Transcribe German Audio to Text — Free, No Upload, Private

German is a language that punishes careless transcription. Compound nouns have to come out as one word, Rechtsschutzversicherung and not three fragments; umlauts and the ß must be written rather than approximated; nouns need their capitals; and the verb at the end of a long clause has to land in the right place. Whisper handles all of this well enough that the output reads as written German, which is what makes it editable instead of disposable.

This page opens with German selected, so the model never has to guess the language from the first seconds, which can go wrong with a Swiss or Austrian speaker. The recording, whether a Vorlesung, a meeting or an interview, is processed entirely on your own device: the model is downloaded to the browser the first time, and after that nothing leaves your computer, which keeps the DSGVO question simple, since no third party processes the audio.

What still needs a human: names of people and firms, figures and dates, and English terms dropped into German sentences, which can be written as they sound. Strong dialect lowers accuracy, and a quiet speaker far from the microphone lowers it more than any accent does. Pick the Fast model for a quick draft, Accurate or Best for a recording you will rely on, and expect a long lecture to take a while on a phone.