India-Specific Voice AI Models Emerge for Mass Market
New developments in India's AI landscape target mass-market and vernacular commerce. Sarvam AI's Vikram 30B model can now run voice-AI chatbots on basic phones costing as little as ₹1,500, supporting 22 Indian languages. Separately, Gnani.ai now offers voice cloning in 12 languages with less than 10 seconds of training audio, enabling personalized merchant-customer communication.
- The push for vernacular AI is supported by the Indian government's Bhashini platform, which provides free APIs, open-source datasets, and pre-trained models for all 22 scheduled Indian languages to accelerate development. - Sarvam AI's Vikram models utilize a "mixture-of-experts" architecture to improve efficiency; the 30B model was trained on 16 trillion tokens of text and is designed for real-time conversations with a 32,000-token context window. - Gnani.ai's voice cloning model, named Vachana TTS, supports 12 Indic languages and has achieved a Mean Opinion Score (MOS) of 4.23, a metric indicating high-quality, natural-sounding speech. - The ecosystem for voice-based transactions is maturing with the National Payments Corporation of India's (NPCI) "Hello! UPI" initiative, which enables conversational payments and fund transfers through voice commands in multiple Indian languages. - A critical regulatory shift for WhatsApp-based businesses is a new policy effective January 15, 2026, which prohibits the use of general-purpose AI chatbots on the Business API, restricting AI to specific functions like customer support and order processing. - Major e-commerce platforms like Flipkart have already integrated multilingual voice search capabilities to better serve customers in Tier 2 and Tier 3 cities, where voice inputs are outpacing text-based searches. - The operational cost for deploying such technologies is becoming more transparent, with voice AI services in India typically priced on a per-minute model, ranging from ₹6 to ₹18 per minute depending on volume and features. - Both Sarvam AI and Gnani.ai are among a select group of startups chosen under the government's ₹10,372-crore IndiaAI Mission to develop sovereign foundational models, ensuring data privacy and India-specific customization.
