Inworld AI

Inworld AI

Inworld AI is a realtime voice AI lab and model provider run by Theai, Inc. in Mountain View, California, founded in 2021. Developers use its APIs for text-to-speech, speech-to-text and full-duplex voice conversations that chain STT, an LLM and TTS over one WebSocket. On demand, TTS-2 Flash costs $15 and TTS-2 $25 per million characters, with lower paid-tier rates.

Key Features

• Realtime TTS-2: Streaming speech synthesis with under 100 ms time to first byte and natural-language steering of pace, tone and emotion.
• Voice cloning and design: Create a custom voice from 15 seconds of audio or from a text description, and keep the same voice across 200+ languages.
• Realtime STT-1: Streaming speech recognition with voice profiling (emotion, accent, age, pitch) and configurable turn-taking, from $0.15 per hour on demand.
• Realtime API: Speech in, speech out over a single WebSocket or WebRTC connection, with custom voices and tool calling.
• LLM Router: Routes requests across 220+ models from providers such as OpenAI, Anthropic and Google at provider rates.

Use Cases

• Companion and social apps: Give AI companions and social characters expressive, low-latency voices.
• Voice agents: Build customer support and assistant agents that listen and respond in real time.
• Games and learning: Voice characters in games, media and language-learning apps such as Talkpal.

Getting Started

Developers can start on the free On-Demand plan and test voices in the TTS Playground: https://inworld.ai