Gnani AI launches Timbre v3.0, bringing expressive AI speech to Indian languages
Built in-house, Gnani’s multilingual text-to-speech model combines natural speech, real-time responsiveness, and audio tags that give businesses and creators control over how every line is delivered.
Gnani AI, a frontier AI company building proprietary foundational models and enterprise agentic AI, today announced Timbre v3.0, its expressive text-to-speech model. Built for enterprises, developers and content creators, Timbre v3.0 brings natural multilingual speech and directed vocal expression into a single model, supporting everything from live customer conversations to audiobooks, learning experiences and video voiceovers.
“Building speech AI for India means paying attention to how people actually speak and listen,” shared Bharath Shankar, Chief Product and Engineering Officer, Gnani AI. “A date or an amount needs to be unmistakably clear. A story needs pauses, changes in pace and expressions. With Timbre v3.0, we are bringing that clarity and control into one multilingual model, so businesses and creators can shape both what a voice says and how it says it.”
“Consider a story being narrated in Hindi. A character hesitates before answering, lowers their voice to a whisper and then breaks into a laugh. The words tell only one part of the story. The pauses, pace, and delivery carry the rest. The same attention to delivery matters in everyday business conversations. A payment reminder needs to read an amount clearly. An insurance message needs to give the listener time to absorb a date. A learning app needs a voice that can hold attention through an explanation.”
“Timbre v3.0 brings these needs together: clear pronunciation, voices that retain their character across languages, and control over expression at the moments that matter. Today over 1 lakh hours of voice calls are powered by Timbre”, concluded Shankar.
EXPRESSION THAT CAN BE DIRECTED, LINE BY LINE
Timbre v3.0 supports multiple audio tags: short cues placed within text that guide how a line is spoken. Creators and developers can direct a voice to laugh, giggle, sigh, whisper or shout at a chosen point, add pauses, slow the delivery down or lift it with excitement.
Non-verbal sounds such as coughs, sniffles, yawns and exhales blend into the spoken output, giving storytellers and content teams more ways to shape a performance. These controls support character dialogue for games and animation, podcasts, advertisements, social video voiceovers and e-learning.
Audio tags work across all supported languages. For text that does not require explicit direction, the model automatically adapts delivery to the meaning, tone and punctuation of each sentence.
ONE MULTILINGUAL MODEL, A CONSISTENT VOICE ACROSS REGIONS
Timbre v3.0 supports Hindi, Bengali, Gujarati, Kannada, Malayalam, Marathi, Punjabi, Tamil, Telugu and Indian-accented English, with multiple prebuilt male and female voices.
A unified multilingual model preserves a voice’s character as it moves between languages, helping businesses maintain a recognisable brand voice across regions. Natural pacing and intonation are paired with pronunciation and text normalisation capabilities that turn numbers, dates, currencies, symbols and abbreviations into spoken language.
For banking, financial services and insurance, these details are central to the experience. For example, a voice agent might say, “Your EMI of ₹12,500.50 for loan account ending 4092 is due on 15/10/2026, at an annual interest rate of 8.75%.”
Text normalization turns these written formats into natural speech in the listener’s language, distinguishing monetary amounts, individual account digits, dates and percentages. Adjustable speaking rates let teams slow down important information, giving listeners time to absorb it, or move more quickly through routine updates.
BUILT FOR LIVE CONVERSATIONS AND LONG-FORM CONTENT
Timbre v3.0 delivers first audio in under 200 milliseconds, with streaming playback that begins before synthesis is complete. Long-form synthesis supports extended narration and responses without breaks, generating high-quality audio outputs. This provides flexibility to use the same voice across phone systems, applications and digital content.
Enterprise applications include voice agents and IVR systems for payment reminders, collections, KYC and onboarding, insurance renewals and claims, appointment reminders and customer support across financial services, insurance, healthcare and telecom. Beyond customer conversations, Timbre v3.0 supports audio articles, in-app guidance and accessibility tools that read content aloud in a listener’s preferred language.
