Get the lecture as text
Most lectures already exist as a file or a link: a recording on your phone, an .mp4 from Zoom or Teams, a video on YouTube or the course site. VoiceKeyboard's FILES screen turns any of them into a .txt file. Whisper, the recognition model, runs on your own PC, so a recording of a seminar does not leave your laptop.
- A recording on disk: drag it onto FILES, or press Browse. Audio (mp3, m4a, wav and others) and video (mp4, mkv, mov and others) both work.
- A video link: switch FILES to LINK and paste the address. The sound is downloaded and recognised on your PC.
With Pro, tick "Timestamps in the text": every phrase begins with its time, like [00:41:12], so you can go back to the exact place in the recording when a sentence in your notes makes no sense. Ask your lecturer before recording a class; many universities have rules about it.
![The FILES screen of VoiceKeyboard transcribing three files: a meeting recording, lecture.mp4 and an interview. The lecture's text appears line by line with times like [00:21:01]; BATCH PROGRESS shows a bar per file.](/assets/img/shots/files-batch-en-1920.webp?v=7ed8c10de7)
A routine that uses the transcript
Having the whole text is easy to confuse with having learned it. The transcript earns its keep when you use it to check yourself, not to reread.
- Skim and mark. Read the transcript fast, with Ctrl+F for the terms from the syllabus. Copy the three or four passages that carry the lecture into your notes, with their times.
- Close it and explain out loud. Open a blank page in Word, OneNote or Notion, press Ctrl+1 and explain the topic in your own words, as if to a friend who missed the class. VoiceKeyboard types it as you speak.
- Compare. Put your explanation next to the transcript. What you left out or got wrong is exactly what to review.
- Next day, repeat step 2 without looking. It goes faster each time.
Steps 2 and 4 are retrieval practice: pulling an idea out of memory instead of reading it again. A large review of study techniques in psychology (Dunlosky and colleagues, 2013) rated practice testing among the most useful methods, while highlighting and rereading came out low. Saying it is the quickest way to do it, since most people speak faster than they type, and the dictated text is there to compare afterwards.
Notes by voice, in any app
Dictation is not only for the routine above. It types into whatever window has the cursor: a Google Doc, a flashcard app, a chat with the study group, a prompt for an AI tutor. A few things help:
- The free plan gives 2,000 words a day with any engine, which is plenty for notes and summaries.
- On a train or in a library with bad Wi-Fi, switch RECOGNITION to LOCAL. The offline models work without the internet; see offline speech recognition on Windows.
- With Pro, voice commands put in a new line, a comma or brackets when you say them, which keeps structured notes tidy.
- Studying in a second language? Dictate your explanation in that language. Practising out loud is useful in itself, and the text shows you what you said.
Summaries and flashcards from the transcript
A transcript is good material for an AI assistant. Give the .txt to ChatGPT, Claude or Gemini and ask for a summary with the times of key points, or for ten questions to test yourself. Check its answers against the transcript: a model can be confidently wrong, and the times make the check quick.
What Free covers, and when Pro is worth it
| Free | Pro | |
|---|---|---|
| Recordings and links | 1 file and 1 link a day, up to 60 minutes each | no limit, any length, several at once |
| Timestamps | no | yes |
| Dictation | 2,000 words a day | no daily limit |
| Voice commands | no | yes |
The 60-minute limit is the one students meet first: a 90-minute lecture does not fit in Free. If you transcribe lectures every week, Pro at $9.99 a month billed yearly ($14.99 month to month) removes that limit and adds timestamps. If you need it only for exam season, take one month and cancel; a refund is possible within 14 days.
Languages
Whisper, which does all FILES work, hears about 99 languages and writes each in its own script. It writes the words in the language of the lecture and does not translate. For dictation the choice of engines depends on the language; each of the five interface languages has its own guide, starting with voice typing in English.
Start with one lecture you already have. Install VoiceKeyboard, open FILES, drop the recording in and read the result. Then try the explain-out-loud step with Ctrl+1. For lectures on YouTube, getting a transcript with timestamps has the details.
