NotebookLM's AI podcasts speak 50+ languages

NotebookLM’s podcast hosts just went multilingual.

When Google first launched Audio Overviews late last year, the internet couldn’t stop listening to two surprisingly natural AI hosts turn boring PDFs into entertaining chatter. The catch? It was primarily an English-only party, even though NotebookLM had already rolled out to over 200 countries.

Now, Google has announced that Audio Overviews support more than 50 languages, powered directly by Gemini’s native audio capabilities.

Ditching the language barrier

The update brings a brand new Output Language setting directly inside NotebookLM. Whatever language you pick—whether it is Hindi, Turkish, Afrikaans, or something else—both your generated Audio Overviews and your chat responses will follow suit.

You can switch this setting on the fly, which completely changes how multilingual research works.

Google pointed to a handy classroom example: a teacher can pull together a Portuguese documentary, a Spanish research paper, and an English field report on the Amazon rainforest. Students can take that messy pile of foreign sources, hit generate, and listen to a conversational breakdown in their own native tongue.

Google noted that this rollout is an early look at what native audio can do, with plans to keep refining the experience based on user feedback.

Why it matters

For everyday knowledge workers, students, and researchers, this turns the dreaded chore of parsing foreign-language documents into something you can easily digest on your commute. You no longer have to run texts through a clunky translator before trying to wrap your head around them.

For the broader AI landscape, this is a clear demonstration of Gemini's native audio strength. While most competitors still stitch together a large language model with a robotic text-to-speech engine, Google is generating conversational banter natively across dozens of languages. It sets a much higher bar for what an AI study buddy should sound like.

In our view, rolling this out widely within a free product is a brilliant land grab for global user engagement. The big risk, however, is cross-lingual hallucination. When an AI summarizes a complex Spanish study into Hindi audio banter, subtle nuances can easily get lost in translation. It is fantastic for quick orientation, but you will still want to verify the primary sources before citing them in serious work.