Kõnele is an Android app that offers speech-to-text user interfaces and services to other apps. Many apps contain a text area or a text field (e.g. a search bar) that can be edited using the input method editor (IME), aka the virtual keyboard. Kõnele provides an IME that allows text to be created via speaking. Many apps (e.g. intelligent assistants, keyboard apps, navigation apps) also contain a microphone button that is linked to either the standard Android speech recognition activity or the standard Android speech recognition service. Kõnele implements both the activity and the service.
In the background, Kõnele uses two speech recognition servers. One supports grammar-based speech recognition, the other supports continuous full-duplex speech recognition. Both servers focus on Estonian speech recognition, but both are based on a fully open-source stack that makes them easy to deploy and customize for other languages.
The main goals of the Kõnele project are to provide Estonian speech recognition on the Android platform, and to provide grammar-based speech recognition for voice command applications. Kõnele is an open, configurable and powerful alternative to the speech recognition apps and services currently available on Android.
Kõnele provides two user interface components that can use any speech recognition service/language available on the device:
- the voice search panel can be called from other apps (typically by pressing a microphone button), and returns the recognition results to the app; in case the panel is opened via the Kõnele launcher icon, custom rewrite rules allow one to define which app is launched to interpret the recognition results (by default, the results are used to perform a web search using the default browser);
- the IME can be used to dictate into any text field in any app, edit the result with swipe commands, and execute user-defined editor commands (e.g. “select”, “copy”, “replace”, “undo”, “move cursor to next regex match”)
Kõnele also provides two speech recognition services that are pre-configured to use Estonian speech recognition:
- a continuous full-duplex service supports audio input of unlimited length where dictation results are returned already while dictating;
- a grammar-based speech recognition service supports English and Estonian voice actions (e.g. alarm clock, unit converter, address search), for the list of existing grammars see http://kaljurand.github.io/Grammars/.
The services can be used via Kõnele’s own user interface components or by external apps like Arvutaja.
Kõnele requires very few permissions (only the access to the microphone is essential), see the details in the manifest.
Kõnele is free and open source software. Visit the Kõnele GitHub page for the source code, bug reporting, etc. The information and downloadable APK-packages for all the released versions are on the Releases page.
“Kõnele” is the Estonian word for “to speak” (imperative form).
Kõnele is part of the project that was awarded Estonian Language Deed (Keeletegu) for the year 2011. See more here.