PDF to audiobook
PDF is the hardest input format to do well, because a PDF describes a page layout rather than a book. Here is what NarrationOS does about that, and where the limits are.
Before converting, it helps to know which kind of PDF you have. Try selecting text in your PDF reader.
A PDF has no idea what a chapter is. It knows about glyphs at coordinates. Naive extraction produces audio full of page numbers, running headers repeated every page, and sentences broken in half at column boundaries.
The PDF parser in NarrationOS reconstructs reading order, detects headings and subtitles, strips repeated page furniture, and avoids splitting chapters at false boundaries. It is genuinely better than a plain text dump, and it is still the format most likely to need a look over before you render.
Once the text is right, the PDF to audiobook conversion itself is the same as for any book: pick a voice, render, and export an MP3 with chapter markers or a WAV file.
With PDFs specifically, it is worth reading through the imported text first. A long render of mis-extracted text is wasted time, and problems are far easier to fix at the text stage.
NarrationOS includes a text editing mode for exactly this: fix a broken paragraph, delete a stray header, or repair a hyphenated word split across a line break. Then narrate.
OCR quality depends entirely on scan quality. A clean 300 DPI scan of printed text usually reads well. A skewed phone photo of a dim page, a heavily stylised typeface, or handwriting will produce errors, and those errors will be narrated faithfully as written.
If your source is a poor scan, expect to do some cleanup. This is a limitation of recognising text from images, not something any narration tool can fully solve.
Yes. Image-only PDFs can be run through OCR so the pages are converted to text first, then narrated. Results track the quality of the scan: clean, straight, reasonably high-resolution scans work well, while dim or skewed photographs produce recognition errors.
That is page furniture leaking into the extracted text. NarrationOS strips repeated headers and footers during parsing, but unusual layouts can still slip through. Open the imported text, remove the stray lines, and render again.
Yes, clearly, when you have the choice. An EPUB carries real chapter structure and needs no layout reconstruction or OCR. Use PDF when it is the only format you have. See the EPUB guide for that path.
Yes. Once the text is extracted, with OCR first if it was a scan, NarrationOS narrates it and exports an MP3 with chapter markers and tags, or a WAV file. The chapter markers follow the headings detected in the PDF, so check them in the imported text before you render.
Narration