How to Transcribe Voice Notes Privately on a Linux Laptop

There are three private ways to turn voice notes into text on a Linux laptop. You can dictate straight into your apps with an offline tool like Speech Note or Handy. You can transcribe recorded voice note files with Buzz or whisper.cpp. Or, since most voice notes are recorded on a phone anyway, you can transcribe on the phone with an on-device app and move only the text to the laptop. All three keep your audio off other people’s servers.

Which approach fits how you take voice notes? #

Option 1: Speech Note, a notes app with offline speech to text #

Speech Note is a Linux app for taking and reading notes by voice. Its Flathub listing says that “text and voice processing take place entirely offline, locally on your computer, without using a network connection.” It transcribes from the microphone and from audio files, and it can also read text aloud and translate between languages. GPU add-ons are available for AMD and NVIDIA cards.

flatpak install flathub net.mkiol.SpeechNote

Open it, download a speech model for your language from its settings, and start dictating. It suits people who want voice notes to live in a dedicated app.

Option 2: Handy, push-to-talk dictation into any app #

Handy is a small, free, open-source app for Linux, Windows and macOS. You set a keyboard shortcut, hold it, speak, and let go. Handy transcribes locally with Whisper or Parakeet V3 and pastes the text into whichever app has focus: your editor, terminal, browser or notes app. Voice activity detection trims silence.

It’s the closest thing to a system-wide voice keyboard on Linux, and it’s completely offline. It doesn’t transcribe files, so pair it with one of the tools below for recordings.

Option 3: transcribe recorded voice notes with Buzz or whisper.cpp #

For audio files, Buzz is the friendly option. Install it from Flathub, open a file, choose a model and language, and export text:

flatpak install flathub io.github.chidiwilliams.Buzz

For a folder of notes, whisper.cpp is quick to script. Once it’s built (see its README), a short loop converts and transcribes everything in a folder:

for f in notes/*.m4a; do
  ffmpeg -loglevel error -i "$f" -ar 16000 -ac 1 -c:a pcm_s16le "${f%.*}.wav"
  ./build/bin/whisper-cli -m models/ggml-base.en.bin -f "${f%.*}.wav" -otxt
done

Each note gets a matching .txt file. The base.en model is fast and fine for clear speech in English. Use small or medium for accents or noisy recordings, and a multilingual model for other languages.

Option 4: transcribe on your phone, move only the text #

If your voice notes start on your phone, you can skip moving audio files. Private Transcribe runs Whisper on iPhone and Android with no account and no upload. It works in airplane mode after a one-time model download. You record the note, the phone transcribes it, and you send the text to your laptop however you like: paste it into a message to yourself, or use a local tool such as KDE Connect.

There’s no Linux version of the app. The point is that the transcription happens before anything reaches the laptop, and the audio can stay on the phone. You can even have the app delete it automatically and keep only the text.

How to move audio from phone to laptop privately #

If you’d rather transcribe on the laptop, avoid routing recordings through a cloud drive:

• USB cable: the simplest option, and nothing leaves the room.
• Syncthing: peer-to-peer sync between your own devices, with no central server holding your files.
• KDE Connect or GSConnect: send files and clipboard contents between Android and a Linux desktop over your local network.

Get better recognition from your laptop’s mic #

Laptop mics sit next to the keyboard and the fan, which is the worst place for speech recognition. A cheap USB headset or clip-on mic usually improves results more than a bigger model does. Check that the right input device is selected. pavucontrol works for both PulseAudio and PipeWire. Record a test sentence before trusting a new setup.

For more on hardware, see better microphones for speech recognition accuracy.

Keep voice notes private on Linux #

Local transcription removes the vendor, but your notes are still files on a laptop. Use full-disk encryption (LUKS, usually offered by your distro’s installer), don’t keep notes in a folder synced to a cloud service unless you mean to, and delete audio once you’ve got the text you need.

For longer recordings with several speakers, see how to transcribe interviews offline on Linux. For live dictation tools beyond Linux, see real-time speech to text with zero cloud uploads.

Frequently asked questions #

What’s the easiest offline speech to text for Linux? #

For dictation, Handy: install it, set a shortcut, and talk. For a notes app with speech built in, Speech Note from Flathub. For turning audio files into text, Buzz. All three are free and work offline.

Does offline Whisper run on an older Linux laptop? #

Yes, with smaller models. whisper.cpp is designed to run efficiently on CPUs, and the tiny and base models run on modest hardware. Larger models need more memory and much more time without a GPU.

Can I transcribe voice notes in languages other than English on Linux? #

Yes. Use a multilingual Whisper model, available in Buzz, Speech Note and whisper.cpp. Handy’s Parakeet V3 model also detects the language automatically. Accuracy is highest for widely spoken languages.

Is it better to transcribe voice notes on my phone or my laptop? #

If the notes are recorded on your phone, transcribing there saves you moving audio files around, and the audio never has to leave the phone. The laptop is better for long files, big batches and multi-speaker recordings, where more processing power and larger models help.