Text to speech
German text to speech
Our German voices are recorded in German, not an English model with an accent bolted on. Play them below, then read how the model handles compounds, umlauts and the way Germans actually say numbers.
Listen
5 native German voices
Every clip on this page was generated by Kugel 3 in German. Nothing here is a dubbed English sample.
Noah
Native recording
Olivia
Native recording
Gustav
Native recording
Samantha
Native recording
Nick
Native recording
Details
What we tuned for this language
Umlauts and compound nouns
Words like Lieferantenrechnung or Kfz-Zulassungsstelle are single tokens to a reader and a stumbling block to a model trained mostly on English. Our German training data is German, so long compounds keep one stress pattern instead of breaking into pieces.
Numbers, dates and IBANs
German inverts the digits of two-digit numbers, writes dates day-first and groups IBANs in fours. Our normalizer expands them the German way before the voice ever sees them, so 21,5 % is read as a percentage and not as a date.
Your own vocabulary
Product names, medical terms and Swiss or Austrian place names are where generic German TTS gives itself away. The pronunciation dictionary lets you pin those words once and have every voice say them the same way.
FAQ
German text to speech: common questions
How many German voices does KugelAudio have?
This page plays 5 personas recorded natively in German. The full voice catalogue is larger, and the same Kugel 3 model speaks 26 languages, so one voice can move between German and English inside a single conversation.
Does it pronounce umlauts and the sharp s correctly?
Yes. ä, ö, ü and ß are part of the German training data rather than characters mapped onto the closest English sound. For names and jargon that even German speakers disagree on, the pronunciation dictionary fixes the reading for your whole account.
Where is German speech generated?
On our own GPUs in the EU. KugelAudio is a German GmbH, so there is no US parent and no CLOUD Act exposure, and the same models can be deployed inside your own data centre behind your firewall.
Put this voice in your product
One API for streaming speech, a pronunciation dictionary for your own vocabulary, and hosting inside the EU or inside your own data centre.