Part of First Steps in the Playground
Cartesia Speech-to-Text Playground
CartesiaReal-time voice AI models for developersAbout this screen
A transcript with timings
Cartesia's speech-to-text playground shows a sample call transcribed by ink-2: a timestamp (0:00.545) next to the text, which handles an account number and an amount exactly, followed by small chips for time and latency and a "Copy" action. A model select (ink-2) and "Build with Ink" sit above it.
Turn detection as sliders
The right panel sets language hints (with "Try with ink-preview"), keyterms in a text area, and turn detection: Balanced, Patient or Responsive presets, then sliders for start_threshold, eager_end_threshold, end_threshold and end_timeout_ms, each explained with labels such as "Sooner" and "More certain".
Three ways to test
At the bottom, "Upload", "Speak" and a green "Try next" button let you switch from the sample to your own audio or microphone.






