LMNT
Low-latency AI voice API built for real-time agents, games, and apps.
About
LMNT is a text-to-speech platform built specifically for real-time voice AI. It streams audio in roughly 150 to 200 milliseconds, which makes it a fit for conversational agents, voice bots, and games where a pause before speech breaks the illusion.
It offers studio-quality voice cloning from a short recording, dozens of built-in voices, and multilingual generation, with no concurrency or rate limits on API plans. Reviewers list customers such as Khan Academy, Vercel, and Unity, and the Palo Alto company is backed by investors including Elad Gil and Conviction. A free playground and API tier let developers test voices before committing.
What it can do
- Sub-200ms streaming text-to-speech
- Voice cloning from a short audio sample
- Multilingual synthesis with mid-sentence switching
- REST and WebSocket APIs with Python and Node SDKs
Setup
Try voices in the web playground, then integrate the streaming API with an API key.
Prerequisites
An LMNT account and API key for production use.
Use cases
Works with
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 115; Trending score 76/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.