Introducing Cohere Transcribe: a new state-of-the-art in open-source speech recognition

Cohere has introduced Transcribe, an open-source automatic speech recognition (ASR) model designed to advance the accuracy and usability of speech-to-text technology across various real-world applications. Transcribe's development focused on minimizing word error rate (WER) to ensure high transcription fidelity, achieving top-ranking accuracy on HuggingFace’s Open ASR Leaderboard. This conformer-based model supports 14 languages and is optimized for both GPU and local environments, making it suitable for practical deployment. It boasts exceptional throughput, converting minutes of audio into text within seconds, and has been praised for its performance in handling everyday speech with reliability and speed. Available for download on platforms like Hugging Face, Transcribe can be integrated via API for experimentation or deployed in production with Cohere’s Model Vault, offering flexible pricing and infrastructure management options.