Inworld opensources Voice Isolator, SAVA models
Inworld AI has open-sourced its Voice Isolator and SAVA (Semantic-Acoustic Voice Activity) models, alongside a technical report for its Realtime TTS-2 engine.
By open-sourcing core audio processing components, Inworld is clearly moving to establish its technology stack as a standard for real-time AI voice applications.
- –Voice Isolator enables clean input by removing background noise while preserving speaker nuances.
- –SAVA enhances voice activity detection by combining semantic and acoustic signals, crucial for responsive voice agents.
- –Realtime TTS-2 documentation provides transparency into the low-latency speech generation architecture.
- –This move strengthens Inworld’s developer ecosystem and positions it as more than just a closed character engine.
DISCOVERED
1h ago
2026-11-18
PUBLISHED
1h ago
2026-11-18
RELEVANCE
AUTHOR
inworld