Inworld

Models

Inworld-TTS-1.5 is an advanced text-to-speech (TTS) model that converts written text into natural, expressive, and human-like speech. Designed for low latency and real-time performance, it supports hi...

Inworld TTS 1.5

Readme

Inworld AI Models on each::labs

Inworld AI is a leading developer of real-time voice AI platforms and Agent Runtime solutions, specializing in high-performance Text-to-Speech (TTS) models, voice cloning, and conversational AI agents for immersive applications. The company excels in creating dynamic, low-latency voice technologies that power non-player characters (NPCs) in gaming, streaming assistants, and interactive media experiences, with proven partnerships including NVIDIA, Xbox, Disney, Ubisoft, and Meta. Through each::labs, developers gain seamless API access to Inworld's cutting-edge models, enabling integration into games, AI companions, and real-time voice agents without managing complex infrastructure.

Positioned at the forefront of the AI ecosystem for entertainment and consumer applications, Inworld AI transforms static interactions into lifelike, responsive conversations. Their technology supports sub-200ms latency TTS, multilingual capabilities across 15 languages, and model-agnostic orchestration for LLMs from providers like OpenAI, Anthropic, Google, and Mistral. On eachlabs.ai, Inworld's tools join a unified platform offering access to over 150 AI models, making it the go-to hub for builders seeking real-time voice AI innovation.

What Can You Build with Inworld?

Inworld AI focuses on voice AI categories, including ultra-low-latency Text-to-Speech (TTS), voice cloning, and Agent Runtime for orchestrating conversational pipelines with speech-to-text (STT), LLMs, and tools. These capabilities shine in gaming for dynamic NPCs, streaming for intelligent assistants, and enterprise voice agents for support or sales.

Text-to-Speech (TTS): Inworld's TTS-1.5 models deliver top-ranked realism with P90 latency under 250ms, supporting emotion controls and non-verbal sounds. For example, game developers use it to generate natural NPC dialogue that adapts to player actions, enhancing immersion in open-world titles.
Voice Cloning: Instant zero-shot cloning from 15 seconds of audio or professional fine-tuning with 30+ minutes creates custom voices ready in minutes. Creators build personalized AI companions, like a virtual coach mimicking a user's favorite streamer for motivational feedback during workouts.
Agent Runtime: This C++ core handles real-time conversational AI, unifying TTS, STT, LLMs, and external tools with observability and A/B testing. It's ideal for gaming NPCs that evolve narratives based on player behavior or streaming assistants managing live interactions.

A concrete scenario: Imagine developing a multiplayer RPG where NPCs respond unpredictably to quests. Using Inworld via each::labs API, send this prompt: "Player approaches a suspicious merchant in a fantasy tavern. Generate dialogue with hesitant tone, offering a shady deal on enchanted armor, while scanning for player inventory hints." The Agent Runtime processes it in sub-200ms, outputting cloned voice audio: "Psst, traveler... I see that rusty sword of yours. Trade it for this armor of shadows? No questions asked." This creates emergent storytelling, proven with 20 million players in commercial games.

Inworld's TTS-1.5 Max tops benchmarks like Artificial Analysis TTS Arena for quality, beating competitors in blind tests with 59-60% win rates, while TTS-1.5 Mini prioritizes extreme latency at half the cost. Multilingual support for languages like Hindi, Chinese, Japanese, and Korean enables global deployments, from esports NPCs to international customer support bots.

Why Use Inworld Through each::labs?

each::labs positions itself as the premier unified platform for AI model access, delivering Inworld's real-time voice AI alongside 150+ models from top providers in a single, production-ready API. This eliminates provider lock-in, letting developers switch between Inworld TTS, image generators, or video tools without rewriting code.

Key advantages include:

Unified API Approach: One endpoint for Inworld's low-latency TTS and Agent Runtime, plus seamless integration with other models for multimodal apps like voice-enabled video editors or NPC-driven simulations.
SDK Support: Robust SDKs for Node.js, Unreal Engine, and more, with quickstarts for TTS APIs and MCP tool integrations, accelerating from prototype to production.
Playground Environment: Test Inworld models interactively in the browser-based playground, tweaking prompts, voices, and latencies before scaling.
Production-Ready Features: SOC 2 Type II compliance, GDPR zero-data-retention options, HIPAA readiness, and on-premise deployment ensure enterprise security.

Pricing is developer-friendly: Agent Runtime is free, with TTS-1.5 Max at $10 per million characters (~1¢/minute) and Mini at $5/million—25x lower than some alternatives—billed consumption-based with a free tier. Benchmarks confirm Inworld's edge in quality, latency, and cost, making each::labs the smartest path for voice cloning APIs or gaming AI tools.

Getting Started with Inworld on each::labs

Sign up at eachlabs.ai for instant access to Inworld models via the intuitive Playground, where you can experiment with TTS prompts and voice cloning in seconds. Dive into comprehensive API documentation and SDKs to integrate into your Unity or web app, starting with sample code for real-time agents. Begin building immersive experiences today—your first API call unlocks Inworld AI's full potential on the platform designed for scalable innovation.