Apple SpeechTranscriber

Apple SpeechTranscriber is an on-device, long-form speech-to-text transcription module, part of the new SpeechAnalyzer framework in iOS 26 and macOS Tahoe, engineered for speed and accuracy.

Apple SpeechTranscriber, introduced with the SpeechAnalyzer framework in iOS 26 and macOS Tahoe, provides on-device, long-form speech-to-text transcription. This module is specifically designed for sustained audio over minutes or hours, handling lectures, podcasts, and multi-speaker meetings with high accuracy. Benchmarks show it can be over 2x faster than Whisper's Large V3 Turbo model, processing a 34-minute, 7GB video file in just 45 seconds. It operates entirely on-device, ensuring privacy and eliminating network overhead, and supports a growing list of languages including English, Chinese, French, German, and more. While it currently doesn't support custom vocabulary for long-form transcription, it represents Apple's modern, recommended path for advanced speech-to-text capabilities.

https://developer.apple.com/documentation/speech/speechtranscriber