Hi! I vibecoded this with my daughter. It uses the SpeechAnalyzer framework in iOS 26, so all speech-to-text processing happens entirely on-device. One interesting thing we noticed is that real-time transcription is noticeably faster in English than in other languages. My guess is that the English model benefits from more training data compared to Spanish, French, German, etc. It works slightly better with earphones, but it’s surprisingly good even if you just point your phone away from you and speak in a normal voice. The transcription runs a few seconds behind, but it can handle continuous speech quite well, and you quickly get used to the delay.

You can use the app to communicate at a distance, during presentations, when learning a language, or if you have difficulty hearing. Some unexpected use cases we’ve found: ad hoc captions when the TV subtitles are already set to another language, a backup when people can’t hear me on Zoom calls, captions while listening to the radio, or trying to figure out what Bad Bunny is singing about. The haptics help you "feel" the words that are displayed when your screen is pointed away from you.

Let me know if you find other interesting uses for this app. Also, please let me know how it works on an iPad (I'm too cheap to buy one and you can't speech to text in the emulator).