v2.1.0 puts five Indian languages live on voice — Hindi, Marathi, Tamil, Bengali, and Telugu — shipping on day one of 2026. This is the release AIVA was really built for: a caller is understood in the language they actually speak, not just English or Hindi.
Every language ships with full code-switching support — callers mix English mid-sentence constantly, and AIVA keeps context instead of resetting. Native scripts render correctly across the dashboard, transcripts, and exports, and the acoustic models recognise regional accents within each language rather than a single “standard” variant.
What shipped in this release
- Voice agents for Hindi, Marathi, Tamil, Bengali, and Telugu — auto-detected per call.
- Code-switching between any of the above and English, mid-conversation.
- Native-script transcripts and CSV/JSON exports for every supported language.
- Regional-accent recognition trained on real-call audio, not studio recordings.
Available to every account from launch. Set language: "auto" to let AIVA detect the caller's language, or pin one with "hi", "mr", "ta", "bn", or "te". Why we built this first.