Cohere Unveils ‘Transcribe’: A New Benchmark in Open-Source Speech Recognition


Cohere has launched Transcribe, a leading automatic speech recognition (ASR) model that’s open-source and primed for practical, everyday use. The model boasts top-tier accuracy on Hugging Face’s leaderboard, shining with the lowest word error rate (WER) among contenders like Whisper and ElevenLabs Scribe. Designed for flexibility, Transcribe efficiently processes speech into text across 14 languages, benefiting applications from meeting transcriptions to real-time customer support. For businesses, it’s a dream—easy to integrate, whether locally or via Cohere’s Model Vault for hassle-free deployment. Cohere sets a new standard for enterprise speech intelligence while planning to weave Transcribe more tightly with their AI orchestration platform, North. Available now on Hugging Face, it’s a superb blend of accuracy, efficiency, and open access.
Read more at Cohere…