How do I migrate from ElevenLabs to LMNT and claim the 500,000 free migration credits?

Most teams switching from ElevenLabs to LMNT want two things: a drop‑in migration path that doesn’t break production, and a fast way to unlock the 500,000 free migration credits so they can test at scale before committing. You can do both in a day: wire up LMNT’s streaming TTS to your existing agents or games, then submit a simple proof of prior ElevenLabs spend to get credits added to your LMNT account.

Quick Answer: You migrate from ElevenLabs to LMNT by mapping your existing TTS calls to LMNT’s streaming API, cloning or selecting equivalent voices, and then validating latency and quality in your staging environment. To claim the 500,000 free migration credits, you sign up for LMNT, share verification of your recent ElevenLabs usage with the LMNT team, and have those credits applied so you can run side‑by‑side tests and production traffic without extra cost.

Why This Matters

If you’re already live on ElevenLabs, any TTS switch has risk: latency regressions, voice mismatches, or rate limits that quietly cap your growth. LMNT’s migration offer is designed to de‑risk that shift with both a production‑grade stack (150–200ms streaming, no concurrency limits, studio‑quality voice cloning from a 5‑second sample) and a big enough credit pool to run real traffic, not just tiny experiments.

Key Benefits:

• Faster conversational latency: LMNT’s 150–200ms low‑latency streaming keeps agents and game characters responsive enough for natural turn‑taking.
• Economical at scale: Character‑based pricing plus 500,000 free migration credits lets you benchmark LMNT against ElevenLabs under realistic load.
• Production‑ready reliability: No concurrency or rate limits and SOC‑2 Type II compliance mean you can migrate serious workloads, not just side projects.

Core Concepts & Key Points

Concept	Definition	Why it's important
TTS migration	The process of switching your text‑to‑speech backend (e.g., from ElevenLabs to LMNT) without changing your product UX.	Lets you upgrade latency, quality, and pricing without rewriting your whole stack.
Streaming voice API	A TTS interface that sends audio chunks over a persistent connection (often WebSockets) as they’re generated.	Enables 150–200ms response times for agents, tutors, and game characters instead of multi‑second delays.
Migration credits	Free LMNT usage credits (500,000 characters in this offer) provided to teams coming from ElevenLabs.	Gives you budget to run side‑by‑side tests, A/B agents, and phased cutovers before paying.

How It Works (Step-by-Step)

At a high level, you’ll (1) map your ElevenLabs usage to LMNT features, (2) plug LMNT into your app’s TTS layer, and (3) request and use your 500k migration credits to scale testing and rollout.

1. Map your current ElevenLabs usage

   • List where you call ElevenLabs today:
     • Chat agents and support flows
     • Tutors, coaches, or voice UIs
     • NPCs or narrators in games
   • Capture the key parameters you’re using:
     • Voices (IDs, styles, genders, locales)
     • Languages and any code‑switching scenarios
     • Average/peak character usage and concurrency
   • This gives you a concrete checklist to replicate on LMNT: voice selection or cloning, language coverage, and load profile.

2. Create and configure your LMNT account

   • Sign up at lmnt.com and open the free Playground.
   • Try the built‑in voices (e.g., “Brandon” for an engaging broadcaster, “Leah” for a cheerful assistant) to find close matches to your ElevenLabs voices.
   • If you have branded or signature voices on ElevenLabs:
     • Use LMNT’s studio‑quality voice cloning with just a 5‑second recording to recreate them.
     • Test across your main content types (short replies, longform narration, emotionally varied lines).

3. Wire LMNT into your TTS abstraction

   • If you already have a “TTS provider” interface, add LMNT as a new backend instead of ripping out ElevenLabs immediately.
   • Use LMNT’s Developer API:
     • Browse the API spec at https://api.lmnt.com/spec.
     • Start with the example prompt: build a Rust app that reads https://text.npr.org/ in the “brandon” voice as a quick end‑to‑end check.
   • For conversational apps and agents:
     • Prefer LMNT’s low‑latency streaming mode (150–200ms) over batch synthesis.
     • Pipe audio chunks to your existing WebRTC, WebAudio, or mobile audio layer as they arrive.
   • For games and interactive characters:
     • Integrate LMNT in the same place you previously called ElevenLabs—usually your dialogue system or NPC controller.
     • If you’re on Unity, treat LMNT as your live voice backend so you can generate lines on the fly instead of pre‑baking everything.

4. Run side‑by‑side tests with LMNT

   • In staging or a small production slice, mirror your TTS traffic:
     • One branch calls ElevenLabs.
     • One branch calls LMNT.
   • Compare:
     • Latency: Time‑to‑first‑audio and full utterance completion; LMNT should sit in the 150–200ms range for streaming.
     • Naturalness: Prosody, emphasis, and multilingual flow, including mid‑sentence switching across LMNT’s 24 languages.
     • Stability under load: Spin up more concurrent sessions; LMNT has no concurrency or rate limits, so you shouldn’t see throttling as you scale.

5. Request your 500,000 free migration credits

   • Once you’ve validated that LMNT can replace (or outperform) your ElevenLabs setup, request the migration credits:
     • Sign into your LMNT account.
     • Contact LMNT via the “Get started” or “Contact us” path and mention you’re migrating from ElevenLabs.
     • Provide simple verification of recent ElevenLabs usage—this is usually:
       • A recent ElevenLabs invoice or billing screenshot (with sensitive details redacted).
       • A quick summary of your use case (e.g., “conversational support agent,” “NPC dialog in Unity,” “language tutor”).
   • The LMNT team will:
     • Apply 500,000 migration credits to your LMNT account.
     • Confirm how those credits map to your approximate traffic (e.g., how many calls or days of usage you can expect under your typical load).

6. Use migration credits to scale rollout

   • With credits applied, run a phased migration:
     • Phase 1 — Canary: Route 5–10% of sessions to LMNT. Monitor logs, audio quality, and user feedback.
     • Phase 2 — A/B: Split traffic 50/50 between ElevenLabs and LMNT. Compare session time, completion rate, and any qualitative comments about the voice.
     • Phase 3 — Cutover: Move 90–100% of traffic to LMNT and keep ElevenLabs as a short‑term fallback until you’re fully confident.
   • Because LMNT doesn’t enforce concurrency or rate limits, you can safely crank up traffic as your confidence grows, without worrying about hitting unseen ceilings.

7. Optimize and expand after migration

   • Tune for your use case:
     • Conversational agents: adjust speech rate and phrasing to capitalize on LMNT’s low latency and natural prosody.
     • Games: create multiple voice clones for different character archetypes using short seed recordings.
     • Multilingual use cases: lean into LMNT’s 24‑language support and mid‑sentence switching to offer localized yet fluid experiences.
   • When usage grows:
     • Move from free + migration credits into LMNT’s paid tiers for a commercial license and better volume economics.
     • If you’re a startup, check eligibility for the Startup Grant (45M credits over 3 months) to extend your runway.

Common Mistakes to Avoid

• Treating migration as a single “flip the switch” event:
  How to avoid it: Keep both ElevenLabs and LMNT wired into your TTS abstraction. Use environment flags or feature toggles so you can gradually raise LMNT traffic share and roll back instantly if needed.

• Skipping voice parity testing:
  How to avoid it: Before you cut over, play the same canonical scripts (greetings, error messages, long explanations) through your existing ElevenLabs voice and the LMNT equivalent or clone. Make sure tone, energy, and intelligibility match your product’s expectations—especially for branded characters or support personas.

Real-World Example

Imagine you’ve built a real‑time language tutor that currently uses ElevenLabs. Students speak into the app, you run speech‑to‑text + LLM, and then ElevenLabs reads back the answer. It works, but under higher load your response time creeps into multi‑second territory and you’ve hit concurrency caps that limit simultaneous users.

You sign up for LMNT, try the “Vesper” and “Leah” voices in the Playground, and decide to clone your existing tutor voice with a 5‑second sample. You then add LMNT as a second backend in your TTS service, use the streaming API to start playback within ~150–200ms, and mirror 10% of production sessions to LMNT. Once you see better latency and consistent student satisfaction scores, you contact LMNT support, show your ElevenLabs invoice, and get 500,000 migration credits. Those credits cover a full month of A/B testing plus the eventual cutover, all without extra TTS spend while you transition.

Pro Tip: When you start your LMNT A/B test, log the per‑turn latency from “LLM finished” to “first audio frame played.” It’s one of the fastest ways to quantify the impact of LMNT’s streaming engine versus your existing ElevenLabs setup.

Summary

Migrating from ElevenLabs to LMNT is mostly about preserving your existing UX while upgrading the engine underneath: you map your current usage, integrate LMNT’s low‑latency streaming API, clone or match your voices, then roll traffic over gradually. The 500,000 free migration credits are there to fund real, production‑like testing—once you’ve verified fit, you share proof of ElevenLabs usage with the LMNT team and they apply the credits so you can confidently ramp. With no concurrency limits, 24 languages, and studio‑quality clones from 5 seconds of audio, you can move your agents, tutors, and games to a stack that’s built for real‑time voice at scale.

Next Step

Get Started