New Open Audio Models 🤗 | Recap with Jeff

This video covers the latest wave of open audio tooling, from Mistral's Voxtral 4B text-to-speech model to Cohere Transcribe for speech recognition and the Hugging Face infrastructure used to run large-scale transcription workflows. It walks through live demos, browser-based transcription with Transformers.js, and a practical UV-script pipeline built on storage buckets, HF Mount, and HF Jobs. If you're building speech apps or batch transcription systems, this is a fast overview of the current open stack. --- Demo Links 👉 Voxtral TTS: https://huggingface.co/spaces/mistralai/voxtral-tts-demo 👉 Cohere Transcribe: https://huggingface.co/spaces/CohereLabs/Cohere-Transcribe-WebGPU 👉 UV scripts for transcription: https://huggingface.co/datasets/uv-scripts/transcription --- 🤓 Topics Covered - Voxtral 4B text-to-speech - Cohere Transcribe speech-to-text - Hugging Face audio pipelines --- ⏱️ Timestamps 0:00 Open audio models and demos 2:44 What Hugging Face storage buckets are 3:39 How HF Mount works 4:02 HF Jobs and wrap-up