Belarusian audio to text: an honest view
Belarusian has far less training data in the speech model than Russian or Ukrainian. Clear audio of one speaker gives a usable draft, but expect to read every line. The most common problem is Russian leaking in: a word spelled the Russian way, or a sentence that drifts into Russian spelling when the speaker's pronunciation is close to Russian. Akanne, the Belarusian habit of writing an unstressed o as а, is a good place to look: a word written with о where Belarusian spells а is usually a Russian slip.
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.
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.
Belarusian letters and spelling
The transcript uses the Belarusian Cyrillic alphabet, including ў and і, the apostrophe in words like сям'я, and no и or щ. Check for these Russian letters during review, since they are a sign the model slipped into Russian.
Two spelling norms are in use: the official one taught in schools and the classical taraškievica favoured by some media and writers. The model does not let you choose, and its output usually follows the official norm, so if you publish in taraškievica, expect to adjust spellings. There is no Łacinka (Latin) output.
Russian, trasianka and language choice
Many recordings mix Belarusian and Russian, and everyday speech in some areas is trasianka, a blend of the two. Choose the language by what dominates:
- Mostly Belarusian: keep Belarusian selected and fix Russian forms during review.
- Mostly Russian with Belarusian phrases: choose Russian.
- Do not rely on auto-detect for short clips; it often picks Russian for Belarusian speech.
Typical recordings and exports
People bring journalism and documentary interviews, oral histories, recorded lectures and literary readings, podcasts produced in Belarus and abroad, and voice messages from family. Researchers in Poland, Lithuania and elsewhere use the CSV export to code interviews line by line. Recordings of older speakers from villages often contain regional words that the model does not know, so give those passages extra attention.
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 Belarusian video to text page.