Using AI to Summarize Music Lesson Transcripts: A Practice Lab Experiment

This article is part of the Practice Lab series where I experiment with AI tools in the practice room. These articles are working experiments to provide insight into uses for AI in music practice and learning. In this article, I share how I tested multiple AI tools to summarize transcripts of my violin lessons.

I have been recording and transcribing my violin and viola lessons for several months, and during this time, have experimented with multiple AI tools to summarize and outline the major topics in my lessons to help me to plan my practice and keep a record of my learning and progress. When I first started recording and transcribing my lessons, I needed a way to synthesize all the information so I turned to AI to help. I use the record and transcribe function within Apple Notes and the transcriptions are often rather messy. But the AI tools did a good job turning these transcripts into actionable summaries with all the important details such as technical corrections, musical ideas, repertoire notes and assignments clearly called out. These summaries fit well in my existing lessons notes structure where I could then layer on my practice priorities for the week and any additional notes I wanted to capture on my repertoire or other assignments.

In this experiment, I tested multiple AI tools to summarize my lesson transcripts.

The tools I tested: ChatGPT, Claude, and NotebookLM/Gemini

Each of the tools I tested has slightly different strengths, while ChatGPT is great for custom prompts and refining the output, Claude, with its longer context window, is excellent at reading long documents like transcripts and writing clear, organized summaries. NotebookLM, which I’ve written about in a previous practice lab article, is best when transcripts and summaries are part of a larger archive where it can be used to connect trends over time and find reoccurring themes across lessons.

The experiment

Over the course of multiple months, I uploaded transcripts to each of the tool, using a similar prompt to ask them to summarize my lessons in 3-4 paragraphs and to provide a bullet-point list of the key topics covered over the course of the lesson. I evaluated each tool’s summaries for accuracy, comprehensiveness, and clarity.

Note on privacy: I received my teacher’s consent to record and transcribe my lessons and she is fully aware of my use of AI to summarize lesson transcripts. I recommend that anyone looking to record, transcribe, and summarize their own lessons this way do the same.

Tool	Best For	Biggest Strength	Biggest Limitation
ChatGPT	Interactive practice planning	Flexible follow-up and custom prompts	Needs careful prompting
Claude	Accurate summaries	Clear, readable synthesis	May need follow-up to become practice-specific
NotebookLM/Gemini	Lesson archives	Connecting lessons over time	Best with organized source material

What worked

AI tools can turn a long transcript into a usable lesson summary

All three tools were able to create lesson summaries that distilled the main points of the lesson including technical focus areas, and nuanced work on repertoire. Because lesson transcripts are long, and Apple Notes transcription is not the most accurate, AI tools were able to make the content of the lesson immediately accessible. The summaries made it possible to easily identify practice priorities based on the lesson.

ChatGPT worked well for interactive practice planning

ChatGPT worked best for practice planning based on lesson transcripts and summaries. It is good for asking follow-up questions, generating practice checklists, or reformatting summaries into assignments or detailed notes.

Example prompts:

• Turn this summary into a 45-minute practice plan.
• Create a checklist of what I should practice before my next lesson.
• Identify the three most important things to focus on this week in my practice.

Claude was useful for clear, readable summaries

Claude is excellent at producing well-organized summaries from long, messy transcripts. I found Claude’s summaries to suit my own needs the best, as they usually required the least editing for accuracy while being written in a clear, neutral voice.

NotebookLM/Gemini worked well as a lesson archive and to identify trends over time

I’ve written about my experiment with NotebookLM where I uploaded all my lesson summaries and transcripts into a single notebook within the tool. NotebookLM was great for organizing all my lessons in one place and producing insights across lessons. While it can summarize individual transcripts, it was most useful to identify patterns across multiple weeks or months.

Example Prompts For Cross-Lesson Insights:

• What technical issues do I encounter across lessons?
• What kinds of practice strategies does my teacher emphasize across lessons?
• Based on the most recent lesson transcript, what are the three most important things to practice this week?

What didn’t work

Transcript accuracy matters

During the time I have been transcribing my lessons, I have noticed that music terminology is occasionally transcribed incorrectly and the AI can compound this. I have found that the AI occasionally attaches teacher comments to the wrong movement or section of a movement or is unable to properly identify the section of the piece worked on in the lesson. It is possible to alleviate some of these issues by providing additional context with the prompt, clarifying repertoire or lesson details upfront.

AI may misunderstand musical context

The transcript captures the words exchanged during the lesson but not the musical examples, of which there are numerous over the course of a music lesson. Because of this, AI doesn’t have the full information it needs to write complete summaries. I have found that the summaries are surprisingly accurate given the tools are missing a large amount of the musical content from a lesson. When it is clear the AI is missing context, it is relatively easy to provide feedback to the AI to get a more accurate revision of the summary.

Generic prompts produce generic summaries

While I generally want a relatively straight forward summary of my lesson, if you want specific details such as practice assignments, it is necessary to build this into the prompt. I have found it useful to ask the tool to use accurate musical terminology, and you may want to ask it to preserve the teacher’s language. Some tools are more likely than others to get creative which results in less accurate summaries.

Improvements

While all the major AI tools do a reasonable job of summarizing transcripts, customizing prompts and working from the cleanest possible transcript leads to the best output. Working the output into a practice journaling or planning system is what will actually make these records of lessons useful to you as a musician.

Suggested improvements:

• Use consistent file names: YYYY-MM-DD Teacher Lesson – Piece – Instrument. This gives the AI additional context and keeps your notes organized.
• Add metadata at the top of each transcript:
  • Date
  • Teacher
  • Repertoire
  • Focus Area
  • Instrument (if useful)
• Use a standard prompt template for each lesson.
• Create a dedicated project or reusable skill in your AI tool to quickly generate lesson summaries.
• Save the final summary somewhere where you can find it later: Notion, Obsidian, Google Docs, NotebookLM, or a practice journal.
• Create a weekly practice plan from the summary immediately after the lesson.
• Compare the AI summary with your own memory while the lesson is still fresh, and make corrections.

Lesson Transcript Summary Prompt

I customized my prompt for this experiment so that it would fit seamlessly into my lesson notes format. The below prompt was the most successful at getting accurate summaries of my lessons that helped me remember my lesson and plan my practice for the week.

Feel free to use and modify the following prompt to summarize your own lesson transcripts:

Summarize this violin lesson transcript following my guidelines. Write 3–4 paragraphs using appropriate violin and music terminology, then provide 5–7 bullet points listing the main topics covered.

Summary Format
Paragraphs (3–4 total)
Structure the paragraphs as follows:
1. Opening / Lesson Overview — Briefly describe the focus of the lesson. Mention the repertoire, movements, or sections covered and any overarching theme (e.g., navigating chromatic passages, bow technique in high-intensity music, clarifying fingering and position work).
2. Technical Work — Detail the specific technical skills addressed. Use precise terminology (see vocabulary list below). Describe what was practiced, what challenges came up, and what corrections or exercises the teacher gave — including the reasoning behind them where possible.
3. Musical / Interpretive Work — Describe any discussion of phrasing, dynamics, articulation, style, or expression. Note musical concepts introduced (e.g., rhythmic pulse in Baroque-influenced writing, phrase-end breathing, bow drops for open tone, de-emphasizing inner voices in chords).
4. Wrap-Up / Practice Assignments (optional but encouraged) — Summarize any specific assignments, practice strategies, or goals the teacher gave for the week ahead.
Bullet Point List (5–7 items)
After the paragraphs, include a concise bullet list titled “Main Topics”. Each bullet should be a short, clear phrase (not a full sentence) naming a specific topic addressed in the lesson. Examples:
	• Shifting strategy in the chromatic passage in the development section
	• Intonation vigilance with open-string ring tones as reference
	• De-emphasizing lower voices in broken chords

Tone and Style
• Write in first person (e.g., “I worked on…”, “My teacher pointed out…”, “The lesson focused on…”).
• Use formal but readable language like a well-written lesson journal, not a transcript or casual recap.
• Capture key corrections or cues verbatim when the teacher’s exact wording matters (e.g., “think of the forte as a comfortable, beautiful sound rather than showing off how loud you can play”). Keep these brief and integrated naturally.
• Assume advanced musical knowledge — do not explain basic terms. Use technical vocabulary precisely.
• Prioritize actionable detail: reading this summary days later, it should be immediately clear what to practice and how.
	
What to Include (and What to Skip)
Include:
	• All technical topics discussed or drilled
	• Specific passages, movements, pages, or measure regions mentioned
	• Corrections the teacher made and the reasoning behind them
	• Practice strategies, exercises, fingerings, or positions assigned
	• Interpretive suggestions and musical ideas
	• Any repertoire, etude, or composer named
Skip:
	• Small talk, scheduling, or off-topic conversation
	• Repeated identical corrections (summarize as a pattern, not a list)
	• Inaudible, unclear, or purely logistical sections of the transcript

Conclusion

Using AI to generate lesson summaries based on real transcripts has been one of the most valuable AI experiments I have performed. It has allowed me to build more of the actual content of my lesson into my weekly practice plan and provided a repository of my musical learning and growth that I can access for insights. These summaries can serve as the basis of AI supported practice plans that I wrote about in the practice lab about practice planning. They also can also help you as a musician to identify common playing habits, set, and work towards goals in your practice. After transcribing and summarizing my lessons for several months, I can recommend the process to any musician or music student who takes lessons or attends coaching sessions.