Search
On this Thursday, enterprise AI firm Cohere unveiled its inaugural open-source speech model, Transcribe. This automatic speech recognition (ASR) model, boasting around 2 billion parameters, is designed as a lightweight solution capable of running on consumer-grade GPUs. It offers support for 14 languages and can transcribe 525 minutes of audio per minute, rendering it ideal for speech-to-text conversion and content analysis applications, especially catering to the self-hosting requirements of enterprises. On the Hugging Face Open ASR leaderboard, it achieves an average word error rate (WER) of 5.42%, surpassing rivals like Zoom Scribe v1 and IBM Granite 4.0 1B. Cohere intends to incorporate Transcribe into the North platform, furnishing post-transcription workflow assistance, including automated archiving, intelligent summaries, and other enterprise-grade services.
-
C114 Communication Network -
Communication Home
