Armenian transcription: set your expectations
Armenian is a low-resource language for speech models. The output is a rough draft that needs a full review, with corrections expected throughout. Check every name, place and technical term.
Clear audio helps more than anything else. One speaker close to the microphone in a quiet room, such as a lecture, a sermon recorded near the pulpit, or a dictated text, gives the most usable draft. Family gatherings, phone calls, people talking over each other and speech over music give much weaker results. There, the timestamps mostly help you find your place while you type.
Reviewing a draft is still faster than typing Armenian from scratch, especially on a keyboard without an Armenian layout. Play the file, follow the highlighted line, and click any line to hear it again. For family histories, publications or anything you plan to quote, have a fluent reader check the final text, ideally someone who speaks the same variety as the person recorded, since Eastern and Western speakers can read the same draft differently.
Eastern and Western Armenian
The model leans toward Eastern Armenian, the standard in Armenia and among Armenians in Iran and Russia. Western Armenian, spoken in Lebanon, Syria, France, the US and Istanbul, has different pronunciation: sounds written with բ, դ and գ in Eastern are heard as p, t and k in Western, and the reverse. The model can pick the wrong letter as a result, and Western grammar forms may be pulled toward Eastern ones. Plan extra review time for Western Armenian.
Spelling is another split. Armenia uses the reformed orthography, while Western Armenian and many Iranian Armenian writers use the classical one. The transcript will mostly follow reformed spelling, so if you need classical spelling, budget time to convert it.
Armenian script: points to check
The transcript is written in the Armenian alphabet. These are the details a review most often has to fix.
- Punctuation: Armenian uses the full stop ։ and marks questions and emphasis with ՞ and ՜ over the stressed vowel. The model often uses a Latin period or question mark at the end instead.
- Similar sounds such as ծ and ց, or ք and կ, may be swapped.
- The ligature և may be written as եւ, or the reverse.
- Numbers may appear as digits or written out in words, and dates may mix both. Pick one style and apply it across the file.
- Surnames ending in -յան or -յանց usually come through, but the root of the name often needs correcting.
- Russian and English loanwords may come out in Armenian letters, in Latin or Cyrillic letters, or as a misheard Armenian word.
Auto-detect, mixing and who uploads
In Armenia, everyday speech often mixes in Russian words; in the diaspora, English, French, Arabic or Persian. On auto-detect, a clip that opens in Russian or English may be transcribed in that language. This page preselects Armenian, so keep it unless another language clearly dominates the recording.
Common uploads include oral histories and family recordings from the diaspora, church services and talks, university lectures, journalist and research interviews, podcasts, and YouTube videos that need Armenian subtitles.
Uploading and exporting
WhatsApp and Telegram voice notes, phone recordings and MP4 or MOV videos upload as they are. The transcript comes in timestamped paragraphs, and the player highlights the current line. Click a line to replay it and fix the text in place.
Exports to TXT, DOCX, PDF, SRT, VTT and CSV keep the Armenian script as written. To read the content in English, use the Google Translate link on the transcript, or the AI prompt export to ask ChatGPT, Claude or Gemini for a translation or summary. Speaker labels are available on paid plans or with a Minute Pack. Your files are never used to train AI models.