"MAI-Transcribe-1 claims the lowest word error rate across 25 languages on the FLEURS benchmark, averaging 3.8 percent, outperforming OpenAI's Whisper-large-v3."
"MAI-Voice-1 generates 60 seconds of natural-sounding audio in under one second on a single GPU and supports custom voice creation from a few seconds of sample audio."
"The MAI Superintelligence team, formed by Mustafa Suleyman, aims to deliver world-class models for Microsoft over the next five years, with these releases as the first evidence."
"MAI-Image-2 debuted at number three on the Arena.ai text-to-image leaderboard, showcasing its competitive capabilities in image generation."
Microsoft has released three new AI models—MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2—developed by the MAI Superintelligence team. These models are designed to operate independently of OpenAI's technology. MAI-Transcribe-1 boasts the lowest word error rate across 25 languages and outperforms competitors. MAI-Voice-1 generates natural-sounding audio quickly and allows for custom voice creation. MAI-Image-2 has already achieved a high ranking in text-to-image generation. This release signifies Microsoft's commitment to developing its own AI capabilities.
Read at TNW | Apps
Unable to calculate read time
Collection
[
|
...
]