LMNT
Developer-focused AI text-to-speech with low-latency streaming, voice cloning, and support for 31 languages.
About LMNT
LMNT is an AI text-to-speech platform for developers building conversational applications, agents, games, and other products that need generated speech. Its API turns text into speech in 31 languages, supports mid-sentence language switching, and streams audio with stated latency of 150–200 milliseconds. Voice cloning is available from a five-second recording, and every listed plan includes unlimited voice clones. Developers can start in a browser playground or integrate through Python, TypeScript, Go, and cURL examples in the documentation. The free plan includes 15,000 characters. Paid plans raise the included character allowance, add commercial licensing, remove concurrency and rate limits, and charge for additional usage at plan-specific rates. This makes LMNT practical for prototypes as well as production voice experiences, though usage-based overages require teams to track character volume. Organizations handling recordings, text, or generated speech should also account for the service’s collection of uploaded content described in its privacy policy.
Best for
Best use cases
A conversational session streams generated speech while an LLM produces the response text.
A game generates character dialogue in several languages and switches language inside a sentence.
A production service converts recurring scripts to audio while tracking included characters and overage costs.
Compare before choosing
LMNT emphasizes 150–200ms streaming latency for interactive speech workloads.
All listed plans include unlimited voice clones, while paid tiers add commercial licensing.
Monthly tiers differ mainly by included character volume and the per-character overage rate.
Features
Conversational agents
Stream low-latency speech for voice assistants and LLM-powered sessions that need quick spoken responses.
Game characters
Generate multilingual dialogue and cloned character voices for interactive game experiences.
Narration workflows
Turn scripts into lifelike speech for tutorials, stories, broadcasts, and other audio content.
Pros
- The $0 plan includes 15,000 characters, so developers can test real API output before paying.
- Low-latency streaming at a stated 150–200ms supports responsive agents, games, and conversational interfaces.
- All plans include unlimited voice clones, and a clone can be created from a five-second recording.
Cons
- The free allowance is limited to 15,000 characters, which can run out quickly in sustained testing.
- Paid plans use character-based overages, so high or unpredictable traffic can increase monthly cost.
- Commercial licensing and removal of concurrency or rate limits begin on paid tiers rather than the free plan.
Pricing Plans
Free
- 15K characters included
- Unlimited voice clones
Indie
- 200K characters included
- $0.05 per 1K characters after
- No concurrency or rate limits
- Unlimited voice clones
- Commercial license
Pro
- 1.25M characters included
- $0.045 per 1K characters after
- No concurrency or rate limits
- Unlimited voice clones
- Commercial license
Premium
- 5.7M characters included
- $0.035 per 1K characters after
- No concurrency or rate limits
- Unlimited voice clones
- Commercial license
* Pricing information is for reference only. Please refer to the latest announcement on LMNT's official website.