Croatian transcription: what to expect
The speech model handles Croatian well on clear audio, though it has heard less Croatian than Spanish, German or French. A quiet recording with one or two speakers usually gives a draft that needs some editing: a few word endings, names and less common terms. Phone calls, noisy rooms and people talking over each other need more review.
Croatian has many case endings, and these are where small errors show up. A noun may come out in the wrong case or a verb in a slightly different form, especially in fast speech. Reading the transcript while the audio plays catches most of these. The player highlights the current line, and you can click any line to hear it again and fix the text in place.
No conversion is needed. A lecture recorded as MP4, a dictaphone file in WAV or MP3, or a voice note from WhatsApp or Telegram can be uploaded as it is.
Croatian, Serbian and Bosnian
Croatian, Serbian and Bosnian are very close, and the speech model can confuse them. On auto-detect, a Croatian recording may be labeled Serbian or Bosnian. If it is taken as Serbian, the transcript may come out in Cyrillic, or with Serbian word choices such as ekavian forms. This page preselects Croatian so that guess is skipped.
Pick the language by what the speaker uses. Croatian speech is ijekavian (mlijeko, vrijeme) and uses words like tjedan, kruh and tisuća. Choose Serbian or Bosnian for those recordings, and Slovenian for Slovenian, which is a separate language. With Croatian selected, the text should follow Croatian spelling, but a word from a neighbouring standard can still appear, so check vocabulary if the text will be published.
Dialects and regional speech
Standard Croatian is based on Shtokavian, and that is what the model handles best. Kajkavian speech from Zagreb and northern Croatia and Chakavian speech from Istria, the coast and the islands differ in vocabulary and grammar. They are transcribed, but the text tends to move toward standard forms, and strong dialect needs careful review. For a dialect study where every local form matters, plan to correct a good share of the text by ear.
Everyday speech in Croatia also includes English and German loanwords, especially in work and tech conversations. Short borrowed words usually come through, sometimes in their original spelling and sometimes adapted.
Diacritics and details to review
The transcript uses the Croatian Latin alphabet with č, ć, dž, đ, š and ž. Check these points as you go.
- č and ć: they sound close in fast speech and in some regional accents, so words like kuća and noć are worth a look.
- đ and dž: a rare word or a surname may come out with the wrong one.
- Names, small places and company names are where most corrections happen.
- Numbers and dates may appear as digits or written out.
Who uses it and exports
People transcribe interviews for a završni or diplomski rad, journalism interviews, meetings, lectures and podcasts. Croatian families in Germany, Austria, Australia, Canada and the US bring recordings of grandparents, and researchers bring oral-history interviews from the diaspora.
TXT, DOCX, PDF, SRT, VTT and CSV exports keep the Croatian text as written. Word suits an interview appendix, CSV works for coding interviews in a spreadsheet, and SRT or VTT give you Croatian subtitles. Speaker labels are available on paid plans or with a Minute Pack.
RocketScribe writes the transcript in Croatian. To read it in English, use the Google Translate link on the transcript, or use the AI prompt export to ask ChatGPT, Claude or Gemini for a summary or translation. Files are never used to train AI models. Have important text checked by a Croatian speaker, since RocketScribe does not claim legal or medical compliance.