We are looking for Zurich voices

The honest reason is not that Swiss German speech synthesis does not exist — by now it does. The reason is licensing. Every Zurich speech corpus we found is released for research and not for a product. Anyone who needs real, cleanly licensed, consented Zurich German has to record it themselves. On top of that there is what synthetic voices are bad at anyway: pace, mumbling, hesitation, the difference between two people from the same neighbourhood.

Ordered by commitment, smallest first — and the smallest is worth the most to us. Each one says what already exists in the product for it today.

• 01

  Read our lines

  You grew up in or around Zurich. Twenty minutes, once or now and then.

  You read sentences we wrote and say, for each one: that is how it is said, that is not, or nobody says that at all.

  All 390 lines from everyday and care scenes are machine-checked for Zurich forms. An ear from here hears what no rule can: a sentence that is correct and that still nobody says. And these are sentences somebody says to a frightened person at half past six in the morning.

  As it stands

  Every situation is open to read. Point at the line that is wrong through the feedback window on the page.

  Write to us about thisTake a look →
• 02

  Record your voice

  Zurich dialect as a first language — any age, any corner of the canton.

  A few minutes of speech on topics we suggest, with a consent form you read before you agree to anything.

  Training on many different speakers is the best-evidenced listening method there is — and it cannot be built out of a Standard German computer voice reading dialect. The listening lab is built on exactly this.

  As it stands

  What the consent looks like is further down this page. The first voices go into the listening lab.

  Write to us about thisTake a look →
• 03

  Be a speaking partner

  You speak dialect and have half an hour now and then.

  Join a speaking round: a small group, a topic, a time. You simply talk the way you talk.

  The speaking loop in Heidi deliberately ends not in a score but in a room with other people — because that is where a language is spoken. A score is what a product offers instead of that.

  As it stands

  Built: rounds, topics, sign-ups and attendance. Open a round or sign up for one.

  Write to us about thisTake a look →
• 04

  Teach

  You teach German or dialect, freelance or at a language school.

  Tell us what you offer and what your learners would need here. We are not building a marketplace against you — we are building the material underneath.

  A language school has people in a room and knows where learners actually fail. Heidi has checked dialect material and the exercises generated from it. The two fit together.

  As it stands

  No tutor profiles or matching yet. What there is, is a conversation.

  Write to us about this

As a learner you can already open a group and pass on a link, and you can sign up for a speaking round. Both work from zero members — they only get better when somebody above says yes.

SpeakingGroups

You stay in control. We tell you in advance what the recording will be used for, you can withdraw it, and consent for the product is not the same as consent for research. We assume you do not want the second unless you say so explicitly.