What Breton results look like
Breton is a low-resource language for the speech model, with little recorded and written material to learn from. The draft will contain many errors, and on noisy or fast audio it may be more misleading than helpful. It does best on slow, clear speech from one person: a lesson, a reading, a careful interview.
Used that way, it still saves typing. You get timed lines to correct instead of an empty page, and the player follows along so you can replay any word.
Upload the recording as it is: MP3, M4A, WAV, OGG, OPUS, FLAC, a voice message or a phone recording. You can also record from your microphone in the upload box. Nothing needs converting first.
Spelling, mutations and dialects
The model writes in the common modern spelling (peurunvan), with the trigraph c'h, ñ marking a nasal vowel, and ù. Initial consonant mutations are a frequent source of errors: after certain words, b becomes v, k becomes g, t becomes d and so on, and the model may write the unmutated form or the wrong one. Check the beginnings of words carefully.
Breton has four traditional dialect areas: Kerne, Leon, Treger and Gwened (Vannetais). Gwened differs most from the others in pronunciation, so recordings from the Vannes area need more correction. Older native speakers in the countryside often use local words the model has rarely or never seen.
The transcript opens next to a player that highlights the current line. Click a line to hear it again, correct the word in place, and your changes save automatically.
French in Breton recordings
Nearly every Breton speaker also speaks French, and many recordings switch. Interviews with older speakers often drift into French; learners and Diwan pupils may pause in French for a word they do not know. Keep Breton selected when most of the recording is Breton. A French documentary with a few Breton quotes should be transcribed with French selected.
Typical Breton recordings
People bring collecting work with elderly native speakers, family recordings, songs and kan ha diskan sessions, school and evening-class material, radio programmes and local storytelling. Song and chant are hard for any speech model, so expect heavy editing there.
Export TXT, DOCX, PDF, CSV, SRT or VTT, with timestamps and speaker names switched on or off per export. Speaker labels are available on paid plans or with a Minute Pack.
For video files, see the Breton video to text page.