Tagged with

3 articles found

Soprano-80M: The 2000x TTS Speedup That Makes Real-Time Voice AI Finally Real, And Exposes Everything Broken About the Field

A deep dive into Soprano-80M’s breakthrough <15ms latency and 2000x real-time generation, the architectural choices that enable it, and why its stability issues reveal uncomfortable truths about modern TTS development.

#low-latency#real-time-ai#text-to-speech...

microsoft

VibeVoice’s Uncanny Valley: Microsoft’s 90-Minute AI Podcasts Sound Too Human

Microsoft’s VibeVoice model can generate 90-minute multi-speaker podcasts that blur the line between synthetic and human speech, raising ethical questions about audio deepfakes.

#microsoft#text-to-speech#voice-synthesis

Microsoft’s VibeVoice 1.5B Can Generate 90-Minute Podcasts With 4 Voices

Microsoft’s new open-source TTS model can synthesize feature-length audio with multiple speakers, but comes with audible disclaimers and watermarking to prevent misuse.

#AI#microsoft#text-to-speech...