Transcribe audio files offline, with speaker labels

From the tray

1. 1

   Open the tray entry

   Click the Vocalinux tray icon and choose Transcribe Audio File…

2. 2

   Pick a recording

   Select a WAV file, or an MP3, OGG, FLAC, M4A or Opus file if ffmpeg is installed, and click Open.

3. 3

   Download TinyDiarize once

   The first time, Vocalinux asks to download the TinyDiarize model (about 465 MB, one time). Click Download. It is only used for file transcription and never becomes your dictation model.

4. 4

   Wait for the transcript

   A notification says the file is being transcribed. Long files can take a while. When it finishes, a Transcript window opens with timestamps and speaker labels.

5. 5

   Copy or save it

   Use Copy to put the full transcript on the clipboard, or Save As to write it to a text file (transcript.txt by default).

The model download is checked against a pinned SHA-256 digest before use, like every other Vocalinux model. The status line under the transcript reports how many blocks were found and the highest speaker number.

From the command line

--transcribe-file PATH transcribes one file, prints the diarized transcript to standard output and exits:

user@linux:~$ vocalinux --transcribe-file meeting.wav

• It needs no display and no tray, and it runs alongside a Vocalinux instance that is already running, so it works over SSH or from a script.
• It does not download the model. Fetch TinyDiarize once through the tray entry first, otherwise the command stops with a message saying the model is not downloaded yet.
• Because the transcript goes to standard output, you can redirect it to a file with your shell. Errors go to standard error, prefixed with vocalinux: could not transcribe, and the exit status is 1. A file with no detectable speech prints vocalinux: no speech detected.

Output format

Each block is one speaker turn: a [HH:MM:SS] timestamp for where it starts, the speaker label, then the text. Blocks are separated by a blank line. The Transcript window, Copy, Save As and the CLI all use the same layout. The lines below show the shape only; they are not real output.

[00:00:00] Speaker 1: <first turn>

[00:00:14] Speaker 2: <second turn>

[00:00:31] Speaker 3: <third turn>

How speaker labels work

TinyDiarize is a small, English-only whisper.cpp model whose output includes a speaker-turn marker wherever it detects a change of voice. Vocalinux runs it in whisper.cpp's tinydiarize mode and splits the text at each marker. Each new block gets the next number.

That makes the labels positional, not identities. In a two-person interview the transcript reads Speaker 1, Speaker 2, Speaker 3, Speaker 4 as the two people take turns, rather than alternating between 1 and 2. Turn boundaries are useful for reading and editing; assigning names is up to you.

FAQ