Heavily inspired from https://github.com/zats/SpeechRecognition. Goal: After seeing Google i/o 2019 where google introduces Live Transcription of videos, I'm asking myself if we can do the same in iOS. The simplest option is to use SFSpeechRecognizer to tap into audio stream of a video and transcribe the audio buffer. After some googling I found something that I can put together to a simple demo. Naturally Google's solution is a lot superior to what iOS can: clean API, offline, multiple languages etc... Hopefully Apple will introduce something similar in their next WWDC 2019.

Next step: Swift version of MTAudioProcessingTap

Youtube Demo

Some related links to the topic:

• https://stackoverflow.com/questions/53636698/mtaudioprocessingtap-exc-bad-access-doesnt-always-fire-the-finalize-callback#53637508
• https://github.com/gchilds/MTAudioProcessingTap-in-Swift
• https://gist.github.com/omarojo/03d08165a1a7962cb30c17ec01f809a3