Apple’s Latest AI-Enhanced Transcription Surpasses OpenAI’s Whisper
It seems Apple is implementing various behind-the-scenes AI advancements for iOS 26 and macOS Tahoe. While many features build on existing capabilities, the company will also introduce a chatbot-like interaction for users wishing to engage with Apple Intelligence privately via the Shortcuts app, alongside a highly effective speech API that excels beyond OpenAI’s Whisper.
At least, that’s according to MacStories’ John Voorhees in his practical review. He enlisted his son’s help to create Yap, a “basic command-line tool that takes audio and video files as input and provides SRT- and TXT-formatted transcripts.”
During his experiments, he successfully transcribed a 7GB 4K video version of a 34-minute AppStories podcast episode in just 45 seconds and produced an SRT file. After testing other AI transcription models, Apple’s performance exceeded that of all competitors:
Yap: 45 seconds.
MacWhisper (Large V3 Turbo): 1 minute and 41 seconds.
VidCap: 1 minute and 55 seconds.
MacWhisper (Large V2): 3 minutes and 55 seconds.
Although Apple’s AI transcription model is not without flaws and struggled with certain last names and terms like “AppStories,” Voorhees was taken aback by Yap’s efficiency, being 55% quicker than OpenAI’s top model while maintaining equivalent transcription quality.
That being said, when iOS 26 and macOS Tahoe launch, expect to see new applications leveraging Apple’s cutting-edge AI models to analyze spoken language and transcribe information. As these models are freely available for developers, they are set to enhance the audio transcription market.
At present, these features are confined to developers utilizing the beta iterations of iOS 26, macOS Tahoe, and Xcode 26.
