This project demonstrates the recording and transcribing of speech from one language to another (e.g. English to Yoruba) using the Google Cloud Speech API and Apple on-device Speech library.
- An API key for the Cloud Speech API (See the docs to learn more)
- Create a project (or use an existing one) in the Cloud Console
- Enable billing and the Speech API
- Add your iOS bundle identifier to project (e.g com.SpeechRecognition.joshuvi.SpeechRecognitionIOS in my case)
- Create an API key, and save this for later
- An OSX machine or emulator
- Xcode 13 or later - tested with Xcode 13.2
- Cocoapods version 1.0 or later
Ensure you have pods installed on your machine, to check use
pod --version
❗ Do not run this command
pod installbecause pods already exist, see Podfile & Podfile.lock
- Clone this repo
git clone https://github.com/Josh-Uvi/SpeechRecognitionIOS.git - CD into this directory
SpeechRecognitionIOS - Open SpeechRecognitionIOS.xcworkspace with Xcode
- In SpeechRecognitionService.swift, replace
YOUR_API_KEYwith the API key obtained above. - Build and run the app.
- Use command
cmd + Rto build and run on iOS simulator/device - Alternatively, press the play button from xcode to start the build process
- The build process will start up and run on Xcode simulator if you have one configured or on a device if connected
- Say a few words and your speech will be recognized.
- Use command
DONE
- Apple Speech on-device library
- Google Speech-to-text API
- Transcribe live audio recording
- Offline capability available for iOS 13.0 or later
TODO
- Transcribe pre-recorded audio file
- Display transcribed text on an OLED device
- Do performance and load stress test if possible
NICE TO HAVE
- CI/CD pipelines
- Self-hosted Speech Recognition server/engine, using docker technology and Kaldi Speech Recognition Server.
- Run Python program in the app, using Python-Apple-support