Long-text narration: where the real limits areClone a voice and use it to read your text

Long-text narration: where the real limits are

“Unlimited” is one of the most common words in this search, and the honest answer to it is not “of course, unlimited” but a number. Every service has limits; they differ, and they are almost never where people expect: not in the length of the text as such, but in the size of one request. Let us see exactly where they are, and what to do with a forty-thousand-character article.

600Characters per engine callAn internal step. You never see it: the text is split for you.
10,000Characters per request≈ 10–15 minutes of speech. This is the real ceiling.
No capOn requests in a rowThe voice is saved, so the parts sound the same.

Why the ceiling is where it is

It helps to understand the mechanics, because they explain why the number is what it is, and show that nobody set it out of spite.

  • The engine synthesises in short pieces, about 600 characters each. That is its own limit and not your concern: the text is cut up and joined back automatically, with nothing to split by hand.
  • The bottleneck is assembly, not synthesis. The pieces have to be joined and normalised into one file, and that step is what runs up against time. At ten thousand characters it finishes with room to spare; at twenty thousand it is a coin toss, and the consolation prize for losing is a twenty-minute unprocessed WAV.
  • Hence 10,000 as the working limit. It is not “how much is allowed” but “how much goes through reliably”.

What this means in practice

Ten to fifteen minutes of speech in one request is a long article, a full video script or a book chapter. The vast majority of jobs fit in that size whole, with nothing to split.

Work it out on your own text

How many minutes is it, and will it fit in one go

Paste your text. The limit and the price are counted in characters, not words, so the only honest answer to “is it unlimited” is a number.

142 characters13 s of speech

One request, no questions

About 13 s of speech. A text this size goes through in one go.

Open the chat and read it aloud

The pace used here is 11.1 characters per second, measured on our own Russian narration. English usually reads a little faster per character, so take the minutes as an upper estimate rather than a stopwatch.

If the text is longer after all

Then you split it, and one detail makes that painless: the voice is kept between requests. The parts will sound the same, not like three different narrators.

  1. Cut by meaning, not by counter. A part ends at the end of a section or paragraph, not at character 10,000 in the middle of a sentence.
  2. Do not change the voice between parts. That is the main and only thing to keep to.
  3. Join the files into one. Joining audio is a local operation with no separate price; just ask for it.
  4. Listen to the joins. The only place the split can be heard is the transition between parts; if the pause came out short, start the next part with a new paragraph.

You have a 25,000-character article. What happens if you paste it in whole?

On “free and unlimited”

Two words in one search, pulling in different directions. Unlimited free narration does not exist anywhere: a second of synthesis is a call to a paid model, and a service promising unlimited use hides the ceiling somewhere else: a watermark at the start of the file, a cut after the first minute, a queue or a daily limit.

Here there is one limit, and it is your balance. In exchange, narration is one of the cheapest operations: it is counted in characters, and the starter tokens you get on your first visit are enough for real work with texts. No watermark, no trimming, no cap on how many times a day you use it.

One limit that is not in the numbers

This whole chapter is about technical limits, so it is worth naming the one that is not technical. The engine counts the length of the text; whose voice reads it is up to you, and there is one limit here: your own voice, or someone else’s with their consent. A recording taken from someone else’s video is still their voice at any length of text, and fifteen minutes of narration in someone else’s voice is exactly as much worse as it is longer than a test phrase.

Paste a text and listen.

Loading…

What to read next

How the whole job works: clone a voice and use it to read your text. What makes narration sound alive rather than robotic: reading a text aloud with a voice. If the voice you need is not a narrator’s: a character voice for narration.

Frequently asked questions

Is there text-to-speech with no limits?
Unlimited free narration does not exist: every second is a call to a paid model, and services promising unlimited use hide the ceiling in a watermark, a cut or a daily limit. The real limit here is single and honest: 10,000 characters per request, roughly 10 to 15 minutes of speech. The number of requests is not capped, and there are no watermarks or cuts.
How many characters can be read at once?
10,000, roughly 10 to 15 minutes of speech: a long article, a full video script or a book chapter. Within that size the text is split into pieces of about 600 characters and joined back automatically, with nothing to cut by hand.
Why exactly 10,000 characters?
The bottleneck is not synthesis but assembly: the voiced pieces have to be joined and normalised into one file, and that step is what runs up against time. At ten thousand characters it finishes with room to spare; at twice that it is a coin toss, and the price of failure is a long unprocessed file. So it means “this much goes through reliably”, not “this much is allowed”.
How do I narrate a text longer than the limit?
Split it by meaning, at sections or paragraphs rather than by character count, and have the parts read one after another. The voice is kept between requests, so all the parts will sound the same. Then ask for the files to be joined into one: joining audio is a local operation with no separate price.
Will the join be audible?
Usually not: it is the same voice, and the timbre and pace do not change. The only place the split can show is the pause between parts. If it came out too short, start the next part with a new paragraph: the model breathes by punctuation and by the structure of the text.
Is there a daily limit on narrations?
No. There is one limit here, your token balance, and you can see it coming: you can ask the chat what any operation costs before you run it. There are no daily limits, queues or caps on the number of requests.

In the Twelver chat ten to fifteen minutes of speech is read in one request, and a longer text is split into parts that all sound in the same voice. Starter tokens arrive on your first visit, and signing up adds more.

Try it yourself

Everything in this guide runs inside Twelver

One chat for text, images, video, music and voice — no separate services or subscriptions.

Open Twelver chat
Related pageVoice Cloning
Rate your experience