Gnani AI announced Timbre v3.0 on October 14, 2026, describing an expressive text-to-speech model for Indian languages. It says the launch includes 13 voices, 13 audio tags and 10 languages. These are the vendor's launch specifications; the article does not supply an independent native-speaker quality test.
Audio tags are short cues placed within a script to direct delivery, such as a laugh, sigh, whisper or pause. Gnani says tags influence how a line is spoken rather than being read aloud. It also says untagged text responds to meaning, tone and punctuation, so tags can be reserved for selected moments.
The launch is available now through the Gnani app, where users can select Timbre v3.0 and add tags to scripts. Gnani says availability through Pipecat and LiveKit will come later. Those integrations should not be treated as launched, and the source gives no fixed delivery date for them.
Gnani positions the controls for business voice conversations and creator work such as voiceovers, stories and ads. These are proposed applications, not proof of a measured conversion increase or improved customer outcomes. The announcement does not set out a complete price schedule or commercial licensing comparison.
The company advises using tags sparingly, placing them before text they should affect, and having a compliance team review business scripts. It also recommends listening in each language, ideally with a native speaker, and testing real call audio before deployment. Generated expression alone does not make a script accurate or appropriate.
Gnani says Timbre 2.5 continues serving everyday voice-AI conversations and that more voices and languages will be added later. Those future additions remain plans. Buyers should distinguish the current app release from later integrations and check language-specific output and service terms for their own use.