a rough idea how this is happening behind the scene. if any company uses this architecture (speech to text and text to llm) will be mostly killed by OpenAI realtime API https://cdn.discordapp.com/attachments/1211484481848873020/1291120587728752680/voice-arch.drawio.png?ex=66fef13f&is=66fd9fbf&hm=43f2a7855299a6d66cafa258da0d8a970aa86f1fde80046a6cfaa7a229a5758c&