Skip to content
Inworld AI· inworld·· 3 d agoPickAI score62

Inworld open-sources Voice Isolator and SAVA speech models and publishes TTS-2 technical report

Today we’re open-sourcing Inworld's Voice Isolator and SAVA (Semantic-Acoustic Voice Activity) models, and publishing the technical repor...

AI summary

Inworld open-sourced two speech models, Voice Isolator and SAVA (Semantic-Acoustic Voice Activity), and published the technical report behind Realtime TTS-2.

Why it matters

The original post lists the open-source licenses, 48 kHz and 20 ms lookahead parameters of the two speech models, and the latency comparison between SAVA and open-source baselines.

Source: Inworld AI · x.com