Text to Speech Reader: Turn Articles and PDFs into Audio Fish Audio

A reader is a different job from a script generator. A script is written to be performed; a document is written to be read with the eyes, and it arrives full of things that only make sense on a page — page numbers, running headers, footnotes, broken lines. The work of a text to speech reader is to turn that page-shaped object into something that sounds like a person reading it.

This guide covers how to prepare a document, how to keep an hour of listening from sounding mechanical, and what long-form reading costs.

What makes a text reader different from a script generator

• The input is hostile. Articles, reports and PDFs carry artefacts a script never has. Cleaning the text is most of the quality difference.
• The unit is a section, not a line. A reader holds attention across thousands of characters, so pacing, sentence length and paragraph breaks matter more than performance tags.
• The listener is multitasking. Audio documents are heard while driving, walking or working. Clarity beats drama, and a steady voice beats a memorable one.
• The output is a file, not a preview. Long listening happens offline, so the download and the file lifetime are part of the workflow.

How to prepare a document for listening

1. Strip the page furniture. Remove page numbers, running headers, copyright footers, footnote markers and hyphenated line breaks. Anything left in the text is read aloud.
2. Split by section. Cut the document at chapter or heading boundaries into blocks of roughly 2,000 to 5,000 characters. A split keeps the read consistent and lets you regenerate one section later without touching the rest.
3. Fix the punctuation that carries the pacing. A full stop is a breath, a comma is a half breath. Where the original uses a dash or a slash, rewrite it as a sentence; the reader will otherwise swallow the pause.
4. Proof-listen the first two minutes. Check names, numbers and abbreviations before generating an hour. Fixing a pronunciation rule once is cheaper than regenerating a chapter.

Which documents work best

• Articles and newsletters — the most common case, and the easiest: the text is already clean and written in sentences.
• Reports and papers — dense but structured; split by section and read the abstract first to catch pronunciation of specialist terms.
• Books and manuals — the long-form case; chapter-level blocks, a single voice, and a download plan for offline listening.
• Course readings — pair the reader with the course narration workflow when the material is part of a curriculum.

What does a reader cost?

One credit covers about 1,000 characters, so the arithmetic follows the length of the document rather than the length of the recording session:

A book sits inside the $9 Starter pack, and a library of documents fits comfortably in the $29 Creator pack, which also carries commercial usage rights. The download guide covers what a generation produces and how files are kept; the audiobook guide covers the production side when the reading becomes a published title.

Frequently asked questions

Can I paste a whole PDF into the reader?

Usually not in one piece. Long documents read better when split by chapter or section: the engine holds pacing and tone more consistently across 2,000 to 5,000 character blocks, and a split also lets you regenerate one section without redoing the whole document.

Does the reader keep headers, footnotes and page numbers?

No. Clean the text before generating: remove page numbers, running headers, footnote markers and hyphenated line breaks. Those artefacts are read aloud literally, and they are the most common reason a PDF sounds wrong.

What does an hour of listening cost?

Roughly 9,000 to 10,000 characters per hour of speech, so about 9 to 10 credits. A 300-page book at 1,800 characters per page lands near 540 credits, well inside the 1,000 credit Starter pack.

Can I download the audio and listen offline?

Yes. A generation produces an audio file you can download and keep, and the download guide covers where files live and how long they are kept. Offline listening is the normal use for long documents.

Which voices work best for reading long text?

Steady, mid-paced voices with clear consonants. Avoid highly stylised character voices for documents: they are designed for short lines and become tiring across an hour of listening.

Related

• Text to speech download - what a generation produces and how files are kept.
• Audiobook narration - chapter-level production when a reading becomes a title.
• Fish Audio text to speech - the engine, the languages and the player.