Transcribe lecture recordings from a phone, Zoom or YouTube into text with a summary, timestamped paragraphs and answers to your questions.
To transcribe lecture recordings into notes, upload the audio or video file or paste a link to it. The transcript comes back with a short summary on top, timestamped paragraphs and labelled speakers. Then search it, ask it questions, fix the odd term and export it as PDF, Markdown or text to study from.
This guide is for students with a pile of recordings: a voice memo from the front row, a Zoom or Teams recording, a lecture on YouTube or a university video platform. It covers recorded lectures. For captions while the lecture is happening, see Live transcription for churches, lecture halls and council chambers.
The short version, in the VocalStack Dashboard:
When it is done you get the transcript, a title and a summary, and the audio to play back from any line.
Most lecture recordings are files already. The upload takes audio and video: MP3, M4A, WAV, FLAC, OGG, MP4, MOV and WebM all go straight in, and so do most other video formats. A voice memo from an iPhone is an M4A; an Android recorder usually saves MP3 or M4A. Either works as it is, with no converting.
For video, VocalStack transcribes the sound and keeps the picture with the recording, so you can watch the slide that was on screen when a line was said. How long the video is kept depends on the plan (more in the cost section).
Two limits matter for long lectures. On Pay as you go, one file can be up to 2 GB and one transcription up to two hours of audio. On Premium the limits are 5 GB and eight hours. A three-hour seminar on Pay as you go can be trimmed into two parts, each its own transcription.
If the lecture is online, paste the link instead of downloading it. VocalStack checks the link first, shows you its title and length, and then fetches the audio itself.
What works from a link:
From a video site, only the audio is fetched. The picture you see next to the transcript is the site's own player.
What does not work from a link: anything behind your university sign-in. That covers Panopto, Echo360, Moodle, Canvas, Blackboard, a Kaltura MediaSpace on your university's own address, and Teams recordings stored in a work or school SharePoint. VocalStack cannot sign in as you. If the platform offers a download button, download the file and upload it. Lecturers can switch downloads off; if they have, ask them for the file.
As a last resort, the browser extension records the audio of a tab while the lecture plays, and the desktop app does the same for any app on a Mac. Both make a live transcription, which runs in real time and is billed at the live rate, so use them only when there is no file to get.
Lecture recordings fail in predictable ways, and most of the fixes happen before you press record.
Distance and echo. A phone on the desk in row twelve hears the lecturer once directly and several more times off the walls. That blur is the biggest single cause of wrong words. Sit near the front, put the phone on the desk facing the lecturer rather than in a pocket or bag, and if the lecture uses a microphone, sit near a speaker. A recording taken from the lecture hall's own system (the version the university posts) is almost always cleaner than one made in the room.
Questions from the back. A student asking a question twenty metres away is often too quiet to transcribe at all. Lecturers can help everyone by repeating each question into the microphone before answering it.
Technical terms. Standard academic vocabulary, from drug names to Latin anatomy, is recognised as it is. What no tool can know in advance is your lecturer's own shorthand, a new coinage or an obscure surname. In a recorded transcript you correct those in a few clicks (see below). For a live lecture, a Reference Script with the slides or reading list fixes them as they are spoken; that is covered in the live transcription article.
The wrong language. Automatic detection works from the first stretch of speech. A lecture that opens with a minute of chatter, or a lecturer with a strong accent, can be detected as the wrong language. If you know the language, name it before you start. That also helps on noisy recordings.
Audio transcription errors: 7 causes and how to fix each goes through each kind of mistake in more detail.
A transcript of a 90-minute lecture runs to well over ten thousand words. Nobody studies from that. These parts of the transcription turn it into something closer to notes.
The summary on top. Once the transcript is final, VocalStack writes a title and a short summary of one or two paragraphs, keeping the names and numbers that came up. It reads like the abstract of the lecture: enough to remember which week covered what. It is written as prose, and your own notes still matter.
Paragraphs with timestamps. The transcript is split into paragraphs by topic, each with its start time. Use them as headings when you revise. Every line keeps its time in the recording, so you can play back from that moment and hear again the part you did not follow.
Speaker labels for the Q&A. Speaker labels are on When Needed by default: a lecture with one voice reads as plain prose, and once a second voice is heard each line says who spoke. You can name the speakers, so the lecturer's lines are marked by name and the questions stand apart.
Ask about the lecture. Ask is an assistant inside the Dashboard that answers questions from your transcripts. Open a lecture and ask "what were the three conditions for the theorem?" or "where did she explain the difference between mitosis and meiosis?". The answer links to the moments it came from. Ask it about all your transcripts at once to find which lecture covered a topic.
Search every lecture. Search finds exact words across all your transcriptions, and also passages that mean the same thing in different words. It works across languages, so a search in English finds a lecture given in Spanish.
Corrections. Fix a misheard term right in the transcript. The correction is saved to the transcript and carries into every export.
Exports. Download the transcript as PDF, Markdown, plain text, Excel, JSON, or as SRT or WebVTT subtitles. The text, Markdown and PDF exports start with the title and summary, and timestamps can be set per paragraph, per line, or left out. Markdown drops straight into a notes app such as Obsidian or Notion.
Tags for each course. Tag each transcription with its course, and the list of transcriptions filters down to one module at a time.
VocalStack transcribes 100 languages and detects which one is spoken. Each transcription can be translated into up to 10 languages, and the transcript, its summary and its exports switch to the one you pick. A lecture given in German can be read and searched in English, with the German still there to check a term against.
Translation costs nothing on top of the transcription.
For a student who cannot hear the recording, a transcript is the recording. Speaker labels, paragraphs and the summary make it something to study from as well as a record. Universities usually have accessibility rules for recorded teaching, and automatic transcripts may or may not satisfy them; human-checked captions are sometimes required. Correcting the transcript where it matters (names, terms, numbers) closes most of the gap. Lecturers can share a finished transcription through a public link, so students read it in their own language without making an account.
Many universities have a policy on recording lectures, and many lecturers have their own. Some allow it for personal study, some require permission, and some allow it as an accessibility adjustment only. Lecture material may also be the university's or the lecturer's copyright. Check your course's policy, ask the lecturer when in doubt, and keep recordings and transcripts for your own study unless you have permission to share them. This is general guidance, not legal advice.
At today's prices, a recorded file costs $0.20 per hour of audio, billed per second. On Pay as you go each transcription is billed for at least 5 minutes; on Premium there is no minimum. Translation, summaries, Ask and search are included. The current rates are on the pricing page.
At those prices:
Partly. A new account gets $5 of one-time signup credit and asks for no card. At the file rate that is about 25 hours of recorded lectures, roughly half a term of one course. After that you top up credit as you need it, from $10.
Pay as you go has a few limits worth knowing for a term's worth of lectures. Source audio is kept for 30 days and uploaded video for 3 days; the transcript itself is kept until you delete it. After 30 days you can still read, search and export the transcript, but not play the audio. Pay as you go also includes 30 Ask questions a day and 60 meaning-based searches an hour; exact-word search has no limit. Premium is a subscription that includes credit every month, keeps audio and video while you subscribe, and removes those limits. The pricing page has the current figures for each plan.
Yes. Paste the YouTube link into a new transcription. VocalStack fetches the audio and transcribes it as a file, at the file rate. The video plays in YouTube's own player next to the transcript.
Not from the link, because those platforms need your university sign-in. Download the recording from the platform if your lecturer has allowed downloads, then upload the file.
It depends mostly on the recording. A lecturer on a microphone, recorded from the lecture hall's system, comes back with few mistakes, and most of them are rare names and course terms. A phone at the back of an echoing hall gives many more errors, and quiet questions from the audience may be lost. Name the language when you know it, record as close to the speaker as you can, and correct the terms that matter.
It gives you the raw material: a summary, paragraphs with timestamps, speaker labels, and an assistant that answers questions about the lecture with links to the moment. Export to Markdown or PDF and edit from there. It does not write flashcards or bullet-point revision notes for you.
Yes. The spoken language is detected from 100, and the transcript and summary can be translated into up to 10 languages at no extra cost.
Yes, as a live transcription, which costs more per hour than a file. See live transcription for lecture halls.
To start, upload your first recording at {{app}}/transcriptions/new/file. The signup credit covers about 25 hours of lectures.