LMNT pricing

About

LMNT builds a low-latency AI text-to-speech (TTS) API aimed at developers shipping real-time voice experiences — voice agents, interactive applications, and audio content generation. The core product is a speech-synthesis API with voice cloning, fronted by a free in-browser Playground for trying the models before integrating. LMNT positions itself on speed and a developer-friendly, no-rate-limit experience, competing with TTS API providers such as ElevenLabs, Cartesia, PlayHT, and Resemble AI.

The buyer is a developer or small product team integrating speech into their own application, with a self-serve path from a free tier up through three paid subscription tiers and a sales-led enterprise plan. LMNT advertises SOC-2 Type II compliance, signaling it courts production and business customers, not just hobbyists.

LMNT was founded by Sharvil Nanavati — who previously originated and led the team behind Google Glass — and raised a seed round in mid-2023 led by Elad Gil and Sarah Guo (Conviction). It is an early-stage, privately held company that positions itself on sub-300ms streaming latency and a developer-first, no-rate-limit experience. Publicized partnerships include Khan Academy, where LMNT’s speech models power voices in the Khanmigo AI tutor. Detailed ARR, valuation, and headcount figures are not publicly disclosed; the company remains a small, focused TTS-API specialist competing against larger voice-AI players.

Pricing summary : Freemium TTS tiers plus per-1,000-character overage

LMNT uses a freemium subscription model with a usage meter built on two dimensions:

1. Monthly subscription tier (the included character pool): Free at $0 (15K characters), Indie at $10/mo (200K characters), Pro at $49/mo (1.25M characters), and Premium at $199/mo (5.7M characters). The plan fee buys a fixed monthly allotment of synthesized characters.
2. Per-character overage (the usage component): Once the included pool is exhausted, paid plans bill per 1,000 characters. The rate drops as you move up tiers — $0.05/1K on Indie, $0.045/1K on Pro, and $0.035/1K on Premium. The Startup Grant plan bills overage at $0.028/1K. (The Free tier has no overage path — it is capped at its 15K allotment.)

LMNT also surfaces “credits” terminology on its Startup Grant page (45M credits over three months), where credits map to characters. This makes it a hybrid pricing model where the subscription and the character-based usage meter are both meaningful.

What makes this different: Every paid tier advertises “no concurrency or rate limits” and unlimited voice clones — LMNT competes by removing the throughput caps and per-voice fees that rival TTS APIs commonly meter, and lets the included-character pool plus a declining overage rate do all the price discrimination.

Pricing by product

Text-to-speech API (self-serve plans)

Playground (free web tool)

Startup Grant (sales-reviewed program)

Sales motions across products: PLG / self-serve for Free, Indie, Pro, and Premium (sign up and buy online); application-reviewed for the Startup Grant; sales-led for the Enterprise plan grant recipients roll into.

Per-character overage rates by tier

The included-character pool plus a declining per-character overage rate is the entire price-discrimination mechanism: heavier users buy up into a higher tier to get both more bundled characters and a cheaper marginal rate. Note: the in-app checkout at app.lmnt.com is bot-protected and could not be captured, so any plan or add-on shown only inside the authenticated billing flow is unknown.

Hidden costs : What heavy-volume TTS users actually pay per month

LMNT’s pricing looks clean on the page, but the real bill is driven entirely by how far past your plan’s included character pool you run — and the smart move is rarely to stay on the tier that “fits” your headline volume. Two archetypes show where the money actually goes.

Archetype 1 — an indie voice-agent builder on Indie ($10/mo) running 1,000,000 characters/month. That is well past the 200K Indie pool, so 800K characters bill at $0.05/1K.

At 1M characters/month the buyer is better off on Pro ($49 flat, since 1M is inside Pro’s 1.25M pool) than on Indie ($50 with overage) — the included pool, not the headline price, decides the cheapest tier. The lesson: LMNT’s declining overage rate plus rising pool means you should price every tier at your real volume, not pick the cheapest base fee. This is the classic prepaid-credits / included-pool trade-off where the marginal rate and the bundled allotment must be modeled together.

Archetype 2 — a production app on Pro ($49/mo) generating 3,000,000 characters/month. Pro includes 1.25M, so 1.75M characters bill at $0.045/1K.

Here Pro-with-overage ($127.75) actually beats Premium-flat ($199) because 3M characters sits comfortably below Premium’s 5.7M pool but the Pro overage rate is cheap enough that paying for usage beats pre-buying the bigger bundle. The break-even only flips toward Premium once monthly volume approaches the point where Pro overage would exceed the $150 step up to Premium — roughly 4.6M+ characters. The takeaway: with characters as the single value metric, LMNT buyers should model overage against the next tier’s flat fee every month, because the cheapest plan changes with volume.

Want to estimate your own LMNT bill? Use the LMNT pricing calculator to model your monthly cost based on character volume and plan tier.

Pricing evolution : How LMNT’s TTS tiers and rates have changed over time

LMNT’s pricing page has been rebuilt twice in its short history, and each rebuild moved decisively toward cheaper, simpler, more developer-friendly terms. The arc, reconstructed from Wayback snapshots of lmnt.com/pricing: a 2023 three-tier “keep it simple” page → a 2024 four-tier ladder that introduced per-clone fees → a late-2024/early-2025 repricing that removed clone fees, slashed overage, raised allotments, and added a no-rate-limit guarantee → a 2025 Startup Grant bolt-on.

Cadence

Tracked range: 2023-09–2026-06. Quarters not listed above were verified stable (0 price changes, 0 SKU additions) across the captured Wayback snapshots.

Notable changes

• 2023-09 → 2023-12 — Stable three-tier model: Free ($0, 10K chars, non-commercial, attribution required), Pro ($7/mo, 30K chars, $0.20/1K overage, $10/mo per extra professional clone), Enterprise.
• 2024-04 — Repackaged to four tiers (“Plans that scale with you”): Indie ($17/mo) and Premium ($199/mo) added, Pro repriced to $49/mo, instant voice clones introduced, Free pool raised to 15K characters, overage cut to $0.17/1K, and the free tier’s non-commercial/attribution requirement dropped.
• 2024-09 — Per-professional-clone fee ($10/mo each) eliminated; every tier, including Free, now advertises unlimited instant and professional voice clones.
• ~2025-01 — Major repricing: Indie cut from $17 to $10/mo; overage slashed (Indie $0.17→$0.05, Pro →$0.045, Premium $0.15→$0.035 per 1K); Pro allotment raised 500K→1.25M and Premium 1.5M→5.7M; “no concurrency or rate limits” added to every paid card; SOC-2 Type II badge surfaced.
• 2025-08/09 — Startup Grant launched (45M credits over 3 months, ~15M characters/month, $0.028/1K overage) and surfaced from the pricing page via an “Explore special pricing for startups” link.

The 2025 repricing in detail

The single most consequential change in LMNT’s history is the late-2024/early-2025 repricing, visible between the 2024-09 and 2025-01 Wayback snapshots. It moved on four dimensions at once:

• Overage collapsed ~70% on Indie: $0.17 → $0.05 per 1,000 characters. Pro landed at $0.045 and Premium fell from $0.15 to $0.035 — making LMNT’s marginal character rate among the most aggressive of the TTS API peers it competes with.
• Included pools grew sharply: Pro went 500K → 1.25M characters (2.5×) and Premium 1.5M → 5.7M (3.8×) for the same flat fee, a large effective discount for committed buyers.
• Indie’s base price dropped: $17 → $10/mo, lowering the entry point to a paid plan.
• A throughput guarantee replaced a throughput upsell: every paid card began advertising “no concurrency or rate limits,” removing the rate-tier ladder rival APIs commonly monetize and leaving the character pool as the sole price-discrimination lever.

Taken together, the repricing reframed LMNT from a conventional metered-clone TTS vendor into a volume-friendly, no-throttle developer API — a deliberate repositioning against ElevenLabs and Cartesia on price and predictability rather than feature count.

What’s unique : No-rate-limit, unlimited-voice-clone TTS pricing

1. No concurrency or rate limits on paid tiers. Every paid LMNT plan advertises “no concurrency or rate limits” as an explicit line item — removing the throughput-tier upsell that rival TTS APIs commonly meter. For a real-time voice-agent builder, unmetered concurrency is the difference between a predictable bill and a surprise at scale, and LMNT turns it from a hidden constraint into a headline promise. This collapses the value metric down to a single dimension: synthesized characters.

2. Unlimited voice clones on every tier — even Free. LMNT does not meter or cap voice clones at any tier, including the $0 Free plan. This is a notable reversal: as recently as 2024-04 the company charged $10/mo per extra professional clone and capped clone counts per tier (2 on Free, 30 on Premium). By removing per-clone fees entirely it eliminated a whole billing dimension, betting that character volume alone is a cleaner, more defensible value metric than a clone-slot count buyers struggle to predict.

3. The character pool is the only price-discrimination lever. With clones unlimited and throughput unmetered, LMNT’s tiers differ on exactly two numbers: the included character allotment and the per-1,000-character overage rate. The overage rate falls as you climb ($0.05 → $0.045 → $0.035), baking volume discounting directly into the ladder instead of negotiating it — a clean hybrid usage model that a developer can fully reason about from the pricing page alone.

4. A declining overage rate that rewards commitment without a contract. Most usage-based vendors reserve their cheapest marginal rate for negotiated annual commits. LMNT instead publishes a lower per-character rate at every higher self-serve tier and an even lower $0.028/1K inside the Startup Grant — substituting a transparent published ladder for sales-led volume discounting, which keeps the entire motion self-serve.

5. A free Playground decoupled from the billing funnel. LMNT’s in-browser Playground (“Our Playground is free … just give us a shout out when you share”) lets developers test the speech models with no billing signup, separate from the API free tier’s 15K-character cap. This splits “try the model” from “integrate and meter,” lowering the barrier to first contact before any character is metered.

Strengths & weaknesses

Billing UX : Self-serve checkout, included-character pools, and grant applications

• In-page “Buy Now” checkout — each paid tile (Indie, Pro, Premium) has a direct “Buy Now” button, and the Free tier a “Get Started” button, taking the buyer into the app.lmnt.com billing flow. (That authenticated checkout surface is bot-protected and was not captured; controls inside it are unknown.)
• Included-character allotment per plan — the bill is structured around a fixed monthly character pool (15K / 200K / 1.25M / 5.7M) shown on every plan card, so buyers can size their tier against expected synthesis volume.
• Per-1,000-character overage display — each paid card states its overage rate (“$0.05 per 1K characters after”, etc.), making the marginal cost above the pool explicit on the pricing page itself.
• “No concurrency or rate limits” guarantee — surfaced as a line item on every paid tile, removing throughput-tier upsell as a billing dimension.
• Free Playground — a no-signup-billing in-browser tool (“Go to Playground”) lets developers test the speech models before they ever hit checkout.
• Startup Grant application flow — a separate “Apply Now” path with a documented 4-step lifecycle (Apply → ~1-week review → register for the startup plan → auto-upgrade to enterprise after 3 months), gating 45M free credits behind a <20-employee eligibility check.

Strategic wins : Pricing decisions that won developer adoption

1. Removing rate limits as a competitive wedge

LMNT turns “no concurrency or rate limits” into a headline benefit on every paid card, betting that developers building real-time voice agents will pay for predictable throughput over a cheaper-looking metered tier. For conversational and agentic workloads — where concurrency spikes are the norm and a rate-limit error is a dropped call — this removes the single most common source of a surprise bill or a degraded user experience. It is a textbook case of reframing the value metric so the buyer pays for the one thing they can predict (characters) and gets the unpredictable thing (concurrency) for free.

2. Free tier with unlimited voice clones as an adoption funnel

A $0 tier that still grants unlimited voice clones lowers the barrier to experimentation before the 15K-character cap forces an upgrade — and the decoupled free Playground lets developers hear the models before they ever create an account. This is a deliberate product-led / self-serve motion: give away capability (clones, model access) generously while keeping volume tight, so the buyer hits the paywall only once they’re already integrated and committed. The 2024 removal of per-clone fees made this funnel materially stronger.

3. Startup Grant as a top-of-funnel program

The 45M-credit grant (15M characters/month for three months, ~330 hours of audio) pulls early-stage startups onto the platform at zero cost, then auto-upgrades them to an enterprise plan after the grant period. By gating it to sub-20-employee teams and excluding prior recipients and one-off projects, LMNT targets durable design-ins rather than free-credit tourists — and the $0.028/1K grant overage rate, lower than any published tier, encourages real production usage during the trial. It is a logo-and-conversion play that mirrors the entitlement-grant patterns increasingly common in usage-based AI businesses.

Areas to improve : Gaps in LMNT’s pricing transparency and packaging

1. Publish enterprise pricing guidance

Enterprise pricing is entirely opaque — even grant recipients only learn they “auto-upgrade to an enterprise plan with the same usage limits,” with no published rate. A concrete fix: publish a starting volume band (e.g., “Enterprise starts around X million characters/month at $Y/1K”) or expose a self-serve volume calculator, the way the published self-serve overage ladder already lets buyers reason about cost. Doing so would let high-volume teams qualify themselves before talking to sales, shortening the cycle and reducing the friction of jumping from a transparent self-serve page to a black-box quote.

2. Surface the in-app billing controls publicly

Key billing controls — spend caps, usage alerts, invoicing, and the exact mechanics of how the included pool resets — live behind the bot-walled app.lmnt.com checkout and are not visible on the marketing surfaces. The fix is to document the billing UX publicly (a help-center page or a screenshot tour), since buyers evaluating an overage-based usage model need to know they can set a hard cap before they commit. Right now a cautious finance buyer cannot confirm spend controls exist without first signing up.

3. Add an overage path to the Free tier

The Free tier hard-caps at 15K characters with no pay-as-you-go option, forcing an abrupt jump to a $10/mo Indie subscription the moment a developer’s prototype gains traction. A small fix — letting Free users add a card and pay Indie-rate overage above 15K — would smooth the conversion curve and capture the “just over the line” prototypes that currently stall. As it stands, the cliff between Free and Indie is a packaging gap in an otherwise smooth self-serve funnel.

Key takeaways

1. Bundle a usage pool, then meter overage. LMNT pairs a flat monthly fee with an included character allotment and a per-1,000-character overage rate. This hybrid keeps revenue predictable for the vendor while letting the bill scale with the buyer’s actual usage — the cleanest structure when cost is driven by a single, measurable unit.
2. Make the marginal rate fall as tiers rise. Overage drops from $0.05 to $0.035 per 1K characters up the ladder. Baking the volume discount into the published tiers (rather than reserving it for negotiated commits) keeps the entire motion self-serve and lets buyers self-select into the right tier without a sales call.
3. Turn removed limits into a selling point. “No concurrency or rate limits” is framed as a benefit, not a silent default. When you stop metering a dimension your competitors monetize, say so loudly — it becomes a differentiator instead of an invisible win.
4. Keep the free tier generous on capability, tight on volume. Unlimited voice clones but only 15K characters lets developers fully evaluate the product’s quality while ensuring any real workload trips the paywall. Generous on capability, tight on volume is the durable free-tier shape.
5. Use a grant program as a funnel, not a giveaway. The Startup Grant trades free credits for early-stage logos and eventual enterprise conversion, with eligibility rules (sub-20 employees, no repeat recipients) that filter for durable design-ins over free-credit tourism.

UBP implications

1. Character-metered TTS favors a hybrid model. With cost driven by synthesized text volume, an included-pool-plus-overage structure aligns price with usage while keeping revenue predictable — and because the unit (characters) maps directly to the underlying inference cost, the meter stays honest as input prices fall.
2. Declining marginal rates can replace explicit volume discounts. LMNT bakes volume discounting into the published tier ladder rather than negotiating it per deal. For self-serve UBP businesses, a transparent declining rate is a powerful substitute for a sales-led discount table — it scales without headcount.
3. Removing throughput caps reframes the value metric. When concurrency and rate limits are unmetered, the character pool becomes the sole value metric, sharpening buyer comprehension and removing a whole axis of pricing complexity. The lesson for UBP teams: fewer metered dimensions usually beats more, as long as the one you keep is the one that tracks value.

Sources

• LMNT pricing page (accessed 2026-06-04)
• LMNT Startup Grant page (accessed 2026-06-04)
• LMNT API / developer site (accessed 2026-06-04)

Bottom line

LMNT prices its low-latency text-to-speech API as a clean hybrid: a flat monthly subscription buys an included character pool (15K Free → 5.7M Premium), and usage beyond it bills per 1,000 characters at a rate that falls from $0.05 to $0.035 as you climb the ladder — with unlimited voice clones and no concurrency or rate limits on every paid tier. Its short pricing history is a study in deliberate simplification: a 2023 three-tier page with per-clone fees and a non-commercial free tier gave way, by 2025, to a cheaper, no-throttle, single-value-metric model that competes with ElevenLabs and Cartesia on price predictability rather than feature count. For any team metering a single usage unit, LMNT is a compact case study in stripping a pricing model down to the one dimension that tracks value.

Want to compare LMNT against other text-to-speech and voice AI pricing? Browse the pricing blueprint.

Pricing timeline : Major events on a vertical axis

Each milestone below corresponds to a public pricing change, product launch, or material adjustment. Major events use a filled marker; minor adjustments use a faded one.

Current four-tier freemium TTS pricing with per-character overage

Jun 2026

LMNT lists Free ($0, 15K characters), Indie ($10/mo, 200K characters), Pro ($49/mo, 1.25M characters), and Premium ($199/mo, 5.7M characters), each with per-1,000-character overage from $0.05 down to $0.035, no concurrency or rate limits, and unlimited voice clones. Structure unchanged from the 2025-01 repricing.

Startup Grant launched and surfaced from the pricing page

Sep 2025

By 2025-08/09 LMNT had launched the Startup Grant (45M credits free over 3 months, ~15M characters/month, $0.028/1K overage, sub-20-employee eligibility, auto-upgrade to enterprise after 3 months) and added an 'Explore special pricing for startups' link plus a 'Startup Grant' nav item to the pricing page. Self-serve tier prices were unchanged from the 2025-01 repricing.

Major repricing: overage slashed, allotments raised, no rate limits, SOC-2

Jan 2025

Between 2024-09 and 2025-01 LMNT cut prices sharply: Indie dropped from $17 to $10/mo; overage fell from $0.17 to $0.05/1K (Indie), $0.045/1K (Pro), and from $0.15 to $0.035/1K (Premium). Pro's allotment rose from 500K to 1.25M characters and Premium's from 1.5M to 5.7M. 'No concurrency or rate limits' was added to every paid card, a free Playground hero was added, and a SOC-2 Type II badge appeared. (Wayback 2025-01.)

Per-clone fees removed — unlimited voice clones on every tier

Sep 2024

By 2024-09 LMNT dropped the $10/mo per-professional-clone fee and the metered clone counts: every tier, including Free, now advertised 'Unlimited instant voice clones' and 'Unlimited professional voice clones.' Subscription prices and character allotments were unchanged from 2024-04 ($17 Indie / $49 Pro / $199 Premium; overage $0.17–$0.15/1K).

Repackaged to a four-tier 'Plans that scale with you' ladder

Apr 2024

By 2024-04 LMNT replaced the single Pro plan with a four-tier ladder: Free ($0, 15K characters), Indie ($17/mo, 200K, $0.17/1K overage, 2 pro clones), Pro ($49/mo, 500K, $0.17/1K, 10 pro clones), and Premium ($199/mo, 1.5M, $0.15/1K, 30 pro clones). Instant voice clones were introduced alongside professional clones; extra professional clones cost $10/mo each. Free tier became commercial-friendly (attribution requirement dropped).

Three-tier 'keep it simple' model: Free, $7/mo Pro, Enterprise

Sep 2023

Earliest captured pricing (Wayback 2023-09 and 2023-12): Free ($0, 10K characters/mo, 2 professional voice clones, non-commercial license, attribution to LMNT required), Pro ($7/mo, 30K characters, 5 professional voice clones, commercial license, $0.20 per 1,000-character overage, additional clones $10/mo each), and Enterprise (contact sales, SLA, dedicated support).