LMNT pricing: should I use Indie, Pro, or Premium for a real-time voice agent, and how do overages work per 1K characters?

Most teams choosing between LMNT’s Indie, Pro, and Premium plans for a real-time voice agent care about two things: will it keep up with conversation, and what will it actually cost once users start talking a lot. The good news is that latency and quality are the same across plans—what really changes is how many characters you get included and how overages are billed per 1K characters.

Quick Answer: Use Indie for early prototypes and low-traffic agents, Pro when you’re moving into production with hundreds or low thousands of users, and Premium when you expect sustained or spiky traffic and need room to scale. All paid plans bill overages per 1K characters, so once you estimate your characters per minute and concurrent users, you can map your agent’s cost with predictable, linear pricing.

Why This Matters

A real-time voice agent lives or dies on two axes: latency and unit economics. Latency is solved by LMNT’s stack—150–200 ms streaming, no concurrency or rate limits, 24 languages, and 5-second voice cloning. Unit economics is on you: pick the wrong plan and you either throttle growth because you’re afraid of overages, or overpay for a tier you don’t need.

When you understand how LMNT pricing maps to characters per minute, session volume, and overage per 1K characters, you can forecast costs as you grow from a single demo to production traffic—without compromising on voice quality or real-time performance.

Key Benefits:

Predictable scaling: Character-based pricing and per-1K overages make it easy to forecast cost per session and per user.
No hidden limits: No concurrency or rate limits means your choice of plan is about volume and price, not technical throttles.
Production-ready from day one: The same low-latency streaming and studio-quality voice clones apply across Indie, Pro, and Premium—so you don’t “upgrade” for quality, only for economics.

Core Concepts & Key Points

Concept	Definition	Why it's important
Characters per 1K	LMNT bills usage in blocks of 1,000 characters of generated speech (roughly 8–10 seconds of spoken audio in English).	This is the base unit for both included usage and overages, so it’s how you estimate cost per turn, per session, and per user.
Included character allowance	The number of characters your plan includes each billing period before overages kick in.	Determines whether your current traffic fits comfortably in Indie, Pro, or Premium, and how sensitive you’ll be to spikes.
Overage rate per 1K	The marginal price you pay for each additional 1,000 characters beyond your included allowance.	This is your true marginal cost of growth—what each extra conversation, user, or hour of voice actually costs.

Note: LMNT’s website lists the exact included character counts and overage rates for Indie, Pro, and Premium. Use those numbers with the math below to size your plan.

How It Works (Step-by-Step)

Think about pricing for a real-time voice agent in three layers: session usage → monthly volume → plan + overages.

1. Model your agent’s character usage

Start with a realistic session profile:

Estimate speech rate and duration.
- Typical English speech: ~140–180 words per minute.
- Average word length: ~4–5 characters (plus spaces and punctuation).
- Rule-of-thumb: 1K characters ≈ 8–10 seconds of speech.
Define a “typical session.”
For a conversational agent:
- Session length: 3–5 minutes.
- LMNT talking time: maybe 40–60% of that (user is speaking the rest of the time).
- Example: 4-minute session × 50% talking time = 2 minutes of LMNT speech.
Convert to characters per session.
- 2 minutes of speech ≈ 12–15K characters.
- So: 1 session ≈ 12–15 blocks of 1K characters.

This gives you a baseline: how many 1K-character blocks does an average user-session consume?

2. Multiply by expected traffic

Now map that to monthly usage:

Estimate active users and sessions.
- Early stage: 50–200 users, 1–3 sessions per day.
- Post-launch: 1K–10K users, heavily skewed by power users.
Compute monthly sessions.
Example:
- 300 daily sessions × 30 days = 9,000 sessions / month.
Compute monthly character usage.
- 9,000 sessions × 15K chars/session = 135M characters / month.
- In blocks of 1K: 135M / 1,000 = 135,000 units of 1K characters.

This is the number you compare against each plan’s included characters.

3. Map to Indie vs Pro vs Premium

With your monthly 1K-character units in hand:

Check included characters for each plan.
- Indie: best for low-volume prototyping and early pilots.
- Pro: tuned for mid-volume production—where you have real users, but not yet consumer-scale traffic.
- Premium: designed for high-volume, multi-agent, or multi-product use, where overages on lower tiers would dominate your bill.
Estimate overages.
- Overages are billed per additional 1K characters above your plan’s included amount.
- Total monthly cost ≈ base plan price + (overage units × overage price per 1K).
Pick the tier where:
- Your base usage fits within the included characters, and
- Expected spikes don’t push you into a scenario where overages cost more than simply moving up a tier.

If you find yourself consistently paying high overages on Indie, it’s almost always cheaper (and simpler) to move to Pro; same logic from Pro → Premium as you reach larger scale.

Choosing Between Indie, Pro, and Premium for a Real-Time Voice Agent

Use this as a rule-of-thumb decision tree based on production reality, not just plan names.

When Indie makes sense

Use Indie if:

You’re in prototype or pre-launch mode.
Traffic looks like:
- < ~5–10K monthly sessions, or
- You’re mostly doing internal demos and UX testing.
Main goals:
- Validate turn-taking and latency.
- Try 24 languages and code-switching.
- Test voice cloning with your own 5-second recordings.

Indie is ideal when your monthly character usage is well below the included allowance and overages, if any, are minor.

When Pro is the sweet spot

Move to Pro once:

You’ve shipped a real product: users expect it to work 24/7.
Traffic looks like:
- Tens to hundreds of thousands of monthly sessions, or
- You’re onboarding partners/tenants building on your agent.
You want:
- Predictable cost per 1K characters at meaningful scale.
- Enough headroom that your growth doesn’t immediately trigger large overage bills.

For most real-time voice agents with steady user growth, Pro is the default “production” tier: you get the same low-latency streaming and studio-quality clones as higher tiers, but pricing is optimized for the scale most SaaS/agent apps hit in their first 6–18 months.

When Premium is the right call

Choose Premium if:

You’re running:
- Multiple high-traffic agents, or
- A platform where others build agents on top of you (B2B2C, marketplaces, dev tools).
Traffic looks like:
- Hundreds of thousands to millions of monthly sessions, or
- You need to absorb traffic spikes without worrying about overage surprises.
You’re ready for:
- Enterprise-style procurement and security reviews (SOC-2 Type II, etc.).
- Negotiated volume economics as usage grows.

At this level, the core question is not “Can the infrastructure keep up?” (LMNT has no concurrency or rate limits), but “What’s the most efficient way to buy 10× more characters?” Premium is built for that.

How Overages Work Per 1K Characters

Overages are simpler than most TTS platforms; there are no surprise “burst” or concurrency surcharges.

Here’s the model:

Usage is tracked per character.
Every character you send to LMNT for synthesis counts toward your monthly total. This includes:
- Narrative content
- System prompts rendered as speech
- Short acknowledgements (“Got it,” “One sec,” etc.)
Billing unit = 1,000 characters.
You’re charged in increments of 1K characters:
- If you generate 12,400 characters over your included amount, that’s 12.4 units of 1K.
- Billing rounds based on LMNT’s pricing rules (typically to the nearest full 1K unit; check your dashboard for exact mechanics).
Overage rate is plan-specific.
- Indie has an overage price per additional 1K characters.
- Pro’s overage rate per 1K is usually lower than Indie’s.
- Premium’s rate per 1K is typically the most favorable for sustained high volume.
Monthly invoice = base + overages.
- Base: your fixed plan subscription.
- Variable: (total characters – included characters) ÷ 1,000 × overage rate.

Because LMNT doesn’t impose concurrency or rate limits, you’re not paying more just because you have spikes; you’re paying proportionally to characters actually spoken.

Quick back-of-the-envelope math

To size your plan, you can do a fast estimation:

1K characters ≈ 8–10 seconds of speech.
1 minute of speech ≈ 6–7K characters.
10 minutes of speech ≈ 60–70K characters.

If your average user hears ~3 minutes of LMNT speech per session, that’s roughly 18–21K characters per session, or 18–21 units of 1K. Multiply by your expected monthly sessions and compare to your plan’s included characters.

Common Mistakes to Avoid

Ignoring characters per response.
How to avoid it: Instrument your agent to log characters per turn and per session from day one. Use this data rather than guesses to select a plan and monitor overage risk.
Staying on Indie too long because it “works.”
How to avoid it: If overages on Indie or Pro become a meaningful share of your bill (e.g., consistently >20–30% of total), run the math on the next tier. Higher plans with lower per-1K rates often reduce your total cost at scale.
Optimizing only for text tokens, not speech characters.
How to avoid it: Don’t just look at your LLM usage. Your LLM might be short and efficient, but if your prompts encourage verbose spoken responses, your TTS bill will climb. Tune your prompting for concise, conversational replies when that makes sense.

Real-World Example

Let’s say you’re shipping a customer support voice agent with LMNT as the TTS layer:

Average session:
- 5 minutes total.
- LMNT speaking ~2.5 minutes (rest is user speech and thinking time).
Characters per session:
- 2.5 minutes × ~7K chars/min ≈ 17.5K characters.
- That’s ~17.5 “units” of 1K characters.

Now traffic:

200 daily sessions at launch → 6,000 sessions / month.
6,000 × 17.5K ≈ 105M characters / month.
In 1K units: ~105,000.

How to choose a plan:

If Indie’s included characters are far below 105M, you’ll pay significant overages—Indie is fine for early testing, but not this steady production load.
Move to Pro:
- If Pro includes, say, ~100M characters, you’re right at the edge.
- Overages are modest (5M characters / 1K × Pro overage rate).
- You get predictable economics and plenty of room to iterate.
If you grow to 10× traffic (60,000 sessions / month ≈ 1.05B characters), the overages on Pro become substantial. That’s the point to:
- Switch to Premium, where the per-1K price is lower, or
- Talk to LMNT about a custom enterprise plan for better volume economics.

Pro Tip: Log and export characters per synthesis request from your backend. After a week of real traffic, you’ll know your true average characters per session and can plug that into each plan’s included characters and overage rate. That’s the fastest way to choose between Indie, Pro, and Premium with confidence.

Summary

For a real-time voice agent, the real decision between Indie, Pro, and Premium isn’t about latency or features—those stay consistently strong across plans with 150–200 ms low-latency streaming, studio-quality 5-second voice clones, and 24 languages. It’s about matching your monthly characters and growth curve to the right mix of included usage and overage rate per 1K characters.

Indie is your prototyping and early-pilot tier.
Pro is the default for most production agents with real user traffic.
Premium (or enterprise) is where you land once you’re running multiple agents or platform-scale usage and want the best marginal cost per 1K characters.

Instrument your app, measure characters per session, project monthly usage, then line that up against each plan’s numbers. That’s how you get a voice experience that feels instantaneous to users and stays predictable on your cloud bill.

Next Step

Get Started