Mistral launched Voxtral, an open automatic speech recognition software, to offer superior accuracy and semantic understanding at lower costs than competitors. Traditional ASR choices involved high-error open-source models or expensive proprietary models. Voxtral starts at $0.001 per minute and reportedly outperforms leading models like OpenAI's Whisper and GPT-4o mini-transcribe. It supports multilingual transcription, allowing inputs of up to 32,000 tokens, and can automatically detect languages. Additionally, it includes functionality for generating audio summaries and answering questions, showing significant advancements in ASR capabilities.
Voxtral bridges the gap between high error open-source models and costly proprietary models, offering state-of-the-art accuracy and semantic understanding at competitive pricing.
Voxtral starts at $0.001 per minute, providing better word error rates than competitors and achieving superior results in transcription tasks across multiple languages.
Collection
[
|
...
]