This project is dedicated to my Dad, who overcomes his own hearing loss everyday and inspires me to write code that can improve accessibility in tech.
The latest version of SFSpeechRecognizer allows offline speech-to-text, amongst other new features. This project builds on the work of other devs to create a demo of how closed captioning could work with video playback.
The benefits of offline transcription is no more data usage, no more need for a data connection at all, and unlimited usage (Apple restricted transcription to 1-minute at a time and for limited hits per day). The downside is that it may not be as accurate nor have the benefit of a constantly learning cloud.
The potential applications, however, are huge for Accessibility purposes. Combined with CoreML/CreateML Speech capabilities, it can potentially open up a number of use cases.
- Danny Bolella - Initial work - TheDB
This project is licensed under the MIT License - see the LICENSE.md file for details
UI is credited to: https://github.com/ChrisMash/AVPlayer-SwiftUI
Buffer is based on the works by: https://github.com/peacemoon/SFSpeechRecognizerRealtimeVideoCaptioning https://github.com/zats/SpeechRecognition https://gist.github.com/omarojo/03d08165a1a7962cb30c17ec01f809a3
Which were all built on top of: https://developer.apple.com/library/archive/samplecode/AudioTapProcessor/Introduction/Intro.html#//apple_ref/doc/uid/DTS40012324-Intro-DontLinkElementID_2
Phew.