android

Android Speech Recognition Example

This example shows how to use ORT to do speech recognition using the Whisper model.

The application lets the user make an audio recording, then recognizes the speech from that recording and displays a transcript.

Set up

Prerequisites

See the general prerequisites here.

Optionally, you will need to be able to record audio, either on an emulator or a device. In the example, a pre-recorded audio is provided.

This example was developed with Android Studio Flamingo 2022.2.1 Patch 1. It is recommended to use that version or a newer one.

Generate the model

This is optional. Currently a usable model file is checked in.

Follow the instructions to generate the model with Olive here.

• The model name is openai/whisper-tiny.en.
• The configuration used is for CPU and INT8.
• Use the --no_audio_decoder option as we don't need the audio decoder in the model in this application.

Copy the model to app/src/main/res/raw/whisper_cpu_int8_model.onnx.

Note: The checked in model was generated with Olive 0.3.1.

Build and run

Open this directory in Android Studio to build and run the example.