Romanian audio and video to text

Regional speech and Moldovan

Romanian is spoken in Romania and the Republic of Moldova; the language there is the same, though with a different accent and some Russian loanwords. Transylvanian, Moldavian, Banat and Wallachian speech also differ. Standard speech, as in news broadcasts, is the easiest case. Aromanian and other related varieties are different languages and are not covered.

Diacritics: ă, â, î, ș and ț

Romanian has five special letters. Text may use ș and ț with a comma below, which is the correct Unicode form, or the older cedilla forms ş and ţ, which many keyboards and websites still use. If your system needs one form, run a find-and-replace. Also check â versus î, since they sound the same and the spelling depends on the position in the word.

What Romanian speakers use it for

University lectures, work meetings, interviews, podcasts and voice messages. Many Romanians live abroad and switch between Romanian and the local language mid-sentence, which is the weakest case for any speech model, so those lines deserve a second listen. Check names of towns and people, since foreign-sounding or regional names are often misspelled.

Which model works best for Romanian

Romanian is supported with a good amount of training material, but less than the largest languages, so the small model makes more slips than it does for English or Spanish, especially on rare words, names and fast or noisy speech. Expect a draft that needs a proper read-through.

On tryscriptorium.com the no-account preview and regular accounts run the small model (the larger medium and large-v3 models run when you self-host Scriptorium). In an account, pick Small in the model menu rather than Auto: Auto prefers the fastest, least accurate model for long files or a busy queue.

The small model manages clear Romanian but may drop or change diacritics, and sometimes mishears similar-sounding words. Names, place names and foreign words are the usual corrections. Short clips with one speaker give the cleanest first draft.

Subtitles and captions in Romanian

Romanian subtitles typically use two lines of up to about 42 characters. Choose one diacritic style, comma-below being the correct one, and apply it to the whole file so that players do not show missing characters. SRT and VTT are UTF-8, and the captions stay in Romanian.

The uploader is set to Romanian

The upload box on this page transcribes as Romanian, so the recogniser does not have to guess the language from the first seconds. If the recording is in another language or switches languages, use Audio to text, which detects the language automatically. Automatic transcription is a draft and no accuracy figure is promised: check names, numbers and anything you will rely on.