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 clean migration path and a way to test LMNT at production scale without blowing their budget. That’s exactly what the 500,000 free migration credits are for—so you can port your existing TTS and voice clones, hit LMNT-level latency, and validate quality before you commit.

Quick Answer: You migrate from ElevenLabs to LMNT by mapping your existing voices and call patterns to LMNT’s streaming TTS API, recreating (or improving) your voice clones, and then testing in your staging/production flows using the free 500,000 migration credits. To claim those credits, sign up for LMNT, reach out to the team with proof of your current ElevenLabs usage, and they’ll apply a one-time migration grant to your account so you can run side‑by‑side tests at scale.

Why This Matters

If your conversational app, agent, or game is already live on ElevenLabs, switching providers isn’t just about sound quality—it’s about latency budgets, reliability, and predictable pricing. A poorly managed migration can break turn-taking, introduce audio lag, or surprise you with throttling right when usage spikes.

LMNT is built specifically to avoid those failure modes: 150–200ms low-latency streaming, studio-quality voice clones from as little as a 5-second recording, 24 languages with mid-sentence code-switching, and no concurrency or rate limits. The 500,000 free migration credits let you validate those claims in your real workload—side-by-side with your existing ElevenLabs setup—before you flip the switch.

Key Benefits:

Risk-free migration testing: Use 500k free credits to run live traffic, A/B tests, and load tests without double-paying during the switch.
Better conversational performance: Move to 150–200ms streaming with no concurrency or rate limits, so your agents and games feel responsive under real load.
Higher-quality voice clones with less data: Recreate or improve your ElevenLabs voices using LMNT’s studio-quality cloning from a 5-second recording.

Core Concepts & Key Points

Concept	Definition	Why it's important
Migration credits	A one-time pool of 500,000 LMNT characters granted to existing ElevenLabs users to support their switch.	Lets you run real migration tests and parallel traffic without extra cost, so you can verify latency, quality, and stability before cutover.
Streaming TTS migration	The process of replacing ElevenLabs TTS endpoints and SDK calls with LMNT’s low-latency streaming API.	Ensures turn-taking and real-time interactions stay smooth, especially for agents, tutors, and in-game characters.
Voice clone parity & upgrades	Recreating your existing ElevenLabs voices (or improving them) via LMNT’s 5-second studio-quality cloning.	Keeps your brand and character identity consistent while potentially improving realism and multilingual delivery.

How It Works (Step-by-Step)

At a high level, migrating from ElevenLabs to LMNT and claiming the 500,000 free migration credits looks like this:

Create your LMNT account and playground baseline
Request and receive your 500,000 migration credits
Map ElevenLabs features to LMNT (voices, languages, streaming)
Swap your API integration and run side-by-side tests
Cut over traffic and scale with LMNT

Below is the process in more detail.

1. Create your LMNT account and baseline in the Playground

Start by getting hands-on with LMNT’s voices and streaming behavior.

Sign up on LMNT
- Go to https://lmnt.com and create an account.
- Access the Playground from the top navigation.
Test built-in voices and latency
- Try voices like Brandon (engaging broadcaster), Leah, Vesper, Natalie, and Tyler.
- Note latency: LMNT targets 150–200ms for low-latency streaming—great for conversational apps, agents, and games.
- Test 24-language support and mid-sentence switching if your ElevenLabs setup relies on multilingual output.
Clone a voice with minimal input
- Use a 5-second recording (or more, if you have it) to create a clone.
- Compare the clone to your existing ElevenLabs voice: timbre, prosody, and stability across different prompts.

This baseline lets you see whether LMNT can match or exceed your current quality and responsiveness.

2. Claim your 500,000 free migration credits

Once you know LMNT can handle your use case, claim the migration grant so you can mirror real usage patterns.

Collect your ElevenLabs details
- Your ElevenLabs account email and organization/workspace name.
- A quick summary of your usage (e.g., “~2M chars/month, streaming TTS for customer support agent,” “voice clones for game NPCs,” etc.).
- Any language and voice requirements (e.g., English + Spanish, broadcaster style, tutor style).
Contact LMNT for migration credits
- Sign in to LMNT and use the contact or support channel (e.g., “Contact us” from the site, or the email/support link in your dashboard).
- In your message, include:
  - Subject line: “ElevenLabs migration – request for 500,000 credits”
  - Your LMNT account email.
  - Your ElevenLabs details and approximate monthly character usage.
  - A short description of what you’re migrating (agents, games, tutors, etc.).
LMNT reviews and applies the grant
- The LMNT team verifies your migration scenario and adds 500,000 characters of migration credits to your account as a one-time grant.
- You’ll see the credits reflected in your usage/limits, ready for integration testing and side-by-side runs.

3. Map ElevenLabs features to LMNT

Next, map the way you currently use ElevenLabs to LMNT’s capabilities. The goal is feature parity or better, without regressions in behavior.

Voices and styles
- Identify your current ElevenLabs voices (e.g., “narrator,” “support agent,” “character A/B”).
- In LMNT:
  - Use built-in voices like Brandon (for broadcast-style), Tyler (smooth storyteller), etc.
  - Create voice clones for your branded or character voices using 5+ seconds of your original reference.
- For each voice, test:
  - Neutral reading.
  - Emotion or emphasis ranges.
  - Different content types (dialog, UI messages, long-form).
Streaming vs. non-streaming behavior
- If you’re using ElevenLabs’ streaming endpoints:
  - Switch to LMNT’s streaming TTS API for your interactive surfaces.
  - Confirm end-to-end latency is within your budget (LMNT targets 150–200ms streaming).
- For batch or offline generation:
  - Map those calls to LMNT’s standard generation API (non-streaming).
Multilingual and code-switching
- LMNT supports 24 languages and can switch mid-sentence like people do.
- If your ElevenLabs setup uses multiple languages or code-switching:
  - Recreate those prompts in the LMNT Playground.
  - Validate accent, pronunciation, and smooth transitions between languages.
Rate limits and concurrency assumptions
- LMNT has no concurrency or rate limits.
- If you’ve previously implemented throttling, careful queueing, or backoff strategies to avoid ElevenLabs limits, plan to simplify or remove those once you’re confident in LMNT’s behavior.
- You can still keep circuit breakers, but you don’t need to protect against artificial ceiling limits.

4. Swap API calls and run side-by-side tests

Now you’re ready to implement LMNT in your codebase and run it alongside ElevenLabs.

Review LMNT’s API spec
- Go to https://api.lmnt.com/spec.
- Identify:
  - Auth pattern (API keys / headers).
  - Endpoints for streaming and non-streaming TTS.
  - Parameters for voice IDs, languages, and configuration.
Create a minimal LMNT integration path
- In your backend or client service that currently calls ElevenLabs:
  - Add a new LMNT client using your preferred language/stack.
  - Mirror the high-level interface (e.g., speak(text, voice, language)).
- For streaming:
  - Use WebSockets or HTTP streaming (depending on your stack) to pipe LMNT’s audio to the client as it arrives.
Implement a dual-provider toggle
- Add a feature flag or config that selects between ElevenLabs and LMNT:
  - e.g., TTS_PROVIDER=elevenlabs or TTS_PROVIDER=lmnt.
- This lets you:
  - Run A/B tests across a subset of users.
  - Roll back quickly if you see an unexpected issue.
  - Compare performance metrics in production-like conditions.
Run migration tests using your 500,000 credits
- Use your migration credits for:
  - Load tests that simulate peak traffic with LMNT’s streaming endpoints.
  - Side-by-side comparisons: same prompt → ElevenLabs vs LMNT → subjective and objective scoring.
  - Latency and stability measurements for conversational agents.
- Track:
  - Median and p95 latency.
  - Error rates and reconnect behavior.
  - User feedback where applicable (e.g., “voice sounds more natural,” “less lag”).

5. Cut over and scale with LMNT

Once you’re confident, you can flip the switch.

Set LMNT as the default provider
- Update your config so TTS_PROVIDER=lmnt for all or most traffic.
- Keep the feature flag in place initially, so you can revert or run controlled rollouts.
Clean up ElevenLabs-specific logic
- Decommission:
  - ElevenLabs-specific throttling, queueing, or retry logic.
  - Any ElevenLabs-only SDK wrappers that aren’t needed.
- Update observability dashboards to track LMNT-specific metrics.
Plan for ongoing scale
- LMNT is built to scale with:
  - No concurrency or rate limits.
  - Affordable pricing that improves with volume.
  - Enterprise plans when you need custom terms or SLAs.
- If you’re a startup or rapidly growing team:
  - Consider LMNT’s Startup Grant (separate from migration credits) for longer-term development runway.

Common Mistakes to Avoid

Treating migration as a simple “endpoint swap”:
LMNT offers 150–200ms low-latency streaming, different voice options, and no rate limits. Don’t just change URLs—update your latency budgets, remove unnecessary throttles, and retune prompts/prosody for best results.
Ignoring voice clone re-tuning:
Copying existing ElevenLabs reference audio 1:1 might not be optimal. With LMNT, you often get studio-quality clones from as little as 5 seconds—take time to choose cleaner, more representative samples and test across different content types.

Real-World Example

Say you run a real-time customer support agent that currently uses ElevenLabs for streaming TTS. The agent handles 300 concurrent sessions at peak, speaks English and Spanish, and uses cloned voices to match your brand.

You sign up for LMNT, hit the Playground, and confirm that a cloned voice plus LMNT’s Brandon and Leah voices can cover your use cases. You contact LMNT, share your ElevenLabs usage, and receive 500,000 migration credits. Your team then:

Implements a small LMNT client service based on https://api.lmnt.com/spec, with a feature flag to switch providers.
Runs a two-week A/B test with 20% of sessions using LMNT streaming. You measure:
- Lower end-to-end latency thanks to 150–200ms streaming.
- Fewer “talk over” incidents in conversations.
- Positive subjective feedback on voice warmth and clarity.
Uses the free credits to stress test 500+ simulated concurrent agents—no rate limits or throttles are triggered.

After the test, you set LMNT as the default TTS provider, remove your previous concurrency guardrails, and keep LMNT’s voices in your design system as first-class brand assets.

Pro Tip: When you run side-by-side tests, log audio identifiers and TTS provider metadata per session. That way, if users say “the new voice sounds better/worse,” you can trace exactly which provider and configuration generated that audio—and iterate quickly on LMNT voice selection or cloning.

Summary

Migrating from ElevenLabs to LMNT is straightforward if you treat it as more than an endpoint swap. Start by validating voices and latency in the Playground, then claim your 500,000 free migration credits so you can run real, production-like tests without double-paying. Map your existing voices and multilingual use cases to LMNT’s streaming API, create or improve your voice clones from short recordings, and use a feature flag to run clean side-by-side comparisons. Once you’re satisfied, flip your traffic to LMNT and rely on 150–200ms streaming, 24 languages, and no concurrency or rate limits to support your agents, apps, and games at scale.

Next Step

Get Started