Cohere Transcribe: Speech Recognition

Cohere has released an open-weight automatic speech recognition model that achieves a 5.4% word error rate, a level considered low enough to replace existing speech APIs in production pipelines. The model is available as Cohere Transcribe, a speech-recognition tool. Speechify’s Windows application also employs local models for transcription and dictation.

本文综合以下全部报道写成，而非依据任何单一来源。