How well Catalan transcription works
Catalan is one of the languages the speech model handles well. On a clear recording with one or two speakers, the result is usually close to what you get for Spanish or French: a transcript that needs light editing rather than a full rewrite. Names of people, small towns and companies are where most corrections happen, as in any language.
Audio quality matters more than accent. A lecture recorded from the back of a room, a phone call on speaker or a group talking over each other will need more checking. Put the microphone near the person speaking and the result improves more than any setting can.
You do not need to convert anything first. A class recorded in Teams as MP4, a podcast in MP3, a WAV from a recorder or a WhatsApp or Telegram voice note can be uploaded as it is, and the audio is taken from video automatically.
Valencian, Balearic and other varieties
Choose Catalan for every variety: Central Catalan from Barcelona, Valencian, Balearic, North-western Catalan from Lleida, and the Catalan spoken in Andorra, Roussillon and Alghero. There is no separate Valencian option in the language list, and you do not need one.
Central Catalan is the variety the model has heard most, so it tends to come out cleanest. Valencian and Balearic speech is understood, but the transcript may not always keep local forms. A Valencian speaker who says "este" or "eixir" may sometimes see a more central form in the text, and Balearic articles such as "es" and "sa" are worth checking against the audio. If the regional form matters, for example in a dialect study or an oral-history project, listen and correct those words in place.
Catalan and Spanish in the same recording
Many conversations in Catalonia, Valencia and the Balearic Islands move between Catalan and Spanish, sometimes inside one sentence. The transcript follows the language selected in the upload box. A Spanish phrase in a Catalan recording may come out in Spanish, or it may be written in a Catalan-like form, and now and then a short Spanish sentence may be rendered into Catalan. Review the passages where speakers switch.
Auto-detect can also mistake Catalan for Spanish, especially on a short clip or one that opens with a few words of Spanish. If that happens the whole transcript can come out in Spanish. This page preselects Catalan to avoid that. If a recording is mostly Spanish with a few Catalan phrases, choose Spanish for that file instead. You choose the language per file.
Accents and spelling to check
The transcript uses standard Catalan spelling. These are the details most worth a look when you review.
- Open and closed accents: à, è, é, í, ò, ó and ú. The accent on è versus é or ò versus ó is easy to get wrong in rare words.
- The punt volat in l·l, as in col·legi or il·lusió. It may appear as a different dot or as plain ll.
- Ç, as in plaça, and the diaeresis in words like veïna or qüestió.
- Apostrophes in l', d' and s' before vowels, and hyphens in pronoun forms such as dona-m'ho.
- English words and brand names used in Catalan are usually kept in their English spelling.
Who uses it and what to do with the transcript
People bring university lectures and thesis interviews (treball de fi de grau or de màster), local radio and podcast episodes, council and association meetings, journalism interviews, and family recordings that need to be kept in writing. Teachers and video makers use it to produce Catalan subtitles.
TXT, DOCX, PDF, SRT, VTT and CSV exports keep the Catalan text as written, including accents and l·l. SRT and VTT give you subtitle files with timings for a video editor. On paid plans or with a Minute Pack, speaker labels show who said what.
RocketScribe transcribes Catalan in Catalan. To read it in English, use the Google Translate link on the transcript. For a summary, notes or a translation, use the AI prompt export and paste the result into ChatGPT, Claude or Gemini. Your files are never used to train AI models. RocketScribe does not claim legal or medical compliance, so have important text checked by a Catalan speaker.