Turn German speech into text on your own device: drop a meeting, an interview, a lecture or a video and get a German transcript, or .srt subtitles with timings. Whisper writes German the way it is printed, with capitalised nouns, umlauts and ß, and usually keeps compound nouns as one word, which keeps the transcript searchable. Standard German comes out best. Strong dialect, Swiss German above all, tends to come back as standard written German or with more mistakes. The recording is processed in your browser and never uploaded.
How it works
- Drop a German audio file or video. German is preselected as the spoken language.
- Pick Accurate (≈77 MB) for dialect, Austrian or Swiss speakers and echoing conference rooms; Fast (≈41 MB) for a quick draft of clear speech. Choose Subtitles under Output first if the goal is an .srt.
- Press Transcribe and check technical terms, names and long compounds, which is where small models slip. Edit in place, then download .txt, .srt or .vtt.
Questions
- Can it transcribe Swiss German (Schweizerdeutsch)?
- Only partly. Whisper has one German setting, learned from written German, and Swiss German is rarely written down. Expect the output to drift into standard German spelling, sometimes rewording a phrase, and more errors than on Hochdeutsch. Swiss speakers using standard German, as on the radio or in a lecture, transcribe like any other German recording.
- Is there German to English transcription?
- Not as one step: the transcript comes out in German. To get English, set Output to Subtitles, download the .srt and open it in the subtitle translator, which in Chrome 138 or later translates every cue on your device and leaves the timestamps alone. Fix the German before translating; a misheard word becomes a mistranslated one.
- What should I proofread in a German transcript?
- Names, numbers and rare compounds first: a long technical compound the model has not met can come back split in two or misspelled. Then sie and Sie, which sound the same and differ only in meaning, so the capital is the model’s guess. Every line plays back from the recording, so checking a word takes one click.
- Does my file get uploaded?
- No. The models and all processing run inside your browser on your own device. You can verify this yourself: open your browser’s DevTools, watch the Network tab, and process a file: there are no uploads. After the first visit most tools also work fully offline.
- How accurate is it?
- The models are OpenAI’s Whisper (tiny and base tiers). On clear speech they are very usable; heavy accents, crosstalk and noisy rooms will show errors. The Accurate tier is noticeably better on difficult audio.
- Which languages work?
- Fifteen, picked in the Language setting: English, Spanish, French, German, Italian, Portuguese, Dutch, Russian, Polish, Turkish, Japanese, Korean, Chinese, Arabic and Hindi. The decoder is conditioned on the one you pick, so it transcribes that language rather than approximating it in English — which also means picking the wrong one returns confident nonsense. The weights understand 99 languages; these are the ones we measured as good enough to offer.
- How long does it take?
- Transcription runs on your processor, so speed depends on the machine: expect around a quarter of the recording’s length, and less on the Fast model than on Accurate.