German speech to text · Free hours for students
Transcribe German audio to text, with speakers and timestamps.
Upload a German interview and get a draft with speaker labels and a timestamp on every speaker turn, then correct it and export to Word. Students in Germany, Austria, Switzerland and 10 other European countries get 7 hours of audio free each month.
7 free hours of audio a month · No card needed
Free hours for students who confirm a university email, for example at LMU Munich, Humboldt-Universität zu Berlin, the University of Vienna or ETH Zurich.
Try an example
Interview about student life, in German
German · Two speakers · Fictional excerpt
Interviewerin 00:12
Was hilft dir, wenn dir eine Hausarbeit über den Kopf wächst?
Student 00:18
Es macht echt einen Unterschied, dass wir in der Lerngruppe über die Arbeit reden können. Dann suche ich nicht allein nach einer Lösung.
Interviewerin 00:31
Kannst du von einer Situation erzählen, in der das konkret geholfen hat?
How to do it
How to transcribe a German interview
1. Sign up with your university email
Confirm your university email with the code we send you and get 7 free hours a month, with no card. Student email not recognised? Write to us.
2. Upload the recording and choose German
Upload an mp3, m4a, wav or mp4 and choose German, or let Spirelight detect the language. Choosing it helps when a recording is short or starts with small talk in English. For Swiss German, choose German too: there is no separate setting.
3. Listen, correct and export to Word
Check compounds, names and specialist terms with the built-in player, rename the speakers (for example to I and B1) and add the marks your transcription rules ask for. Export to Word, plain text or SRT.
Accuracy
What makes German hard for speech recognition
Accuracy depends mostly on the recording, but German adds weak spots of its own. A long, rare compound such as “Studienabbruchquote” (dropout rate) can come out as separate words, or as a more common word that sounds similar. Regional accents are a matter of degree: a speaker from Bavaria, Saxony or Austria who talks close to Standard German is a different case from broad dialect. For a better draft:
- Note names of people, places and organisations during the interview, so you can check spellings later.
- Record online interviews with each person on their own microphone, and avoid cafés and canteens.
- In group interviews, agree that people take turns.
Dialects
Swiss German, Austrian German and strong dialect
In German-speaking Switzerland, people speak Swiss German dialects in almost every everyday situation but write Standard German, and the dialects have no official spelling. A participant might say “Sie händ aagfange” for “Sie haben angefangen” (they started).
Dresing and Pehl’s handbook lists strong dialect, such as Swiss German, among the recordings where automatic transcription works badly or not at all. Where a system copes, it tends to write Standard German, so the draft is closer to a translation than a transcript. Expect a rough draft at best, check it line by line and be ready to type passages yourself.
That is less of a problem than it sounds, because Swiss handouts often ask for written German anyway. The University of Basel’s cultural anthropology handout, for example, has Swiss German interviews written in Standard German, keeping telling dialect words such as “Bebbi” (a person from Basel). What you check is that the draft says what your participant meant.
Austrian Standard German has words such as “Jänner” (January) and “heuer” (this year). They are correct written German in Austria, so leave them as they are.
Transcription rules
Dresing and Pehl, Kuckartz or GAT 2: what the draft gives you
Many students in Germany, Austria and Switzerland are asked to follow the rules of Thorsten Dresing and Thorsten Pehl, from their free handbook “Praxisbuch Interview, Transkription & Analyse” (9th edition, 2024). Others use the simple rules of Udo Kuckartz and colleagues. Conversation analysis often uses GAT 2 (Selting and colleagues, 2009), which records stress, pitch and measured pauses; an automatic draft gives you the words at most.
Dresing and Pehl’s content-semantic rules (inhaltlich-semantische Transkription, called “einfache Transkription” in older editions) ask for every word, with slurred speech and dialect brought closer to written German: “hammermal” becomes “haben wir mal”. Short signals such as “hm” or “genau” are left out unless they answer a question. Pauses of three seconds or more are marked (…), stressed words go in CAPITALS, unclear words are marked (unv.), and every turn starts a new paragraph with a code such as I: or B1:.
In their view, you cite passages by paragraph number; timestamps are for jumping back to the audio. Check which edition your department expects, and see our guide to verbatim transcription for how much detail you need.
Spirelight’s draft gives you the words, the speaker turns and a timestamp on every turn. It doesn’t follow any rule set by itself, so while you proofread you still:
- Rename the speakers to I, B1 and so on. One change applies to the whole document.
- Check the wording and handle dialect the way your rules say.
- Add pauses, CAPITALS for stress, laughter in brackets and (unv.).
- For the extended rules, check for filler sounds such as “ähm” and add any that are missing. “Word for word” only changes English recordings; for German both settings give the same draft.
- Anonymise names, places and other details that could identify someone.
Writing in English
German interviews, English thesis
Many master’s programmes in Germany, Austria and Switzerland are taught in English, so plenty of theses are written in English about interviews held in German. Spirelight doesn’t translate: the transcript stays in German, which is what you want for the analysis anyway. Practice differs between chairs (Lehrstühle) and institutes, so check your supervisor’s guidelines. A common approach:
- Code and analyse the German transcript, not a translation.
- Translate only the quotes you use and say the translations are yours, for example “(my translation)”.
- Give the German original in a footnote if your chair asks for it, and keep the full transcripts for the appendix.
- Ask a native speaker to check key quotes, especially idioms and irony.
Questions about transcribing German audio
Can I transcribe German audio to text for free?
Yes, if you study at a university or college in Germany, Austria, Switzerland or one of 10 other European countries. Confirm your university email with the code we send and you get 7 hours of audio free every month, with no card and no trial; free hours don’t roll over. Everyone else pays €2 per hour, or €14 a month for Spirelight Plus with 10 hours included. See pricing.
How accurate is German speech to text?
It depends mostly on the recording. Clear Standard German, one speaker at a time, gives the best draft; noise, overlapping speech, strong dialect, long compounds and specialist terms cause more errors. Treat it as a first draft and check every quote against the audio.
Can Spirelight transcribe Swiss German?
Choose German, but be realistic. Strong Swiss German dialect gives a rough draft at best, and where speech recognition copes, it tends to write Standard German. Check the draft line by line, and test one interview before you plan a whole project around it.
Can Spirelight translate German audio to English?
No. Spirelight doesn’t translate: you get a German transcript of what was said. Translate the quotes you use yourself, say the translations are yours and keep the German transcript for your appendix.
Does Spirelight’s transcript follow Dresing and Pehl’s rules?
Not by itself. You get the words, the speakers and a timestamp on every speaker turn. While you proofread, you add pauses, stress, laughter, (unv.) and speaker codes such as I: and B1:, and check for any filler sounds such as “ähm” that your method needs.
Where does my German audio go when I upload it?
Transcription runs on AssemblyAI’s servers in the EU. Your files, transcripts and account are stored with Google Firebase in the US. If you use AI summaries or chat, the text is processed by OpenAI in the US. We don’t offer a data processing agreement (DPA). Our terms and privacy policy have the details.
Can I upload interviews with personal data at a German, Austrian or Swiss university?
It depends on your university’s rules, and this isn’t legal advice. Data protection law applies to students’ own interview projects too, so get informed consent and check with your supervisor or the university’s data protection officer (Datenschutzbeauftragte) before you upload. Some universities may expect a processing agreement (Auftragsverarbeitungsvertrag) with services that handle personal data for you, and we don’t offer one. Switzerland has its own Federal Act on Data Protection.
Other languages
Transcribing another language?
Each of these pages covers what to expect from that language’s recordings. Interview transcription covers file formats, proofreading and English recordings.
Get started
Stop typing out interviews by hand
Your next step is simple: sign up with your university email and upload your recording. You get a draft you can correct and download in Word.
7 free hours a month with an eligible university email · No card · No trial