
Voice AI models and speech-to-text APIs
assemblyai.com (opens in a new tab)AssemblyAI builds Voice AI models and sells access to them as developer products: speech-to-text for recorded and streaming audio, speech understanding and an LLM gateway, with a playground for trying them out. Thousands of customers build on the models, Granola, Fireflies, Figure AI and CallRail among them, which together generate more than 600 million inference calls a month and over a million hours of processed audio a day, adding up to upward of two billion end-user experiences. The company stays small on purpose: fewer than 100 people bring in roughly $500K of annual recurring revenue each. It has offices in San Francisco and New York and keeps planning and approvals light, with tools and information open to everyone.