It reads the document, not the page
A printed page carries more than the text: a book’s title at the top of every page, a page number at the bottom, a table laid out in columns. Read out word for word, those would interrupt every few minutes. DocSpeak takes the body of the document and leaves out the running heads, headers, footers and page numbers. A table is read a row at a time, its cells one after another.
A PDF is first turned into a Word document by the same engine as PDF to Word, so a clean, digital PDF reads best.
One MP3 for each chapter
Every Heading 1 in a Word file starts a new chapter, and each chapter becomes its own MP3, named after its heading in the document’s own script and numbered in order, all in one zip. A short document, or one with no headings, is a single MP3. A chapter of only a few lines is joined to the one after it, so a title page does not become a file of its own.
The files are mono MP3s at 32 kbps, about a quarter of a megabyte for each minute of audio, small enough to keep a whole book on a phone.
How long the audio is, and how long it takes
The voice reads about 12 characters a second, so a page of 2,000 characters is about three minutes of audio. Before anything is spoken, DocSpeak counts the pages and the minutes of audio in your file and shows the price.
Making the audio is much faster than listening to it: in the first live job, a two-chapter Telugu Word file gave 5.6 minutes of audio in 23.5 seconds. A long book takes a few minutes.
Hear your own file before you pay
The free 30-second preview reads the start of your own document, so you hear how the voice handles its names, numbers and English words before you spend anything. The voice was chosen by listening to Telugu; no reader has yet checked it in each of the other languages, so the preview is the honest test.
What it does not read yet
A scanned PDF, a page that is a picture of text, is not read yet: convert it with PDF to Word first, which reads it with OCR, and give DocSpeak the Word file. A document typed in an old, non-Unicode font, such as Kruti Dev or Anu, has to be converted to Unicode first with the Legacy Font tool. Text inside pictures is not read.