Transcribe a recorded class or meeting
The Zoom mp4, the Teams recording, the audio from your phone: drop it and get the text, on your PC.
Updated: September 2026 · 6 min read
To turn a recorded class or meeting into text on Windows without uploading the file anywhere: open AppGrabbit › Transcript, drop the recording (mp4, mkv, mov, webm, m4a, mp3, wav, ogg, opus…) or pick it, choose the prompt ("summary" for a meeting, "study questions" for a class) and click Transcribe. The engine runs on your computer; you get a .txt named after the recording and an .srt with timestamps. A 2-hour class takes about 6 minutes with an NVIDIA card or about 25 on the processor. 1 credit · 5 free a month
Where your recording is
The usual path for each tool
- Zoom — the
Documents\Zoomfolder: one.mp4per meeting (and often an.m4awith just the audio, which is enough and lighter). - Microsoft Teams — the recording lives in OneDrive or SharePoint: open it and use Download; it arrives as
.mp4. - Google Meet — in Google Drive, folder
Meet Recordings; download it as.mp4. - OBS, the Windows recorder or your phone —
.mp4,.mkv,.m4a,.wavor.3gp; all go in the same way.
Why not upload it to a website
A meeting holds names, figures and decisions that aren't yours to give away. A class is usually the teacher's material. Transcription websites keep the file for as long as their terms say, and their free plan runs out after 30 minutes of audio. AppGrabbit does everything on your PC: the file never leaves your disk and there is no length cap.
How it works and how long it takes
The app pulls the audio out of the recording with FFmpeg, transcribes it with Whisper (whisper.cpp) on your PC, skips music and silences with a voice detector, and writes the .txt (with the prompt you pick on top) and the .srt. With an NVIDIA card it takes about 30 seconds per 10 minutes of recording; on the processor alone, about 2 minutes per 10 with the Fast model or about 6 with the Accurate one. The first time it downloads the engine once (it tells you the size) and carries on by itself.
What to do with the text
A one-hour meeting is about 9,000 words: it fits in ChatGPT, Claude or Gemini. With the "Summary" prompt on top, you paste the .txt and get the decisions and the to-dos; with "Study questions", a class becomes a quiz to practice; with "Key ideas", an interview becomes quotes with their minute in the .srt.
What to know, honestly
The engine doesn't separate speakers: the text comes out as running text, and in a meeting with several people you recognize the voices by context and minute. When two people talk at once, part is lost. A recording with the microphone far away or with echo comes out worse than a class with the teacher close by. The Fast model gets about 8 words in 100 wrong and the Accurate one about 5, mostly proper names; the .txt says so at the top and the .srt takes you to the exact second.
Step by step
- 1Find the recording — Zoom: Documents\Zoom. Teams: download it from OneDrive or SharePoint. Meet: Google Drive › Meet Recordings. The full .mp4 or just the audio both work.
- 2Open AppGrabbit › Transcript — Install it free if you don't have it (109 MB, Windows 10 and 11). Windows will say it protected your PC because the installer is not code-signed: More info › Run anyway.
- 3Drop the file or pick it — The card tells you how long it will take on your PC. The first time it asks to get ready: pick Fast or Accurate and click Set up and transcribe.
- 4Pick the prompt — Summary for a meeting, study questions for a class, key ideas for an interview, or no prompt.
- 5Click Transcribe and paste the .txt into your AI — The .txt and the .srt land in your Transcripts folder and in the Library, named after the recording.
Frequently asked questions
No. A 3-hour class goes in the same way; it just takes longer and needs disk space for the temporary audio (about 350 MB for 3 hours, deleted when it finishes).
No. The text comes out as running text with timestamps in the .srt. For a meeting, the minute and the context are usually enough to know who said what.
Yes, once downloaded as mp4. The app never signs in to your Microsoft account; it works with the file.
No. It works the same on the processor, it just takes longer (about 2 minutes per 10 minutes of recording with the Fast model). If you have an NVIDIA card, the app uses it on its own.
The engine detects the language; if the class mixes languages, you can fix it (English, Spanish or Portuguese).
1 credit per recording, however long. Free comes with 5 credits a month and, when you hit 0, you can claim 1 more every 48 hours. Pro (USD 8.99 a month) or the 7-day pass (USD 3.99) remove credits.
Related guides
Your classes and meetings as text, without leaving your PC
Windows 10/11 · Drop the recording and go · 1 credit, 5 free a month