PDF to audiobook in Hindi: turn a book or your notes into MP3

A book you have no time to read can be listened to on the bus, on a walk or while cooking. Turning a PDF into audio is simple when the file is clean, and frustrating when it is not, because most of the trouble comes from the file and not from the voice. This guide spends its time there: what to check before you upload, how to fix the usual problems with Hindi PDFs, and what a full book costs.

Which files work

The audiobook maker reads four kinds of file: EPUB, PDF, DOCX and TXT, up to 20 MB and up to 1,000,000 characters per book. EPUB and DOCX usually give the cleanest result, because they hold the text in reading order. PDF is the one that causes trouble, for two reasons that are worth checking first.

Check 1: is the PDF made of text, or of pictures?

Open the PDF and try to select a sentence with your mouse or finger. If you can highlight the words, the PDF contains text and will work. If the whole page selects as one block, it is a scan: a set of photographs of pages. There is no text in it to read, so it has to go through OCR (text recognition) first.

For a short scanned document, Google Drive can do this free: upload the file, right-click it, choose Open with, then Google Docs. Drive detects the language by itself. Google's help page puts the limit at 2 MB per file, so this suits a chapter or a set of notes. A whole scanned book needs a dedicated OCR tool.

Check 2: does the Hindi text copy correctly?

This one is specific to Hindi and catches many people out. Copy a line from the PDF and paste it into any text box. If it pastes as Hindi, you are fine. If it pastes as a jumble of English letters and symbols, the PDF was typeset in an old non-Unicode font such as Kruti Dev. It looks like Hindi on the page, but the file actually contains Latin characters, and any text to speech tool will read out nonsense.

The fix is the same as for a scan: run the pages through OCR so you get real Unicode Hindi, or use a font converter if you have the original document.

Clean the file before you upload

The audiobook maker reads a PDF's text as it finds it. That means everything on the page gets read, including the parts your eye skips.

• Page numbers, headers and footers. A book title repeated at the top of every page will be spoken on every page. If you can, convert the PDF to DOCX, delete them and upload the DOCX.
• Footnotes. They are read where they sit on the page, in the middle of the story.
• Tables and two-column pages. These can come out in the wrong order. Check a page of the result before converting the whole book.
• The front and back matter. Copyright pages, the index and the bibliography make dull listening. Delete them.

Steps on the website

1. Log in with Telegram. It is free and takes a few seconds.
2. Choose your engine and voice in Studio. For Hindi, the free Microsoft voices are Madhur and Swara.
3. Open Audiobook and upload the file.
4. Check the summary: number of characters, estimated audio length and the cost. Nothing is charged until you press start.
5. Start the conversion. A progress bar shows how far it has got; a long book takes a while because it is generated in many small parts.
6. Download the MP3. A long book is split into several files so each one stays a manageable size.

There is an option to send the finished audiobook to your Telegram as well, which is the easiest way to get it onto your phone.

Steps in Telegram

The Telegram bot does the same job without opening the website: send the file to the bot, check the cost it shows, confirm, and it sends the MP3 parts back in the chat. Your account, voice and credits are the same in both places, so you can set the voice on the website and convert from your phone.

What does a whole book cost?

It depends on the length of the book and the voice. A page of a typical paperback holds very roughly 1,500 to 2,000 characters, and an hour of audio is about 54,000 characters.

The free Microsoft voices are limited to 200,000 characters a day, so a book longer than that cannot be converted in one go on a free voice. You have two options:

• Split it. Save the book as a few files of under 200,000 characters each, by chapters, and convert one a day.
• Use credits. With the ElevenLabs and Cappy voices a character costs one credit, so a 1,000,000-character book needs 1,000,000 credits. Gemini AI voices cost two credits per character. The pricing page shows the current plans.

If a conversion fails part-way, the credits it used are returned to your balance automatically.

What you are allowed to convert

Convert what you have the right to use: your own writing, class notes, documents from work, and books that are out of copyright. Premchand's novels and stories, for example, are in the public domain in India. A recent bestseller is not, and converting a copy you downloaded from somewhere does not make it yours to share. DRM-protected EPUBs from ebook stores cannot be read by the tool at all.

Good uses beyond novels

• Revision notes. Students can turn typed notes into audio and listen on the way to an exam.
• Long reports. A fifty-page PDF is easier to get through on a commute than at a desk.
• Your own drafts. Hearing your writing read aloud is one of the quickest ways to catch clumsy sentences.
• For someone who finds reading hard. Older relatives or anyone with poor eyesight can listen to a book in their own language.

Narrating videos and not books? The YouTube voiceover guide covers writing a script that an AI voice reads well.