Is speech to speech the future of voice AI, or just the future hype cycle

​Speech to speech models are one of the most debated shifts in voice AI right now. Some builders think it is the future of real time voice. Others think it still has too many gaps to be production ready.

​Hosted by Mahima Maurya and Unio, India's largest Voice AI builder community.

​Hemant, CTO of Dinodial, is not debating it from the sidelines. Dinodial is already running speech to speech models in production today.

​In this live Q&A, Hemant will share what building on speech to speech actually looks like once you move past the demo stage, the tradeoffs, the breakpoints, and what it takes to make it work in the real world.

​What we'll cover

​Why Dinodial chose speech to speech over the traditional STT, LLM, TTS pipeline

​Latency, cost, and control tradeoffs in a real production environment

​Where speech to speech models still struggle today, interruptions, function calling, and reliability

​What broke in production that never showed up in early testing

​Whether speech to speech is ready for use cases like BFSI, support, and sales, or still maturing

​An honest, builder level take on the theme question, is speech to speech really the future of voice AI

​Who should attend

​Founders, AI engineers, and product folks building voice products who want a grounded answer from someone actually shipping speech to speech, not just talking about it. Bring your questions, this is a live Q&A, not a lecture.