Concepts
Speech Input
Browser speech support exists in the runtime, but it should be understood as one input layer over the same local-first execution model.
Need the shortest path?
Use Quick Start when you want the smallest working route → scan → graph → run flow first.
Something feels off?
Inspect scan output, graph state, active route/context, and dispatch in a practical order.
Need the full surface?
Browse all runtime methods, hooks, extension points, and lifecycle APIs.
Cinder includes a browser-first speech layer, but speech is only one way to enter the runtime.
What speech does
Speech ultimately feeds the same command pipeline used by text input.
That means speech still ends up at:
- local resolution
- optional fallback
- dispatch
- reflex output
Main speech APIs
Cinder.listen()Cinder.stopListening()Cinder.speechStatus()Cinder.onTranscript()Cinder.onSpeechError()
Honest expectation
Browser speech depends on Web Speech API availability. For browsers with none at all (Firefox, older Safari), setting speech.engine: "auto" with a configured fallback endpoint enables a server-side recording-and-transcription path instead - it reuses the same endpoint text tier-3 fallback already uses, not a separate config surface. It can't stream interim results the way native SpeechRecognition can. See Using Speech for setup.
Why text still matters
Text commands remain first-class because they are useful for:
- unsupported browsers
- testing and debugging
- command palettes
- explicit application command bars
- reproducible integration checks
Practical usage pattern
A common flow is:
- start listening
- observe final transcripts
- auto-run them or handle them manually
- fall back to text whenever the browser does not support speech cleanly
Related pages
Continue reading