Committee), political composition of the partisan bands. [Prof] [Textbook] [Slides] - A distinction the professor returns to often: patriotic war, civil war, class war (Claudio Pavone's reading). The textbook mentions it in passing. [Prof] - Role of the Allies and the CVL, dissolution of the formations after the war. [Textbook] [Slides] 5. The turning point of 1946: the referendum and Constituent Assembly - Lieutenant decree and universal suffrage (for the first time including women). [Prof] [Textbook] [Slides] - Result of the referendum on 2 June 1946, geography of the North/South vote. [Prof] [Textbook] - Election of the Constituent Assembly and composition of the three mass parties (DC, PCI, PSIUP). [Textbook] [Slides] Three useful observations for study: - The entries with only [Prof] are the most important ones for the oral exam: they're the points the professor added out loud and that are missing from the other sources. In particular, Galli della Loggia's reading of 8 September and Pavone's reading of the Resistance. - The entries with only [Textbook] are detail the professor chose to cut in class. Useful for a written exam, less so for an oral. - The Verona Manifesto shows up only in the textbook. If the professor talked about it out loud and I missed it, it's worth re-listening to the audio around the minute when the discussion moves from the fall of Fascism to the RSI. I've read the three materials. The map that comes out has five top-level sections. 1. The fall of Fascism (25 July 1943) - Grandi's motion to the Grand Council and the arrest of Mussolini. [Prof] [Textbook] [Slides] - Role of the monarchy and Vittorio Emanuele III in the decision: the professor in class insists at length on the King's premeditation, the textbook is more cautious and speaks of a "combination of circumstances". [Prof] [Textbook] - Popular reaction in the squares and the illusion of immediate peace. [Textbook] 2. The armistice and 8 September - Castellano-Smith negotiations in Algiers, Badoglio's announcement. [Prof] [Textbook] [Slides] - Flight of the King and the government to Brindisi, dissolution of the army. [Prof] [Textbook] - The point the professor in class insisted on most: 8 September as the "death of the homeland" in Galli della Loggia's reading, explicitly cited. This reading isn't in the textbook. [Prof] 3. The Italian Social Republic and German occupation - Mussolini freed at Gran Sasso, founding of the RSI at Salò. [Prof] [Textbook] [Slides] - Verona Manifesto of November 1943, socialization of firms. [Textbook] - Nazi massacres on Italian soil (Marzabotto, Sant'Anna di Stazzema): the textbook devotes a paragraph, the professor cited two specific testimonies. [Prof] [Textbook] 4. The Resistance - CLN and CLNAI (the National Liberation Committee), political composition of the partisan bands. [Prof] [Textbook] [Slides] - A distinction the professor returns to often: patriotic war, civil war, class war (Claudio Pavone's reading). The textbook mentions it in passing. [Prof] - Role of the Allies and the CVL, dissolution of the formations after the war. [Textbook] [Slides] 5. The turning point of 1946: the referendum and Constituent Assembly - Lieutenant decree and universal suffrage (for the first time including women). [Prof] [Textbook] [Slides] - Result of the referendum on 2 June 1946, geography of the North/South vote. [Prof] [Textbook] - Election of the Constituent Assembly and composition of the three mass parties (DC, PCI, PSIUP). [Textbook] [Slides] Three useful observations for study: - The entries with only [Prof] are the most important ones for the oral exam: they're the points the professor added out loud and that are missing from the other sources. In particular, Galli della Loggia's reading of 8 September and Pavone's reading of the Resistance. - The entries with only [Textbook] are detail the professor chose to cut in class. Useful for a written exam, less so for an oral. - The Verona Manifesto shows up only in the textbook. If the professor talked about it out loud and I missed it, it's worth re-listening to the audio around the minute when the discussion moves from the fall of Fascism to the RSI.

Replay

Three things to notice. The map is short, five sections with two or three sub-entries each: an outline you can hang in your mind, not a full summary. The tags track the source, so you know without reopening the files where each piece comes from. The final observations are the added value of this step: they tell you what to do with the map, they don’t repeat what it contains.

A practical note on the audio: once you have the transcript, you can ask the AI for a targeted lookup inside it (“at what minute of the transcript does the professor talk about the Verona Manifesto?”) and it gives you the spot. Without the transcript, the same check would mean re-listening to 45 minutes.

What NOT to do

Don’t turn your own notes into something you no longer recognize. The study map has to help you remember what you saw in class and what you read: so it has to resemble your material, not replace it. If you end up with an outline that has nothing of your way of writing or organizing points, that’s a signal you’ve delegated too much. Put your own emphases back in.

Don’t use the audio transcript as the primary source if you have the textbook. The professor’s voice in class is low fidelity: they speak off the cuff, skip steps they take for granted, get dates wrong, make quick references. The textbook is high fidelity: it was edited, it’s precise, the dates are there, the citations too. The transcript is useful for the points where the professor adds something of their own that isn’t in the textbook (readings by other authors, examples, emphases) or to reconstruct the order in which they chose to handle the topics. For substance, the book wins. When the two sources say different things about a verifiable fact (a date, a citation, a legal article), take the textbook and add a margin note: “the professor in class said X, but the book says Y”. At the oral exam you read the room: some professors appreciate that the student noticed the discrepancy, others prefer the textbook version and that’s it.

Check what you’ve understood

📝 Check-in with your AI

Four questions to fix the method

5-7 min of check-in

Paste the text into your AI. It'll ask four progressive questions about the lesson, one at a time, with feedback.

Copy the text below and paste it into your favourite AI (Claude, ChatGPT, Gemini, or another). Or download the file and give it to your AI as an attachment.

Copy the text Download Open

What comes next

Now the course material is composed into a map that holds the pieces together. When exam week arrives you need something else: turn that map into something you can query, simulate the tests, fix the weak spots. The next lesson takes on this step.