OpenAI announced on July 29 the release of two new transcription AI models, "GPT Transcribe" and "GPT Live Transcribe," made available via API. GPT Transcribe, designed for pre-recorded files, delivers high-accuracy asynchronous processing at $0.0045 per minute, slashing error rates by more than half compared to Whisper. GPT Live Transcribe, optimized for real-time applications at $0.017 per minute, features noise resilience capable of accurately transcribing speech even amid musical instrument playback. Both models demonstrate improved comprehension of context and specialized terminology, with anticipated applications spanning meeting minutes generation, multilingual communication, and accessibility support.