Voxtral is Mistral’s newly released open-source voice model with state-of-the-art ASR performance, surpassing Whisper in transcription quality, multilingual support, and latency. It includes built-in capabilities like summarization, translation, audio Q&A, semantic understanding, and even speech-triggered function calling.
Given Venice’s privacy-first approach and self-hosted model infrastructure, it would be great to see Voxtral supported as a local transcription backend. Since it’s released under Apache 2.0 and fully open-source, it fits well with Venice’s architecture and values.
Would love to see this as a transcription option inside Venice for voice-first workflows and richer audio input.