Skip to content
Speech to text, built for real audio

German speech to text for voice agents

PyAI Hear turns spoken German into text while the speaker is mid sentence, whether the voice comes from a car dashboard, a browser microphone or a support line. The first partial arrives in about 200 ms, so your assistant can plan a reply early.

~200 ms first partialSie and du speakersSwiss and Austrian accents$0.001/min

Explore first. Get a sandbox key when you are ready to build.

PyAI in motionInteractive example
Sample: a 17 second German in-car assistant session, transcribed live
0:00FahrerNavigiere bitte zur nächsten Ladesäule an der A9 Richtung München.
0:05AssistentDie nächste Ladesäule liegt in 14 Kilometern bei Ingolstadt Nord. Soll ich die Route starten?
0:10FahrerJa, und ruf danach meine Werkstatt an.
0:13AssistentRoute gestartet. Ich melde mich, sobald wir ankommen.
Drop an audio file, or record a clip

Try it free: 3 transcriptions, up to 3 minutes each, no account needed. After that, create a free account with $50 in credit.

  1. 1Upload or recordA file or a mic clip, up to 3 minutes
  2. 2Hear transcribesSpeech recognition tuned for real audio
  3. 3Read it hereThe transcript shows right on this page

A sample of how Hear returns German voice agent transcripts, with per segment timestamps and speaker labels.

Live German transcription for any voice assistant

A German speech to text API for voice agents has to follow a speaker in real time, not process a recording later. PyAI Hear takes audio as a stream from a mobile app, a web page, an embedded device or a phone line and answers with partial text in about 200 ms. Live German transcription from Hear can feed assistant logic you write yourself, or PyAI Omni can run listening, reasoning and speaking for you in one session. Every accuracy figure here was measured on English, so record German speakers from your own product and test before launch. The same engine also works as a German speech to text generator for the transcripts your team keeps after each session.

Key takeaways

  • First partial: roughly 200 ms after the speaker's audio starts, in any channel.
  • Accuracy: PyAI measured 4.76% WER on rough 8 kHz audio and 1.59% on clean speech, in English, with no separate German score.
  • Compounds: words like Kundennummer or Ladesäule are single German words, so register up to 5 custom vocabulary terms for the ones your users say often.
  • Price: $0.001/min, billed by the whole minute, with no platform fee and free empty transcripts.
  • Omni timing: end of turn detected about 196 ms after speech stops, and voice to voice near 390 ms median.
  • Migration: whisper-1 compatible, so an existing client moves over by editing the base URL.

How German speech becomes text in a live session

Real time German speech to text means Hear reads the audio while the person is still talking, so the assistant never waits for a finished clip.

Capture the voice

The speaker talks into an app, a browser microphone, a car head unit or a phone line, and your client collects PCM16 audio.

Stream it over WebSocket

Your client sends frames of about 20 ms each down one WebSocket per session, with the language set to German.

Read the partials

Hear returns a first German partial in about 200 ms and revises the text as the rest of the sentence arrives.

Let the agent act

Your assistant, or Omni with its turn detector, takes the final German text and replies, searches or triggers the next step.

What Hear handles in live German speech

Spoken German brings long compounds, two registers of address and strong regional sound. Each point below applies whether the voice reaches you from a dashboard, a kiosk or a laptop.

Partials before the verb lands

German often places the main verb at the end of a clause. Hear streams partial text in about 200 ms and revises it once the verb arrives, so build your intent logic to act on final text and use partials for early hints.

Long compound words

A speaker who says Kfz-Haftpflichtversicherung or Kundennummer uses one long word, and spelling matters when your assistant matches terms. Register up to 5 custom vocabulary terms for compounds your users repeat.

Swiss, Austrian and Bavarian sound

Speakers from Zurich, Vienna and Munich sound different, and PyAI publishes no regional German score. Send samples from each region through a sandbox key, which never bills, and compare the output yourself.

Car cabins and busy stations

Road hum, fans and station announcements sit under the voice. PyAI reports 4.76% word error rate on noisy and accented audio, measured on English, so test German recordings from your real environment.

Digits and amounts

With numerals handling on, postcodes, prices and order numbers return as digits, and inside Omni the turn detector waits while a speaker is still reading out a long number.

Speaker turns and timestamps

Verbose JSON marks each speaker turn with timestamps, which helps when you review how an assistant handled Sie and du or audit a session later.

Accuracy and speed

Hear's word error rates are PyAI's own benchmarks, measured on English and reported at industry parity. No German error rate is claimed. Omni timing comes from a full Omni session.

1.59%Word error rate, clean audio
4.76%Word error rate, 8 kHz phone and noisy
~200 msTo the first partial transcript
92 to 247xReal time, batch throughput

Where German live transcription fits a voice agent

From Hamburg to Graz, German speakers talk to assistants in cars, apps, shops and factories. German voice assistant speech to text has to work wherever the speaker is, and these six jobs show the spread.

In-car assistant

Take navigation and charging requests on the Autobahn hands free, and keep the driver's eyes on the road while the agent plans the route.

In-app banking help

Let a customer ask about a transfer or a card limit inside a mobile banking app, with amounts returned as digits.

Warehouse voice logging

Workers in a Lager call out bin numbers and quantities while their hands stay on the pallet, and the system writes each entry.

Smart home control

Turn spoken German into commands for heating, blinds and lights on a connected home hub.

Website support bot

Add a microphone button to a shop or utility site so visitors can describe a problem instead of typing it.

Customer support phone line

Answer a service number with an agent that captures account details and passes harder cases to staff.

Formats, limits and output

What a German live transcription request sends and receives
InputA live WebSocket stream of PCM16 German audio from any client, or a file of up to 25 MB. A whisper-1 client works after you swap the base URL.
OutputGerman partial transcripts while the person speaks, then final text. Choose text, JSON or verbose JSON with timestamps and speaker turns.
ModesSync file call for short clips, live streaming for apps, devices and voice sessions, and background jobs for long recordings that run 92 to 247 times real time.
TuningSet the language to German, switch on numerals handling for postcodes and prices, and add up to 5 custom vocabulary terms for brands and place names.
Billing$0.001/min of audio, rounded to the whole minute. An empty transcript costs nothing, and a busy engine returns a retry, not a charge.

Connect German speech to text to your assistant

Hear follows the OpenAI transcription format, so a German assistant stack usually changes only the base URL. Docs and examples live at PyAI SDKs.

SDK

On your server, the SDK opens one stream per German session and hands partial and final text to your handler as soon as Hear produces it.

MCP

Through the MCP server, an agent framework can call Hear as a tool, for example to transcribe a German voice note before deciding what to do.

CLI

In a terminal, the CLI streams a German sample and prints partials as they arrive, a fast way to check vocabulary terms before release.

Where Hear fits

PyAI is voice infrastructure. Hear, Speak and Omni share one stack in one region, so recognition, turn detection, reasoning and speech stay in one session for a German assistant. Hear is not a dialog builder and does not write your assistant's replies. Use Hear inside an agent you already run, or choose Omni to handle the whole exchange. Figures are published on PyAI benchmarks.

Get a sandbox key → $50 free credit. No credit card. Sandbox keys never bill.

German speech to text for voice agents FAQ

How fast does Hear return German text?

Hear sends the first German partial about 200 ms after audio starts and refines it while the person keeps talking. In a PyAI Omni session, the end of a turn is detected about 196 ms after speech stops, and the full loop from voice in to voice out runs near 390 ms median.

Does PyAI publish a German accuracy score?

No. PyAI reports 1.59% word error rate on clean audio and 4.76% on 8 kHz phone, accented and noisy audio, and both results were measured on English at industry parity. No German result is claimed, so record speakers from your own app or device and compare transcripts before launch.

Can I plug Hear into an assistant I already built?

Yes. Hear is whisper-1 compatible, so your client usually changes only its base URL. Stream PCM16 frames of about 20 ms over a WebSocket with the language set to German, then pass partial and final text into the assistant logic you already run, in any app or device.

What does German speech to text cost with Hear?

Hear charges $0.001 per minute of German audio, billed by the whole minute, with no platform fee, and an empty transcript costs nothing. New accounts receive $50 in credit without a credit card, and sandbox keys never bill, so complete test sessions are free.

Should I use Hear or Omni for a German assistant?

Pick Hear when you already run your own reasoning and speech output and only need German text. Pick Omni when you want listening, turn detection, reasoning and speaking in one session, with end of turn around 196 ms and voice to voice near 390 ms median, and a detector that waits through unfinished thoughts.

Prices checked November 2026. Hear figures from pyai.com/benchmarks.

Start transcribing with PyAI Hear

Spin up a sandbox key and turn calls, meetings and recordings into clean text in minutes. No credit card, and $50 in free credit.

Get a sandbox key →