Bengaluru-based startup’s research on multilingual zero-shot text-to-speech synthesis earns a slot at one of the world’s most competitive speech and signal processing conferences, presenting on May 7, 2026
Convin.ai, an AI-powered conversation intelligence platform, today announced that its research paper has been accepted at ICASSP 2026 — the International Conference on Acoustics, Speech and Signal Processing — organised by IEEE (Institute of Electrical and Electronics Engineers). The paper will be presented on May 7, 2026 in the session SLP-P40, marking a significant milestone for the company and for Indian AI research on the global stage.
About ICASSP
ICASSP is widely regarded as the flagship conference of IEEE’s Signal Processing Society and one of the most competitive and prestigious forums in the fields of speech, audio, and signal processing worldwide. The conference draws top researchers and engineers from leading global universities and technology companies, with acceptance rates that reflect the high calibre of submissions. Being accepted at ICASSP is considered a mark of excellence in applied AI and speech research.
About the Research
The accepted paper, titled “IPACue-TTS: Integrating Prosody and Articulatory Cues in Conditional Flow Matching for Multilingual Zero-Shot TTS,” introduces a novel approach to text-to-speech (TTS) synthesis. The system integrates prosody and articulatory cues within a conditional flow matching framework to achieve natural-sounding, zero-shot speech synthesis across multiple languages — without requiring speaker-specific training data.
The research was authored by Giridhar Pamisetty and Atul Shree (Director) of Convin.ai and addresses one of the most challenging problems in conversational AI: producing high-quality, expressive, and human-like speech across diverse languages and speaker profiles. This has direct implications for Convin’s core product, which is used by sales and customer support teams across India and globally.
About Convin.ai
Convin.ai is an AI-powered conversation intelligence platform that helps businesses improve the quality and efficiency of their customer-facing interactions. Used by sales, support, and quality assurance teams, Convin’s platform leverages advanced speech and language technologies to transcribe, analyse, and coach agents in real time. Headquartered in Bengaluru, India, Convin.ai serves clients across multiple geographies and industries. For more information, visit www.convin.ai.
Founded by IITians, Built for India
Convin.ai was founded by four IIT alumni who shared a vision of building world-class AI from India, for the world. The founding team brings deep expertise spanning speech and signal processing, machine learning, and enterprise software — a combination that has enabled Convin to build its entire AI stack in-house, from the ground up, without reliance on third-party models or APIs. This engineering-first DNA is what drives the company’s research ambitions, including the ICASSP 2026 paper.
Customer Base & Industry Impact
Trusted by over 100 enterprises across BFSI, insurance, healthcare, and public services — including marquee names such as Kotak Life, ICICI Bank, SBI Life, and others — Convin manages millions of customer interactions every month. The platform powers the full lifecycle of customer conversations: AI Phone Call agents handle inbound and outbound voice interactions in multiple languages around the clock; a Real-Time Assist agent provides live compliance prompts, knowledge base lookups, and guided scripts to human agents during calls; and a Post-Conversation Intelligence engine automatically audits and scores 100% of interactions, replacing the industry norm of manually sampling just 1–2% of calls. This end-to-end approach, from automated voice AI to real-time coaching to post-call analytics, is what sets Convin apart in the market.
Proprietary Indian-Language AI Stack
What makes Convin unique in the Indian AI landscape is its fully proprietary technology stack. The company has built its own Speech-to-Text engine natively trained on diverse Indian languages, regional accents, and code-mixed speech across 10+ languages including Hindi, Tamil, Telugu, Kannada, Bengali, Marathi, Gujarati, Malayalam, Punjabi, and Odia. Its in-house Text-to-Speech models — the very technology underlying the ICASSP 2026 paper — generate expressive, human-like Indian voices with natural prosody and low-latency, real-time delivery. Convin has also fine-tuned sovereign large language models for secure, on-premise government and enterprise use, ensuring that all citizen and customer data remains within the client’s own infrastructure. This sovereign AI approach, with zero third-party LLM dependencies, is rare among Indian AI startups and positions Convin as a trusted technology partner for data-sensitive sectors.
Government Collaborations & Recognition
Convin has actively collaborated with government entities to deploy AI for public good. Its flagship 3D Indian Sign Language (ISL) rendering system, developed for Western Railway, converts live announcements — such as train status updates and emergency alerts — into accurate ISL gestures via photorealistic 3D avatars displayed on station signage. Currently serving millions of passengers across the Western Railway network, it is the only scalable, real-time ISL solution deployed on India’s public infrastructure, and is adaptable for airports, hospitals, and judicial offices. Beyond Western Railway, Convin is currently piloting its advanced AI voice bot with Kochi Metro for automated customer grievance handling and with the Tamil Nadu Skill Development Corporation (TNSDC) for conducting student surveys — further demonstrating the platform’s versatility in public-sector applications.The company’s innovation has been widely recognised: Convin was awarded Rail Analysis Startup of the Year at the 7th Rail Analysis Innovation & Excellence Summit 2026, and is a top-rated G2 performer across seven categories, ranked among India’s leading AI startups. Convin holds comprehensive compliance certifications including ISO 27001, SOC 2, GDPR, HIPAA, PCI DSS, CCPA, and RBI data mandates, ensuring enterprise-grade security and regulatory readiness for all deployments. Except for one signal processing patent, all of the research and engineering behind this ICASSP-accepted work was carried out entirely at Convin.ai.
