Repository navigation
Voice control
regis57 edited this page Oct 9, 2026
·
1 revision
Beranda listens with the microphone of the device that shows it (tablet or computer), never the Pi's.
- Settings page, box 11 · Voice control: tick Show the microphone button, Save.
- Open Beranda in Google Chrome. Edge offers speech recognition but its service fails ("network"); Firefox and Safari do not offer it to web pages.
- Use the secure address:
https://instead ofhttp://, same address and port. The first time, accept the "not private" warning (Advanced → Continue): the certificate is made by Beranda. - Tap the round microphone, allow it, speak: "what's the weather", "play France Inter", "next station"...
The other way round step 3: a one-time Chrome setting,
chrome://flags/#unsafely-treat-insecure-origin-as-secure with Beranda's http:// address.
In the same browser, settings page → Voice control → Test the microphone on this device.
| Message / code | Meaning | What to do |
|---|---|---|
insecure |
the page is http | open it with https://
|
not-allowed |
microphone blocked | address bar icon → Microphone → Allow; check the system's privacy settings |
audio-capture |
no microphone | plug one in (headset, webcam) |
network |
the browser's speech service is unreachable | use Google Chrome; check the internet connection |
unsupported |
the browser cannot listen | use Google Chrome |
The screen also reports its result to the Pi: it shows in the diagnostic file.
Privacy: Chrome sends the sound to Google to turn it into text; Beranda only gets the text and keeps nothing. Full guide: docs/VOICE.md.
Beranda · MIT · code · issues · buy me a coffee