What is Inworld TTS?
Inworld TTS is a low-latency text-to-speech and voice cloning API for apps, games and real-time voice agents. It is made by Inworld AI, a research lab originally known for AI game characters, which acquired the voice agent platform Ultravox in October 2026. Its Realtime TTS-2 family includes TTS-2 for expressive speech with natural language steering and TTS-2 Flash, which Inworld says reaches about 25 ms time to first byte.
Inworld TTS competes on quality and price. The company claims the top spot for realtime TTS on the Artificial Analysis leaderboard, supports 200+ languages with cross-lingual voice cloning, and charges per million characters at rates well below many rivals. Instant cloning needs only 5 to 15 seconds of audio. Developers use it for voice agents, game characters, narration and accessibility, and it integrates with LiveKit, Pipecat, Vapi and Voximplant.
Main Features
- Generate speech with Realtime TTS-2 or the faster TTS-2 Flash.
- Stream audio over WebSocket for real-time voice agents.
- Clone a voice instantly from 5 to 15 seconds of audio.
- Create professional voice clones from 30+ minutes of audio on higher plans.
- Speak 200+ languages and keep a cloned voice across languages.
- Control speaking rate, temperature, pronunciation and timestamps.
- Plug into LiveKit, Pipecat, Vapi and other voice frameworks.
How to Get Started with Inworld TTS
- Create an Inworld account at inworld.ai and get an API key.
- Try voices in the TTS playground and pick or clone a voice.
- Call the REST or streaming API from your app, or use a framework integration.
- Move to a paid plan for lower per-character rates and more custom voices.
Inworld TTS Pricing
The On-Demand plan is free to start with up to 70 minutes of TTS, then $25 per million characters for TTS-2 or $15 for TTS-2 Flash. Creator costs $25 per month with $25 of credits and lower rates, Builder costs $100 per month and Developer $300 per month. Enterprise rates go as low as $5 per million characters.
Who Should Use It?
- Developers building real-time voice agents.
- Game studios voicing characters and NPCs.
- Startups that need cheaper TTS at scale.
- Teams migrating from other TTS providers.
- Creators building narration and audio apps.
Inworld TTS Alternatives
- ElevenLabs AI: broad voice library and creator tools
- Deepgram: speech-to-text and TTS from one provider
- Hume AI: emotionally expressive voice models
Frequently Asked Questions
Is Inworld TTS free?
Yes, you can start free with up to 70 minutes of text-to-speech. Paid plans start at $25 per month.
What is Inworld TTS used for?
Inworld TTS is used to add realistic, low-latency voices to apps. It powers voice agents, games and narration.
Does Inworld TTS support voice cloning?
Yes. Instant cloning needs 5 to 15 seconds of audio, and professional cloning is available on higher plans.
Inworld TTS vs ElevenLabs: which is better?
Inworld is usually cheaper per character and built for real-time agents, while ElevenLabs offers a larger set of creator tools.