Back to Blog
•7 min read

How to Turn an English PDF Into a Video in Another Language

Can AI turn an English PDF into a Hindi, Spanish or Arabic video lecture? Yes: how the output-language setting works, step by step, and what to check.

By ScholarlyGuides
Share:
How to Turn an English PDF Into a Video in Another Language editorial illustration

The short answer

Yes. You do not need to translate the PDF first. Upload the English PDF to Scholarly's PDF to video tool, open the creation panel, set the output language to the language you want (Hindi, Spanish, Arabic, Portuguese, German, Japanese and many more), and generate. Scholarly reads the English pages, writes the lecture script directly in the language you picked, narrates it, and renders chaptered scenes with on-screen text and captions in that language.

The same setting works for the other formats made from that PDF: a podcast, flashcards or a quiz can all come out in a different language from the source.

How the output language works

Every creation in Scholarly has an output language:

  • Auto (the default) follows the language of your source. An English PDF gives an English lecture; a German PDF gives a German one.
  • A named language overrides the source. Pick Hindi and the lecture is scripted and narrated in Hindi, whatever language the PDF is written in.

You can set a default output language once in Settings, or change it per creation in the language row of the creation panel. The picker also lists regional variants, such as Spanish (Mexico), Spanish (Spain), Brazilian Portuguese, Canadian French, Austrian German and Dutch (Belgium).

Two details matter for the video itself:

  1. The narration is told the language explicitly. Scholarly does not leave the speech engine to guess the language from the text. For languages the main engine cannot be told, such as Dutch, Polish, Czech or Greek, narration is routed to a multilingual speech model that is given the language by name.
  2. Everything on screen follows the language too. Titles, key points and diagram labels are written in the output language, captions and the transcript match, and built-in English scene labels are left out of non-English lectures instead of appearing untranslated. Arabic and Hebrew lectures lay out right to left, and the scenes carry fonts for Devanagari, Arabic, Thai, Chinese, Japanese and Korean.

Step by step

1. Upload the PDF

Free accounts can upload one file of up to 8 MB, and PDFs are processed up to 32 pages. Premium accepts files up to 300 MB and PDFs up to 1,000 pages, and up to 10 files can feed one video.

2. Pick the output language

In the creation panel, set the language. If your audience speaks a regional variant, pick it: for example, Spanish (Mexico) is narrated with a Mexican Spanish voice while Spanish (Spain) uses a Castilian one.

3. Add an instruction about terms

English technical terms often stay in English in real speech, especially in Hindi, Arabic and many workplace settings. Say which ones to keep:

"Keep 'cash flow', 'EBITDA' and product names in English. Audience: finance team in Mumbai."

4. Generate and review

Rendering usually takes 10 to 25 minutes. Watch the first minute with a fluent speaker and check names, numbers and terms. If something is off, adjust the instruction and regenerate.

Which languages work best

The output-language picker covers many more languages than most people need. We have written a page for each of the twelve languages people create in most often on Scholarly, explaining exactly how each one is handled, with a short sample in the language:

The full list, including French, Italian, Korean, Dutch, Turkish and Polish, is on the languages page.

Limits worth knowing

  • It is not a certified translation. The script is generated in the target language from your source. For legal, medical or safety-critical material, have a qualified reviewer check it.
  • One accent per variant. Portuguese is narrated with a Brazilian voice today, and Canadian French uses the same French voice as French. There is one Arabic entry, with no separate dialect option.
  • Names can be misread. Personal names, company names and uncommon terms are the most likely errors in any language. Spell out a reading in the instructions if one comes out wrong.
  • The interface stays English or Japanese. The material can be in Hindi or Arabic; the Scholarly app itself is currently in English or Japanese.

Free vs paid

A free account includes one lifetime AI Creation Credit, enough for one roughly 2-minute video at 720p in any output language. Premium ($30/mo or $144/yr) includes 15 AI Creation Credits a week, target lengths up to 45 minutes, 1080p and watermark-free MP4 downloads.

If you are doing this for a team, for example English SOPs that need Spanish and Hindi versions, see how to make training videos in other languages.