# Inworld AI

> Inworld AI began by building conversational game characters. It now sells the realtime voice, model routing and orchestration infrastructure that developers use to make games, companions, learning apps and customer experiences talk back. Its argument is economic as much as theatrical: a voice experience only survives at scale if the latency, quality and bill all work together.

- **Founded:** 2021
- **Headquarters:** Mountain View, United States
- **Founders:** Ilya Gelfenbeyn (Co-founder; former CEO), Michael Ermolenko (Co-founder and CTO), Kylan Gibbs (Co-founder and CEO)
- **Products:** Inworld Character Engine, Inworld Runtime, Realtime TTS, Realtime API, Realtime Router and Inference
- **Notable:** Origins made conversational NPCs playable in a 2023 detective demo., Death by AI's creators said switching to Inworld helped the game reach profitability after its original AI services proved too costly., Talkpal reported a 40% TTS cost reduction, 4% retention lift and 7% feature-usage lift in a four-week A/B test.

## Products & services

- **Inworld Character Engine** — Tools for creating AI-driven characters with dialogue, memory, behavior and game engine integration.
- **Inworld Runtime** — Graph-based orchestration for models, speech, memory, knowledge and tools in interactive applications.
- **Realtime TTS** — Text-to-speech API with custom voices and voice cloning; TTS-2 adds audio-context conditioning and natural-language direction.
- **Realtime API** — Persistent speech-to-speech connection combining recognition, reasoning and voice output.
- **Realtime Router and Inference** — Model routing across third-party LLMs and Inworld-hosted open models, with unified billing and experimentation.
- **Realtime STT** — Streaming speech recognition for conversational applications.

## Achievements

- Origins made conversational NPCs playable in a 2023 detective demo.
- Death by AI's creators said switching to Inworld helped the game reach profitability after its original AI services proved too costly.
- Talkpal reported a 40% TTS cost reduction, 4% retention lift and 7% feature-usage lift in a four-week A/B test.
- Published an open source training repository for its text-to-speech research.

## Latest updates

- **2026-08** — Released Realtime TTS-2, designed to condition voice output on previous audio and accept natural-language voice direction.
- **2026-06** — Published lower pricing across TTS, STT and inference, with spend-based tiers.
- **2026-01** — Released TTS-1.5 with lower reported first-audio latency and improved expressiveness.
- **2025-10** — Described its shift from character demos toward a broader runtime and voice infrastructure stack.

## Links

- Website: https://inworld.ai/
- LinkedIn: https://www.linkedin.com/company/inworld-ai
- Twitter/X: https://twitter.com/inworld_ai
- GitHub: https://github.com/inworld-ai
- YouTube: https://www.youtube.com/@inworldai
- Instagram: https://instagram.com/inworldai_dev
- Facebook: https://facebook.com/inworldai/
- TikTok: https://tiktok.com/@inworld_ai

---

Profile page: https://yespress.io/inworld-ai
Published by YesPress — https://yespress.io
Last updated: 2026-09-30