Speech Android
An on-device speech SDK for Android offering streaming speech recognition, text-to-speech, voice activity detection, and noise cancellation, built on ONNX Runtime with Qualcomm NNAPI acceleration.
Speech Android is an on-device speech SDK for Android, built on ONNX Runtime and a shared speech-core engine, providing low-memory streaming speech recognition covering 25 European languages, with an optional larger model for broader coverage, text-to-speech, voice activity detection, and noise cancellation, all running locally with no cloud APIs and no data leaving the device. This repository is the Android packaging layer: a Kotlin SDK, a JNI bridge, and a demo app, while the underlying C++ engine and ONNX model wrappers live in a separate speech-core repository, pulled in as a Git submodule; an Apple counterpart SDK and a Linux/embedded build, including automotive platforms, are also available.
A separate control-demo app assembles the full pipeline into a working offline voice agent: voice-activity detection feeds streaming speech-to-text, which feeds a small model that decides on-device actions, which are executed and read back through a text-to-speech reply, reporting per-stage latency while running in about 1.2 GB of RAM. The Android documentation lists demo APKs and supported models with their download size, peak memory, and language coverage.