simonwillison.net
New audio models from OpenAI, but how much can we rely on them?
OpenAI announced several new audio-related API features today, for both text-to-speech and speech-to-text. They’re very promising new models, but they appear to suffer from the ever-present risk of ac...