Google just rolled out a new feature in NotebookLM: Video Overviews. It’s meant to turn your notes into short narrated videos with slides, basically an AI-generated presentation based on your sources. Officially, this only works in English for now. But that’s not entirely true.
I gave it a try using some custom instructions, asking the system to use only Hungarian throughout the video. To my surprise, it worked—almost perfectly. The narration was fluent, and most of the phrasing sounded natural, with just a few minor glitches. You can check it on Reddit.
The only part that didn’t follow the language switch was the visuals: the on-screen text and slide labels still showed up in English, no matter what.
The trick was to use a prompt like this:
This is a special Video Overview created entirely in [Language].
#### Special Instructions
- This video will be presented **exclusively** in [Language]. All narration, slide content, and on-screen text must be in [Language] for the full duration of the video.
- Do **not** include English or any other languages, except when necessary to clarify a culturally or technically specific term.
- Ensure that all terminology, examples, and explanations are adapted to suit the expectations of a [Language]-speaking audience.Of course, you can swap in any language. I imagine that languages like Spanish, French, or German—ones with better LLM training coverage—would work even more smoothly.
This isn’t something Google highlights in their announcement, but the potential is there. Even though multilingual support is still “coming soon” officially, the underlying models clearly can handle more than just English if nudged the right way.
So if you’re already using NotebookLM and want to experiment with non-English summaries, try slipping a few instructions into your prompts. You won’t get full localization just yet, but for audio narration, it’s surprisingly solid.
If anyone figures out how to get the slides to follow the same language switch, let me know. That part still seems locked in English for now.