Mistral AI Releases Open-Source Speech Model Voxtral Transcribe 2
Paris-based startup Mistral AI launched Voxtral Transcribe 2 on Wednesday, a pair of open-source speech-to-text models designed to operate on-device for a fraction of the cost of existing solutions, according to VentureBeat. The company claims the new models transcribe audio faster and more accurately than competitors, marking a significant development in the competitive voice AI landscape.
The Voxtral Transcribe 2 models are designed to process sensitive audio without transmitting it to remote servers, a feature that sets them apart from offerings by American tech giants, VentureBeat reported. This on-device processing is particularly appealing to enterprise customers who require automated customer service and real-time translation capabilities.
In other tech news, Nintendo released a new Virtual Boy peripheral for the Switch, The Verge reported on Tuesday. While the original Virtual Boy was a commercial and critical failure, its infamy has contributed to its enduring appeal. Charles Pulliam-Moore of The Verge noted that the new peripheral is "a well-built piece of nostalgia, but its games are just too stuck in the past."
Meanwhile, in Washington, D.C., a new AI math startup has reportedly cracked four previously unsolved problems, according to Wired. The breakthrough occurred during a math conference in the city. Separately, Best Buy is offering the 128-GB Pixel 9a for $399, a $100 discount, Wired reported. The phone is available in Iris, Obsidian, Peony, and Porcelain.
Discussion
AI Experts & Community
Be the first to comment