Speech is an search online voice recorder that checks and helps you improve your pronunciation.

Try it online at: https://speech.sege.dev


How it works?

Under the hood, Speech uses Vosk – the speech recognition toolkit, to check your voice recording and figure out what you are trying to say.

The audio recording is done using the Web Audio API, and encoded into WAV format using the WebAudioRecorder.js library.

How to run it locally?

First, you need to download a Vosk model at https://alphacephei.com/vosk/models and extract it to the model folder in the source directory.

Also, you will need to have libvosk.so or libvosk.dylib in your computer’s library path. You can download it from the vosk-api repository, and copy the library file to somewhere like /usr/local/lib.

Then you are ready to run the project:

$ cargo run

The frontend will be built automatically using Parcel, and the server will be started at http://localhost:3000.


View Github