Free Text to Speech Limits Compared: Real Caps, Checked

Almost no free text to speech tool is uncapped, and the cap is usually hidden in the unit. The tools that really have no limit run the voice on your own device, and each of them gives something else up: no MP3 download, no scanned-page reading, a large model to download first, or a licence that stops you publishing the audio. Every figure below comes from the vendor's own page, linked in the table or beside the claim. As a rule of thumb, about 1,000 characters is about one minute of speech.

Limits last verified: 22 October 2026. Next full re-check: January 2027. Where a vendor does not publish a number, the table says not publicly stated rather than guessing.

Why Every Free Tier Counts Something Different

Vendors use different units, so the numbers cannot be compared directly. That is why nobody can answer "which free tool gives the most" without a calculator. The units you will meet:

• Characters per day or month. NaturalReader counts per day, Luvvoice per month.
• Credits. ElevenLabs gives 10,000 a month, AnySpeech 5,000 once, and Listnr 1,000 "to start", without saying whether they recur.
• Hours per month. ElevenReader only.
• Minutes, once. Murf's 10 minutes never refill.
• Nothing stated. Speechify publishes no number for its free plan, though its own blog mentions a monthly premium-voice-minutes limit.

To put them on one scale, the table adds an estimate column using a rough rule: about 1,000 characters is about one minute of speech. That ratio is the one ElevenLabs' own pricing implies when it says 10,000 credits is about ten minutes. The estimates are ours, not any vendor's, and a vendor never states a word count, so treat them as a guide to scale, not a promise.

Free Text to Speech Limits by Tool

The "free allowance" column is what each vendor states, in its own unit. "Our estimate" is our conversion. Numbered notes under the table carry each tool's caveats. Every row was checked against the linked page on 22 October 2026.

The chart puts the cloud-voice allowances on one scale: minutes of audio you could generate in a month, on our rule of thumb. Only two reach hour scale.

Vendors change free tiers without notice. If a figure here no longer matches the linked page, the vendor's page is right. LOVO and Play.ht are left out of the table because we found no usable free tier for either; see below.

Which Free Tools Have No Cap, and What Each One Costs You

A free text to speech tool is only truly uncapped where the voice runs on your own device. That covers NaturalReader's free system voices, TTSReader's browser voices, soundtools.io, unlimitedtts.com and the built-in readers on your phone and computer. The catches matter more than the headline:

• No MP3. NaturalReader says its free voices cannot be used to convert to a downloadable MP3, with or without a subscription.
• A big first download. soundtools.io pulls a model of about 100 MB and does not run in Safari. unlimitedtts.com needs about 380 MB and may fail on an iPhone. Neither takes a PDF or a Word file.
• Voice quality set by your device. Device voices sound like your operating system, not like a cloud neural voice.
• A licence. TTSReader's free output may not be published or sold.

Cloud neural voices are metered at every vendor that states a cap, and vendors sometimes contradict themselves. Luvvoice's homepage says "no word limit" beside a 10,000-character monthly cap, and NoteGPT says "unlimited text length" beside 20,000 characters per session. When a page says both, the number is the one that bites.

Premium Promises That Are Not Free

Speechify's 1,000,000 words a month. This is the most repeated number in the category, and the page that states it never mentions free users. Speechify's usage-limits page guarantees it to Premium subscribers for 2026, against a 150,000-word monthly baseline, and says extra words after the year are discretionary. The page also contradicts itself on when the guarantee ends, giving both December 31 and "January 1, 2026". Speechify publishes no number for the free plan, although its own blog mentions an unquantified monthly premium-voice-minutes limit, and its pricing page calls the free voices "robotic sounding". If you rely on the million-word figure, check it again in January.

ElevenLabs' free credits. The 10,000 monthly credits come with no commercial licence, and published output must credit ElevenLabs. A half-credit rate you may have seen for the faster models applies to the API only. In the web app, every model costs one credit per character.

One-Time Trials, Monthly Allowances and Tools That Disappeared

One-time totals are not monthly allowances. Murf's pricing FAQ says there is no time limit on the free plan, but you can use it only until the 10 minutes are gone. Google's AI summary for a limits-comparison search still says "10 minutes of free audio monthly", which is wrong. AnySpeech's 5,000 credits and TTSReader's 5,000 premium characters are also totals, not refills. Listnr says 1,000 credits "to start" and does not say whether they recur. It is still operating, with a 2026 copyright line and blog posts dated 12 October.

Two names that keep appearing in older lists had no usable free tier when we checked. On 22 October, lovo.ai, its pricing page and genny.lovo.ai all returned an error page, so we could not see any plan. Meta acquired the team behind PlayAI, the company that ran Play.ht, in July 2025 (TechCrunch). Competitor blogs say the product shut down at the end of 2025, but we found no official notice, so we say reported. When we checked, the play.ht domain did not resolve.

Murf, Listnr and AnySpeech are voiceover tools for creators, and LOVO and Play.ht were too. They are the wrong category for reading a book to yourself, and their free tiers work as demos.

Free Limits for a Scanned Textbook

Here is what the limits mean for a 300-page textbook. Assuming about 2,000 characters a page (our assumption, since page density varies), that is about 600,000 characters.

• NaturalReader Lite voices: 20,000 characters a day, so roughly a month of daily sessions. Its free system voices have no stated cap.
• ElevenLabs: the free month covers under 2 percent of the book.
• ElevenReader: 10 hours a month, which the vendor itself compares to about a 400-page book. That is the largest metered cloud allowance in the table, though its handling of scanned pages is contested.
• ReadLoudly Free: 10 pages at a time, so a 300-page book is about 30 batches.

If the book is scanned, the filter is different. Community answers from Microsoft and Adobe say Edge's read aloud and Acrobat Reader's Read Out Loud need a real text layer, and Acrobat shows an empty-page warning on scans. Word's behaviour on scans is not publicly stated. NaturalReader's free plan does not include OCR, and Speechify lists OCR as a Premium feature. See Adobe for the Acrobat route. Free routes that do exist work per capture: Android's Select to Speak has camera recognition in 18 languages, and an iPhone's Live Text works one photo at a time. Google Drive's free text recognition requires a file of 2 MB or less, which a long scanned book is likely to exceed. For the full workflow, see scanning a book to audio. ReadLoudly's scanned PDF reader includes OCR on the free plan.

Built-In and School Options: Edge, Word, Apple, Android, Chrome, Bookshare

Before you install anything, try what you already have. If your document has a real text layer and a device voice at your own pace is enough, these cost nothing beyond a licence you may already hold (Word), and none documents a usage cap.

• Microsoft Edge (Edge 149 or later, Windows 10 or 11 and macOS) reads web pages and PDFs aloud. Only a few voices work offline.
• Microsoft Word is free only if you already have Office 2019, 2021 or Microsoft 365. Microsoft says Read Aloud does not store your content or audio.
• Apple Spoken Content works system-wide. The best voices are downloads of 100 MB or more.
• Android Select to Speak reads any screen, with camera recognition for text on paper.
• Chrome Reading mode is free but not available on every website, and its help page does not mention PDFs or Google Docs.

Two programmes are worth knowing about if you have a qualifying print disability. Bookshare is free to qualified US students from pre-kindergarten through college, with a default of 150 books in 30 days and no IEP or 504 plan required. Learning Ally costs $135 a year per household unless a school licence or fee waiver covers it, and offers human-narrated audiobooks. Both work from a catalogue, not from your own uploads. For the research behind these tools, see what the evidence shows about text to speech for dyslexia and ADHD.

ReadLoudly's Free Plan, Stated Plainly

We make ReadLoudly, so here is the Free plan stated plainly.

• Voices: the free voices your device provides, with no credits used.
• Pages: 10 pages at a time. Load the next 10 whenever you like. Paid plans process whole documents.
• Upload size: up to 50 MB, with or without an account.
• OCR: included on the free plan.
• Storage: guest uploads are kept about 24 hours. Free accounts keep a book for 30 days after you last open it.
• Ads: the free plan shows ads.

Premium voices and whole-document processing are on paid plans, and the full table is in the ReadLoudly plan table. To try the free plan, use ReadLoudly's text to speech tool.

How We Checked, and What Changed

Each figure was read from the vendor's own pricing, help or terms page on 22 October 2026. Where a vendor publishes no number, the table says Not stated. The minutes estimate is ours, from the ratio above.

Our next full re-check is January 2027, and the date at the top moves only after a real re-check.

Changelog. 22 October 2026: every limit checked against the vendor's own page.