Urdu Video to Text, When YouTube Has No Urdu Captions

You found an Urdu lecture, a talk show or a dars on YouTube and you want the words as text, to read, to search or to give to an AI. You click Show transcript and there is nothing, or the captions are in some other language. Your internet and your browser are fine.

The reason is simple and it surprises people here in Pakistan: YouTube does not make automatic captions in Urdu at all. Its own list has Hindi, Punjabi and Bengali but not Urdu. So you have to make the text from the audio yourself, and then make sure your PDF or app can show Urdu script. Both parts are explained below.

Does YouTube make automatic captions for Urdu videos?

No, and you can check it yourself. Google's automatic captions help page lists every language YouTube can caption by itself, and Urdu is not on it, although Hindi is. Spoken Urdu and Hindi sound almost the same, but they are written in completely different scripts, so a Hindi caption track is no use if you want Urdu text.

An Urdu video only has Urdu captions when the creator typed them or uploaded a captions file. Big news channels and some teachers do this. Most channels do not, so for most Urdu videos the only thing to work from is the sound.

What the video has	Can you get Urdu text from YouTube?	What to do
Urdu captions uploaded by the creator	Yes, from Show transcript	Copy it, or use a tool that reads the captions
Only auto captions in another language	No, wrong language and script	Transcribe the audio
No captions at all	No	Transcribe the audio

What is the free way to turn an Urdu video into text?

OpenAI's Whisper is free, runs on your own computer, and Urdu is one of the languages it was trained on. You use it from a terminal, which is the window where you type commands instead of clicking buttons. It needs Python and ffmpeg installed first, and that page explains the setup.

Save the video or its audio to your computer, and open a terminal in that folder.
Type whisper "lecture.mp4" --model medium --language Urdu --output_format txt with the real name of your file.
Wait, because Whisper listens to the whole thing, and an hour of Urdu on an ordinary laptop can take longer than the video itself.
Open the .txt file in Notepad or Word, which both show Urdu properly.

The --language Urdu part matters more than it looks. Whisper guesses the language from the first few seconds, and because spoken Urdu and Hindi are so close, it can guess Hindi and write the whole lecture in Devanagari. Telling it the language stops that. Use the medium or large model for Urdu, because the small ones make far more mistakes in languages other than English, and the Whisper paper shows accuracy varies a lot from one language to another.

Why does my Urdu text turn into boxes or come out backwards?

Two separate problems cause this. Urdu is written right to left, and its letters join and change shape depending on the letter next to them. A tool that was built only for English lays the letters out left to right and does not join them, so the words come out reversed and broken.

The boxes are a font problem. Text is just numbers until a font draws it, and if the PDF or app has no font with Urdu letters in it, it draws an empty box for each one. Windows has Urdu fonts, but some are an extra you add yourself, from Settings, then Optional features, and Microsoft explains that on its missing fonts page.

How does VideoDoc handle an Urdu video?

I live in Pakistan and a lot of what I learn from is in Urdu, often with the slides in English. An early version of VideoDoc printed Urdu in the PDF as rows of empty boxes, which is exactly the problem above, and fixing it was one of the first things I did.

Now VideoDoc uses the video's own captions first when they exist, which is much faster. When there are none, it transcribes the audio on your own computer. The PDF prints Urdu right to left with the letters joined, and if your PC has no font for the language, VideoDoc warns you before it starts instead of handing you a page of boxes. You also get screenshots of the slides or the board, each beside the words that were being said, plus a Markdown copy to paste into an AI.

Honest limits

English is still the strongest language in VideoDoc, and Urdu transcripts have more mistakes, especially with names and when the speaker mixes in English words, which nearly every Pakistani speaker does. VideoDoc writes Urdu in Urdu script, not Roman Urdu, and it does not translate. It does not label who is speaking. It works on public videos only, and how long it takes depends on your computer.

One Urdu lecture, words and slides

Try it on the Urdu video you keep rewatching.

Free to try on your own computer. Pro is $19 once, for two machines.

Try free in your browser Get Pro · lifetime

If the video does have captions, my guide on getting the transcript of any YouTube video covers the quick route. And for a full lecture you want to study from, there is one on turning a lecture recording into searchable notes.

Quick questions

Does YouTube have automatic captions in Urdu?

No. Urdu is not on YouTube's list of automatic caption languages, though Hindi is. An Urdu video only has Urdu captions if the creator uploaded them.

Can Whisper transcribe Urdu?

Yes, Urdu is one of the languages Whisper supports. Set the language to Urdu so it does not write Hindi, and use the medium or large model, because the small ones make many more mistakes outside English.

Why does Urdu show as boxes in my PDF?

The PDF has no font with Urdu letters, so it draws an empty box for each one. Use a tool that embeds an Urdu font, or add the Urdu fonts in Windows from Settings, Optional features.

Can I get the text in Roman Urdu?

Not directly from Whisper or VideoDoc, which write Urdu in Urdu script. You can ask an AI chat to convert the Urdu text into Roman Urdu afterwards, and check names by hand.

Pick the Urdu lecture you keep going back to, set the language to Urdu, and turn it into text you can actually search tonight.