Books

How to Narrate a Scanned Book: Prepare the Text First

To narrate a scanned book, first turn its page images into checked text outside TGRadio. The app does not include OCR, so a PDF without a text layer is unsuitable for import. Test several pages first: a single file may contain both recognized pages and ordinary page images.

Test copying on several pages

Open the PDF in a viewer, select a sentence, and paste it into a text editor. Repeat this near the beginning, middle, and end, plus a page with unusual formatting. Zoom level and sharp-looking letters do not establish whether text can be extracted. What matters is the pasted result, including its word order.

If copying works on only some pages, inspect the others separately. One recognized page does not make the entire scan ready for narration.

Recognize the images with an external tool

Use a separate text recognition program that lets you specify the document language and save the result. This preparation happens outside TGRadio. For skewed or cropped pages, first find a better copy or correct the image using available tools. Recognition cannot reliably recover letters missing from the scan.

For example, a dark strip near the binding may hide the beginning of every line. Test recognition on that difficult page before processing the whole book.

Check mistakes that change the meaning

Compare recognized text with the page image line by line in difficult passages. Pay particular attention to negatives, names, numbers, and punctuation between words. For example, a missing “not” reverses a statement, while confused digits and letters can corrupt a date. Make corrections in an external editor; TGRadio has no text or pronunciation editor.

Go beyond the opening paragraphs. Inspect small print, damaged pages, and passages in another language, because recognition quality can vary within the same book.

Remove printed-page artifacts

Inspect page and paragraph boundaries. Repeated running headers, page numbers, and words split across lines may interrupt sentences. For example, join a word broken by a line-ending hyphen when that hyphen exists only because of the page layout. Preserve genuine hyphenated spellings by comparing them with the original.

Check tables and formulas separately. TGRadio does not recognize their images, and an unordered string of recognized symbols will not automatically become a clear spoken explanation.

Import the checked copy and start playback

Save the corrected text, for example as TXT, or prepare a PDF with a verified text layer. The import file must be no larger than 10 MB, and the extracted book text is limited to 2 million characters. Reducing a PDF’s file size does not create a text layer or correct recognition errors.

In Books, choose Upload a book and add the prepared file. Select the book and voice, then press Play. Compare available audio passages with the scan wherever wording remains doubtful; subsequent fragments are prepared as you listen.