Inworld AI

Overview

Inworld AI has pivoted from its original game NPC platform to become a real-time Voice AI infrastructure provider. As of May 2026, it holds the #1 position on the Artificial Analysis TTS leaderboard with its TTS 1.5 Max model (P90 sub-200ms latency). Game developers use it for: real-time NPC voice with emotion markup (anger, joy, sadness, fear), instant voice cloning from 15 seconds of reference audio, word-level and viseme-level timestamps for lipsync animation, and a full speech-to-speech pipeline (STT → LLM router → TTS) in one API call. The original NPC Studio and Unity/Unreal game SDKs are no longer the core product — the current offering is a developer API.

Best For

• Games needing real-time NPC voice with sub-200ms latency
• Developers wanting viseme-level lipsync timestamps for 3D characters
• Voice agents and interactive NPCs using speech-to-speech pipeline

Game Development Use Cases

Key Features

• #1 ranked TTS (Artificial Analysis, May 2026)
• P90 sub-200ms real-time latency
• Emotion markup (anger, joy, sadness, fear, surprise)
• Viseme + word-level timestamps for lipsync
• Instant voice cloning (15 seconds of audio)
• Realtime speech-to-speech API (STT + LLM + TTS)

Pricing

Freemium

On-Demand: Free (40 min TTS trial). Creator: $25/mo. Developer: $300/mo (20% rate discount). Growth: $1,500/mo (40% discount). Enterprise: custom (rates as low as $10/1M chars). TTS-2 model: $35/1M chars on-demand.

Commercial Use in Games

On-Demand free tier: 40 free minutes of TTS for prototyping, non-commercial. Creator ($25/mo) unlocks commercial use. Pricing based on characters processed — verify current rates at inworld.ai/pricing before production commitment.

Pros and Cons

Compare

Related Articles