AI Text-to-Speech Tools for Narration, Compared on What an Hour of Audio Actually Costs (2026)

ElevenLabs, Murf, Speechify Studio and WellSaid all sell "text to speech", but they meter it four different ways — by character, by generation time, by the second, and by downloaded minutes only. For video narration, courses and audiobooks, that metering choice matters more than the headline monthly price. This guide compares the published plans on cost per finished hour, commercial terms and language support.

Last updated: November 2026 · Prices in USD, checked against vendor pricing pages in November 2026 · How we compare: this guide is based on publicly available vendor information, not hands-on testing. We have not listened to these tools side by side, and there are no paid placements. Every cost figure below is arithmetic on published prices, shown with its workings.

Affiliate Disclosure: Some links on this page are affiliate links. If you sign up through them, we may earn a commission from the vendor at no extra cost to you. This is how the site is funded. It does not change which products we include or the facts we publish — every price and feature below is taken from the vendors' own public pages.

Quick picks by use case

Best for flexible, character-priced volume

ElevenLabs

from $6/mo

Starter plan. Commercial license starts here; the free plan has none.

  • Priced in credits — roughly 1 credit per character on the main models
  • Japanese included, alongside dozens of other languages depending on model
  • Audiobook and monetized-video use covered on paid plans
Visit ElevenLabs
Best for a video voiceover studio workflow

Murf

from $19/mo

Creator plan, billed annually ($228/year). Monthly billing is $29/mo.

  • Priced by voice generation time: 2 hours/month on Creator
  • Studio editor built around narrated videos and presentations
  • Japanese is on the published language list; commercial rights on paid plans
Visit Murf
Best for steady volume on one annual bill

Speechify Studio

from $100/year

Studio Starter, sold by the year. Not the same product as the Speechify reader app.

  • Priced in seconds: about 24 hours of voiceover a year on Starter
  • Cheapest flat rate at higher volumes in our worked examples below
  • Japanese is on the Studio voiceover language list; commercial rights on paid Studio plans
Visit Speechify Studio
Best for English corporate narration with licensed voices

WellSaid

from $10/mo

Starter plan, billed annually ($120/year). Monthly billing is $19/mo. English voices only on self-serve plans.

  • Only downloaded minutes count — generation and retakes are unlimited
  • Voice catalog built on licensed voice-actor recordings
  • Japanese and other languages are Enterprise-only — see the detail section
Visit WellSaid

Need Japanese narration on a self-serve plan? On published information, your shortlist is ElevenLabs, Murf and Speechify Studio. WellSaid's Japanese voices sit behind its custom-priced Enterprise tier.

First, check the tool still exists

Resemble AI no longer sells voice generation to new customers. In an announcement dated August 13, 2026, the company said it had stopped selling text to speech, voice cloning and related voice products to new customers and is focusing the whole company on deepfake detection. Existing voice customers remain supported, and its open-source voice models remain available as research releases — but a new reader cannot buy it as a narration tool, so it is not in the comparison table. Older comparison articles still list its old voice plans; treat those prices as historical.

Two other names from older TTS roundups are also gone: Play.ht shut down on December 31, 2025 after Meta acquired its team, and LOVO (Genny) went offline after the company filed for Chapter 7 bankruptcy in May 2026. If a guide recommends either as a current buying option, its pricing information is out of date.

Side-by-side comparison

ElevenLabsMurfSpeechify StudioWellSaid
Best forCreators and publishers who want pricing that scales with script length — audiobooks, podcasts, multilingual narrationVideo and presentation voiceovers made in a visual studio editorSteady narration volume paid for once a year — channels, courses, explainersEnglish-language corporate training and brand content, with heavy re-takes
How usage is meteredCredits: about 1 credit per character on Multilingual / v3 / v4 models; Flash and Turbo models cost less per character via the APIVoice generation time (hours per month). Re-rendering an edited block consumes its length againStudio credits: 1 credit per second of voiceover. Dubbing and avatar video burn credits at different ratesDownloaded minutes only. Generation, previewing and retakes are unlimited on paid plans
Free optionYes — 10,000 credits/month, no commercial license, attribution requiredYes — 10 minutes of generation in total, no downloads, no commercial rightsYes — 600 credits, no downloads, no commercial rightsTrial — preview-oriented, no commercial rights
Paid plans fromStarter $6/mo (30,000 credits). Creator $22/mo (121,000). Pro $99/mo (600,000)Creator $19/mo billed annually ($29 monthly) for 2 hrs/month. Business $66/mo billed annually ($99 monthly) for 8 hrs/monthStudio Starter $100/year (86,400 credits/year). Studio Creator $300/year (345,600 credits/year)Starter $10/mo billed annually ($19 monthly) for 240 downloaded min/year. Pro $33/mo billed annually ($49 monthly) for 2,160 min/year
JapaneseYes — Japanese is among the 29 languages of Multilingual v2; newer models cover more languagesYes — Japanese is on the official language list (33 languages per the help center)Yes — Japanese is on the Studio voiceover language listEnterprise only. Self-serve Starter / Pro / Business plans are English voices only
Commercial usePaid plans only (Starter and up). Covers monetized video, podcasts and audiobooks you have the rights to; rights in audio made while subscribed are kept after cancellationPaid plans (Creator and up). Monetized YouTube and streaming use is covered; a Business license is recommended for companiesPaid Studio plans only. The vendor's FAQ grants ownership and perpetual commercial rights for your own projects. The separate reader subscription is personal use onlyPaid plans include full commercial rights. The trial grants none
Broadcast & resale limitsYou cannot resell the service itself, and some outputs (for example sound effects sold as standalone libraries) are excluded. Check each retailer's own AI-narration rules for audiobooksTV and radio broadcasting rights are Enterprise-only. Reselling the voice library, or training AI models on Murf voices, is prohibitedVendor wording is framed around "your own projects" — confirm client-deliverable and resale scope in the terms before relying on itNo self-serve voice cloning; custom voices are an Enterprise conversation. Confirm deliverable and resale scope in your contract
Trade-off to knowCosts follow your script's character count, not the finished minutes — dense scripts cost more for the same runtimeThe clock runs on generation, including re-renders, not just on the final cutAnnual billing and an annual credit pool: the cheap per-hour rate assumes you actually use the allowanceMost expensive per downloaded hour of the four at low volumes — its value appears when you retake a lot

Prices are USD, before tax, as published by the vendors in November 2026. Speechify sells three separate products — a reader app, Studio, and an API — with separate subscriptions; only Studio prices are shown here. Plans change — always confirm the current price and what's included on the vendor's site before you buy.

What one hour of narration costs: a worked example

Because the four tools meter differently, headline prices mislead. Here is the same job priced four ways, using only published figures. The job: one finished 60-minute narration per month, from an English script. Using the vendor's own rule of thumb of roughly 1,000 characters per minute of speech, that script is about 60,000 characters. This is arithmetic, not a listening test.

ToolPlan that covers 60 min/monthWorkingEffective cost
ElevenLabsCreator, $22/mo60,000 characters ≈ 60,000 credits. Starter's 30,000 credits are not enough, so Creator (121,000) is the covering plan$22/mo
MurfCreator, $19/mo billed annually60 minutes fits inside Creator's 2 hours/month of generation time — with room for roughly one full re-render$19/mo ($29 on monthly billing)
Speechify StudioStudio Starter, $100/year86,400 credits/year = 1,440 voiceover minutes/year, an average of 120/month — double this jobabout $8.33/mo, billed annually
WellSaidPro, $33/mo billed annuallyStarter's 240 downloaded minutes/year average only 20/month, so this job needs Pro (2,160/year, about 180/month)$33/mo, billed annually

Scale the same job to four hours a month and the order changes: Speechify Studio Creator ($300/year, about $25/mo for an average of 480 voiceover minutes/month) comes out lowest of the flat plans, then Murf Business ($66/mo billed annually, 8 hours/month), then ElevenLabs Pro ($99/mo, 600,000 credits), then WellSaid Business ($160 per user/month, whose 2,880 downloaded minutes/year match this job's 2,880 minutes exactly).

Two things flip the result in real work. First, retakes: character- and time-metered plans charge again when you re-render, while WellSaid only counts the version you download — a script you render ten times before approving can cost several times its final runtime elsewhere, and exactly its runtime at WellSaid. Second, script density: character metering charges for what is written, not what is heard, so a number-heavy script that reads in 40 minutes can still consume 60 minutes' worth of characters at ElevenLabs, while the time-metered tools charge for the 40.

Developer note: all four vendors also sell API access priced per character (for example, ElevenLabs lists $0.08 per 1,000 characters for its main models and $0.04 for its faster ones; Murf's API lists $0.03 per 1,000 characters for its studio model). On pure arithmetic those rates undercut every subscription above — 60,000 characters works out to a few dollars — but an API gives you no studio editor, and plan conditions differ. Treat API pricing as a separate route for people building narration into a product, not as the price of the tools compared here.

The four tools in detail

ElevenLabs — character-priced, widest language reach of the four

ElevenLabs prices narration in credits, and on its main text-to-speech models one credit is roughly one character. The free plan (10,000 credits/month) is for trying voices: it carries no commercial license, and anything you publish from it needs attribution — attribution does not make commercial use allowed. Starter ($6/mo, 30,000 credits) is where the commercial license begins, which makes it the practical floor for monetized work at low volume — about 30 minutes of narration a month on the standard models. Creator ($22/mo, 121,000 credits) covers roughly two hours, and Pro ($99/mo, 600,000 credits) roughly ten. Language coverage depends on the model: Multilingual v2 covers 29 languages including Japanese, and the newer v3 and v4 models publish much broader lists. Paid plans cover monetized videos, podcasts and audiobooks built from material you have the rights to, and the vendor states that rights in audio generated while subscribed remain after you cancel. What a plan does not settle is the destination's own rules — an audiobook retailer's AI-narration policy, for example, is separate from ElevenLabs' license.

STRENGTHS (per published specs)

  • Lowest entry price for a commercial license here: $6/mo
  • Japanese plus the broadest published language coverage, varying by model
  • Fine-grained steps between plans, plus a per-character API for developers

TRADE-OFFS

  • You pay for characters written, not minutes heard — dense scripts and re-renders consume the allowance faster than the runtime suggests
  • Free-plan audio can never be upgraded into commercial use after the fact
  • Model, language count and per-character cost all vary together, so the "price" depends on which model your project uses
See ElevenLabs plans

Murf — a voiceover studio that charges by the generation clock

Murf is built as a studio: you assemble narration in a visual editor alongside slides and video, and you pay for voice generation time. The free plan offers 10 minutes of generation in total, with no downloads and no commercial rights — a preview, not a starter tier. Creator ($19/mo billed annually, $29/mo monthly) includes 2 hours of generation a month, downloads and commercial rights; Business ($66/mo billed annually, $99/mo monthly) raises that to 8 hours a month and adds a business license aimed at companies. The official help center lists 33 languages including Japanese, with dedicated Japanese voices in the catalog. The licensing detail most likely to surprise people is broadcast: Murf's own help pages state that rights for traditional TV networks and commercial radio are Enterprise-only, separate from the standard commercial rights that cover monetized online video. Reselling Murf's voices, or using them to train an AI model, is prohibited.

STRENGTHS (per published specs)

  • Time-based pricing is predictable for video work: you budget in hours of narration, not characters
  • Studio editor, Canva and presentation integrations suit slide-and-video production
  • Japanese is supported on regular plans, with commercial rights from Creator up

TRADE-OFFS

  • Re-rendering an edited section consumes generation time again, so heavy editing eats the monthly hours
  • TV and radio broadcast use needs Enterprise — standard "commercial rights" do not include it
  • Voice cloning is an Enterprise matter, not a self-serve Creator or Business feature
See Murf plans

Speechify Studio — flat annual pricing, with a product-name trap

The first thing to get right about Speechify is that it sells three separate subscriptions: a reader app that reads documents aloud to you, Studio for making voiceovers, and an API. Only Studio grants commercial voiceover rights; a reader subscription does not. Studio meters in credits at one credit per second of voiceover, shared with its dubbing and avatar features, which burn credits faster. Studio Starter ($100/year) includes 86,400 credits a year — about 24 hours of voiceover, an average of two hours a month — with commercial rights and voice cloning. Studio Creator ($300/year) includes 345,600 credits, about 96 hours a year. The free Studio tier (600 credits) has no downloads and no commercial rights. Japanese appears on Studio's published voiceover language list. In our worked examples, Studio's annual plans produce the lowest cost per finished hour at steady volumes — with the caveat that you pay for the year up front and the low rate assumes you use most of the annual pool.

STRENGTHS (per published specs)

  • Lowest effective cost per hour at steady volume in our published-price examples (about $4.17 per hour of Starter allowance, $3.13 on Creator, if fully used)
  • Voiceover, dubbing, cloning and avatar video under one credit pool
  • Japanese is on the Studio language list, with commercial rights from the cheapest paid tier

TRADE-OFFS

  • Reader, Studio and API are separate bills — buying the well-known reader app by mistake gets you no voiceover rights
  • Credits are shared across features that burn them at different rates, so the voiceover hours are a ceiling, not a guarantee
  • Sold by the year: there is no cheap one-month paid test of Studio
See Speechify Studio plans

WellSaid — pay only for the minutes you download, English-first

WellSaid meters the opposite way from the other three: on paid plans, generation and retakes are unlimited, and only the minutes you download count against your allowance. Its voice catalog is built from recordings of licensed voice actors, which is the core of its pitch to corporate training and brand teams. Starter ($10/mo billed annually, $19/mo monthly) includes 240 downloaded minutes a year and full commercial rights; Pro ($33/mo billed annually, $49/mo monthly) includes 2,160 minutes a year; Business ($160 per user/month billed annually) includes 2,880 minutes per user a year plus team workspaces. The trial grants no commercial rights. The limitation that decides this tool for many readers is language: self-serve plans include English voices only. Japanese and other languages exist, but only inside the custom-priced Enterprise tier, alongside translation. There is also no self-serve voice cloning at any published price.

STRENGTHS (per published specs)

  • Retakes are free: you pay for finished, downloaded minutes, not for drafts — the friendliest metering here for revision-heavy production
  • Commercial rights included from the cheapest paid plan, on a catalog of licensed-actor voices
  • Simple budgeting for English training and corporate narration

TRADE-OFFS

  • Japanese and every other non-English language require Enterprise — for Japanese narration, this is effectively not a self-serve option
  • Highest cost per downloaded hour of the four if your workflow is one clean take (Starter's allowance works out to about $30 per hour)
  • No published self-serve voice cloning
See WellSaid plans

How to choose in four steps

STEP 1

Filter by language first

If your narration is Japanese, the self-serve shortlist is ElevenLabs, Murf and Speechify Studio. WellSaid only offers Japanese on Enterprise. Do this before comparing prices — the cheapest plan that cannot speak your language is not cheap.

STEP 2

Count your finished minutes, then your retakes

Estimate finished narration minutes per month, then honestly estimate how many times you re-render. Mostly one-take work favors character or time plans; revision-heavy work is exactly what WellSaid's download-only metering rewards.

STEP 3

Check the license against the destination

"Commercial" is not one thing. Monetized YouTube is covered by all four paid entry plans here; TV or radio broadcast is Enterprise-only at Murf; selling an audiobook also has to pass the retailer's own AI-narration rules, whatever your TTS vendor allows.

STEP 4

Run your real script through the free tier

All four have a free way in. Paste an actual script — especially a Japanese one, where the English "1,000 characters per minute" rule of thumb does not carry over — and note how much of the allowance one real project consumes before you pick a plan.

FAQ

What does text to speech actually cost per character?

It depends on the route. On ElevenLabs' subscription plans, the effective rate on the main models runs from about $0.20 per 1,000 characters on Starter ($6 for 30,000 credits) down to about $0.17 on the larger plans — that is the plan price divided by its credit allowance, at one credit per character. Its API lists $0.08 per 1,000 characters for the main models and $0.04 for the faster Flash/Turbo models. Murf's API lists $0.03 per 1,000 characters for its studio model. Murf Studio, Speechify Studio and WellSaid do not price by character at all in their studio plans — they meter time, seconds and downloaded minutes respectively, which is why this guide converts everything into cost per finished hour.

Can I sell an audiobook or client video narrated with these tools?

On paid plans, generally yes for content you have the rights to: ElevenLabs' paid plans cover audiobooks and client work, Murf's paid plans cover monetized video and streaming, Speechify Studio grants commercial rights for your own projects, and WellSaid's paid plans include full commercial rights. Two boundaries apply everywhere: you cannot resell the vendor's voices or service itself as your own product, and the destination sets its own rules — audiobook retailers and video platforms have their own AI-narration and disclosure policies that sit on top of the vendor's license. On free plans, commercial use is not allowed at any of the four.

Which of these can narrate in Japanese?

ElevenLabs (Japanese is in Multilingual v2's 29 languages, and newer models cover more), Murf (Japanese is on its official language list) and Speechify Studio (Japanese is on its Studio voiceover list). WellSaid is the exception: Japanese exists only on its Enterprise tier, while its self-serve plans are English-only. Note that character allowances behave differently with Japanese scripts — the roughly 1,000-characters-per-minute figure is an English rule of thumb, so measure consumption with your own script on a free tier before choosing a plan size.

What happened to Resemble AI, Play.ht and LOVO? Older guides recommend them.

All three have left this market. Resemble AI announced on August 13, 2026 that it no longer sells voice generation to new customers and now focuses on deepfake detection (existing customers are still supported). Play.ht shut down on December 31, 2025 after Meta acquired its team. LOVO's Genny service went offline after the company filed for Chapter 7 bankruptcy in May 2026. Any pricing for them in older articles is historical — do not plan a production workflow around them.

Is TV or radio broadcast use included in "commercial rights"?

Not automatically. Murf states explicitly that broadcasting rights for TV networks and commercial radio are Enterprise-only. For the other vendors, broadcast is not singled out in the public plan pages we checked, which is not the same as it being included — if your narration is headed for broadcast or a large paid campaign, confirm the scope in the vendor's terms or in writing before you produce it.

Why are there no voice-quality rankings or listening scores here?

Because this guide is compiled from the vendors' published specifications and pricing, not from hands-on testing. We do not claim to have listened to these tools side by side. Where a judgment appears — for example, that annual Studio pricing is cheap per hour if fully used — it is arithmetic on published prices, shown next to it so you can check it yourself. For voice fit, every vendor here offers a free tier or trial; a short test with your own script tells you more than any score we could invent.