German speech dataset: 1,534 hours in stock
1,534 hours of unscripted German recorded in January and February 2026: 1,012 hours of two-person conversation, each speaker on their own channel, and 522 hours of monologue, with machine transcripts and word-level timestamps. Speakers are mostly in Germany, with Austrian and Swiss speakers too. Download access within 24 hours of a signed order.
- In stock
- 1,534 hours
- Recordings
- 2,984
- Speakers
- 58+
- Audio
- 48 kHz WAV, stereo and mono
- Transcripts
- Machine, word timestamps
- Price
- $35 per hour, full set
Need speakers, topics or recording conditions this set does not cover? We also record speech data to your specification, quoted separately.
The video couldn’t start. Try playing it again or open the video directly.
Hear the German recordings
Unedited excerpts from the sample recordings we send to buyers, with the uncorrected machine transcript for the same stretch of audio.
Business role-play: a job centre appointment
- 00:13:54.2 - 00:13:57.9 Speaker 1 Na, das hört sich ja vielversprechend an. Und, ähm,
- 00:13:58.8 - 00:14:06.5 Speaker 1 sind diese Jobs denn auch, äh, normale Jobs oder ist das eher als Freelancer oder, oder wie muss man das verstehen?
- 00:14:07.8 - 00:14:26.6 Speaker 2 Also viele dieser Firmen sind aktiv, ähm, äh, mit Freelancern. Ähm, dennoch gibt es auch viele, die dann eher, ähm, auf ein Arbeitsverhältnis, ähm, wirklich, ähm, im traditionellen Sinne, ähm, ja suchen und auf der Aufschau, Ausschau sind.
- 00:14:26.8 - 00:14:40.0 Speaker 2 Und dadurch, ähm, gibt es halt dann diese ganzen, äh, positiven Mitbringsel wie, ähm, eine Versicherung, wie, ähm, das, ähm, Essensgeld und, ähm,
- 00:14:41.2 - 00:14:50.0 Speaker 2 ja unter anderem diese, diese speziellen Sachen, die halt bei einem festen Arbeitsvertrag meistens mit dabei sind.
Monologue: why a job becomes a label
- 00:03:32.7 - 00:03:34.5 Speaker 1 die Frage nach deinem Job.
- 00:03:35.1 - 00:03:37.4 Speaker 1 Dein Job ist irgendwie dein Etikett.
- 00:03:38.0 - 00:03:42.4 Speaker 1 Bist du ein Arzt? Wow, dann bekommst du wirklich Respekt. Oder bist du Künstler?
- 00:03:43.3 - 00:03:51.1 Speaker 1 Das zeugt dann von unglaublich viel Mut. Aber kannst du auch davon leben? Oder bist du in der Verwaltung? Aha, sicher, aber langweilig.
- 00:03:52.4 - 00:04:02.2 Speaker 1 Ich gebe zu, wir kategorisieren Menschen sofort anhand ihrer Tätigkeit, die sie einfach jeden Tag ausüben. Und deshalb ist der Druck, einen Traumjob zu haben, eben auch so immens hoch.
- 00:04:02.8 - 00:04:09.4 Speaker 1 Wenn dein Job dein Status ist, dann muss dein Job fantastisch sein, damit du am Ende auch fantastisch bist.
Want whole recordings? Ask for the free sample pack with complete conversations and monologues and their TXT and JSON transcripts. The same sample recordings, with their full transcripts, are on Hugging Face. The excerpts and transcripts are also on GitHub and Kaggle.
What is in the dataset
| Recording type | Recordings | Hours | Average |
|---|---|---|---|
| Conversation, business topicsTwo speakers role-play workplace situations such as job interviews, meetings, sales calls and customer service | 989 | 506.5 | 31 min |
| Conversation, open topicsTwo speakers talk freely about travel, food, music, work, social media and everyday life | 982 | 505.9 | 31 min |
| MonologueOne speaker talks on a topic of their choice, from history and careers to happiness | 1,013 | 521.8 | 31 min |
| Total | 2,984 | 1,534.2 |
Where the speakers are
| Country | Hours |
|---|---|
| Germany | 1,175 |
| Austria | 224 |
| Switzerland | 60 |
| Other or not stated | 75 |
Speakers' home regions include Berlin, Hamburg, Bremen, Hesse, Saxony, Bavaria, Baden-Württemberg, North Rhine-Westphalia, Rhineland-Palatinate and Schleswig-Holstein in Germany, Vienna, Styria and Carinthia in Austria, and Zürich in Switzerland.
What it is good for
- Training and testing German ASR (speech recognition) on unscripted speech, not read sentences.
- Speaker diarization: each speaker in a conversation has their own channel, so turns and overlaps can be read from the channels.
- Voice agents and turn-taking: real interruptions, backchannels and topic changes between two people.
Specifications
| Language | German |
|---|---|
| In stock | 1,534 hours, counted 18 November 2026 |
| Speakers | 58 registered contributors, 48 of whom also recorded monologues. Their conversation partners were not registered, so there are more than 58 distinct voices; that number is not counted. |
| Recorded | 25 January 2026 to 28 February 2026 |
| Audio | 48 kHz WAV. Conversations: stereo, one speaker per channel, plus one file per speaker. Monologues: mono. Bit depth is 16 or 24 depending on the recording. |
| Devices and setting | 96% recorded on a laptop or PC and about 4% on a phone, in quiet indoor settings. |
| Transcripts | Machine transcripts with speaker labels, segment start and end times and word-level timestamps, as TXT and JSON, for 1,533 of the 1,534 hours. Not human-reviewed. |
| Topics | More than 2,300 distinct topic labels, from job interviews and travel agency calls to food, music, AI and philosophy. |
| Licence | Commercial use under a written licence. Voice cloning is excluded. What to check in a speech data licence |
| Delivery | Download access within 24 hours of a signed order |
| Price | $35 per audio hour for all 1,534 hours ($53,690); a part of it is quoted |
Buying for a model? Request a written quote for the full set or the hours you need. Our guide to buying AI training data lists what to check in a sample before you sign.
Get the full German sample pack
Complete sample conversations and monologues with their TXT and JSON transcripts, sent to your inbox. If you already know how many hours you need, say so and we will send a quote with it.
Frequently asked questions
Is the German speech dataset available now?
Yes. 1,534 hours are in stock, counted in our recording database on 18 November 2026. You get download access within 24 hours of a signed order.
What does the German speech dataset contain?
1,971 two-person conversations (1,012 hours) and 1,013 monologues (522 hours), recorded between 25 January 2026 and 28 February 2026. More than 2,300 distinct topic labels, from job interviews and travel agency calls to food, music, AI and philosophy.
Who are the speakers?
58 registered contributors, 48 of whom also recorded monologues. Their conversation partners were not registered, so there are more than 58 distinct voices; that number is not counted. Speakers' home regions include Berlin, Hamburg, Bremen, Hesse, Saxony, Bavaria, Baden-Württemberg, North Rhine-Westphalia, Rhineland-Palatinate and Schleswig-Holstein in Germany, Vienna, Styria and Carinthia in Austria, and Zürich in Switzerland.
What transcripts come with it?
Machine transcripts with speaker labels, segment start and end times and word-level timestamps, as TXT and JSON, for 1,533 of the 1,534 hours. Not human-reviewed. If you need human-verified transcripts, we can quote a review pass.
What audio format is it delivered in?
48 kHz WAV. Conversations: stereo, one speaker per channel, plus one file per speaker. Monologues: mono. Bit depth is 16 or 24 depending on the recording.
Can I use the German recordings for commercial model training?
Yes, under a written commercial licence that names the recordings you buy and the permitted uses. Speech recognition, evaluation and speech understanding are the usual uses. Voice cloning is excluded.
What does the German speech dataset cost?
$35 per audio hour when you license all 1,534 hours, so $53,690 for the full dataset. For part of it, such as a fixed number of hours or one recording type, ask for a quote.
How is it different from free German corpora?
Common Voice and FLEURS are short sentences read aloud, and MLS is audiobooks read aloud. These recordings are unscripted: people talking to each other or one person talking freely for 31 minutes on average, which is closer to what assistants, meeting tools and call analytics have to transcribe.
What should I check when evaluating German speech data?
German compounds can create long words such as "Kundendienstmitarbeiter", and separable verbs move a prefix to the end of a clause. Spoken varieties also differ across Bavaria, Saxony and northern Germany. Check whether the sample's regions, compounds and transcript rules match the speech your system needs to process.