Inworld AI

Inworld AI is a realtime AI research lab and model provider, providing Voice AI built for realtime conversation that feels as human as it sounds. The company builds realtime voice models and LLM inference for consumer applications such as companions, learning, health and fitness, social apps, personal agents and entertainment.

Offering the #1 ranked text-to-speech and speech-to-speech, as well as user-aware LLM routing and speech recognition that also understands user context and emotions. Its voice lineup includes Realtime TTS-2 and Realtime TTS-2 Flash for text-to-speech, Realtime STT-1 for streaming transcription that detects traits such as emotion, style, age and accent, and a speech-to-speech API that lets developers choose the LLM for each session over a single WebSocket or WebRTC connection. Realtime TTS-2 supports about 500 languages and dialects.

Through Inworld Inference, the company serves open models such as Gemma, GLM and DeepSeek on its own fleet via an OpenAI-compatible API. Voice services run in four regions (the US, Europe, India and Asia Pacific), while LLM serving runs in the US, and the company cites SOC 2 Type II, HIPAA and GDPR compliance.