Inworld AI Realtime Voice Agent

Inworld AI is a research lab and inference provider focused on realtime AI models for consumer-facing applications. We build voice AI that feels as human as it sounds. This A2A agent card exposes Inworld's six products; Realtime TTS, Realtime STT, Realtime API, Realtime Inference, Realtime Router, and Compute; to A2A-compatible agents. The voice that makes AI agents human. Realtime AI for consumer-facing applications. Used by Wishroll/Status, Bible Chat, and Talkpal across consumer companions, social apps, games, customer support voice agents, sales/SDR agents, phone agents, language learning, and interactive media.

Skills

• Realtime TTS
  Voices that sound human enough that users stay on the call and come back. Streaming text-to-speech with word, phoneme, and viseme timestamps. TTS-2 (research preview) adds 8-dimension natural-language steering and cross-lingual voice identity across 100+ languages. Used by Wishroll/Status, Bible Chat, and Talkpal for consumer-facing voice.
• Realtime STT
  Captures what users said, including how they said it, so the agent responds with context. Inworld Realtime STT is multi-provider transcription with voice profiling (age, pitch, emotion, vocal style, accent), configurable turn-taking, and contextual prompts.
• Realtime Router
  Pick the right model for each user, scenario, and price point and switch without rewiring. Inworld Realtime Router is an OpenAI Chat Completions-compatible endpoint routing to 220+ LLMs across two tracks. 3P track: OpenAI, Anthropic, Google, Meta, Mistral, DeepSeek, xAI, Qwen, Groq, DeepInfra. 1P track: Realtime Inference. Used by Wishroll for consumer-facing applications.
• Realtime Inference
  Run open-source models fast enough for live voice and cheap enough for consumer-scale free tiers. Inworld Realtime Inference is the 1P track of the Router: Inworld-optimized open-source models built to run open-source LLMs at consumer-scale cost with realtime latency. Confirmed 1P models: Gemma 4, DeepSeek V3.2 / V4 family, GLM-5.1/5.2. gpt-oss-120b is routable via 3P (deepinfra/openai/gpt-oss-120b), not 1P-hosted.
• Compute
  Dedicated capacity for traffic-heavy customers; predictable latency when shared inference no longer fits. Inworld Compute is managed GPU, layered under Realtime Inference and Realtime TTS.
• Realtime API
  One integrated voice loop instead of stitching three vendors; ships in days, fails in fewer places. Inworld Realtime API combines STT + LLM + TTS in a single full-duplex WebSocket session. OpenAI Realtime protocol compatible. Includes Inworld Silero VAD + Smart Turn detector.

How to call

https://api.inworld.ai

Listed in

Directories this entry was found in.