SpeechRecognitionIOS

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.

Prerequisites

• 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

QuickStart

Ensure you have pods installed on your machine, to check use pod --version

❗ Do not run this command pod install because 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_KEY with the API key obtained above.
• Build and run the app.
  • Use command cmd + R to 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.

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