"Doc casts" now available in 50 languages
Breaking language barriers for investigators, journalists, and anyone drowning in PDFs
Google's Audio Overviews can now generate "doc casts" in 50 languages—transforming any uploaded PDF into a narrated, multilingual summary in just minutes. For web researchers, that capability is pure gold.
Rely on this tool to breeze through stacks of documents that are interesting—but not essential—to read cover-to-cover. Two AI hosts will chatter away into whichever niche topics you throw at them in an upbeat, slightly over-the-top style—think "Let's take a deep dive into the world of bean-can manufacturing," or "Welcome to the fascinating universe of workplace etiquette with the Office Behavioral Policy." Often while waiting for trains or planes, I get a first impression of a document that allows me to decide: should I read it completely? The tool breathes life into jargon by casting dry studies as conversational banter, complete with cheesy analogies ("hyenas at an all-you-can-eat trash buffet") but somehow, this works for me. It:
Delivers a helicopter view in minutes, giving you the narrative flow and key takeaways without slogging through every technical detail.
Handles any document, from corporate HR manuals to wildlife-ecology papers, wrapping complex data into an engaging Q&A format.
Sticks to the essentials—highlighting major stats and trends—while letting you steer the discussion with extra prompts or files.
Scales instantly across disciplines and languages, so you get that “aha” moment faster and remember far more than you ever would by reading alone.
Available languages
The new languages are Afrikaans, Albanian, Arabic, Armenian, Azerbaijani, Basque, Bengali, Bulgarian, Burmese (Myanmar), Catalan, Cebuano, Chinese (Simplified), Chinese (Traditional), Croatian, Czech, Danish, Dutch, Estonian, Filipino, Finnish, French (Canada), French (European), Galician, Georgian, German, Greek, Gujarati, Haitian Creole, Hebrew, Hindi, Hungarian, Icelandic, Indonesian, Italian, Japanese, Javanese, Kannada, Konkani, Korean, Latin, Latvian, Lithuanian, Macedonian, Malay, Malayalam, Marathi, Maithili, Nepali, Norwegian (Bokmål), Norwegian (Nynorsk), Oriya, Pashto, Persian, Polish, Portuguese (Brazil), Portuguese (Portugal), Punjabi, Romanian, Russian, Serbian (Cyrillic), Sindhi, Sinhala, Slovak, Slovenian, Spanish (European), Spanish (Latin America), Spanish (Mexico), Swahili, Swedish, Tamil, Telugu, Thai, Turkish, Ukrainian, Urdu, and Vietnamese.
The current limit is 500,000 words per source or up to 200MB for local uploads. There is no page limit. Since this month, NotebookLM has enhanced its ability to understand PDFs that have both texts and images.
The new voices are not on par with the quality of the English voices - a pity if you want to use it for podcasts, but for audio notes it isn’t a big problem.
How to get your language in Google Notebooks? Click on Settings → Output Language.
Here are 10 ways web researchers can use the Audio Overviews, based on how used it myself the last few months:
1. I don't have time to scan 300 pages
Sometimes you just don't have time to read a 300-page report. Maybe you should read it, but how do you find time? It's important you prompt what kind of information you are after first, although, don't make it too specific or you can miss out on details.
When you're investigating money trails and suddenly face a 300-page financial disclosure document, you can instruct the two hosts to focus on tax evasion, changes in financial accounting, or greenwashing - it’s your choice.
2. Wait, these reports don't match
Imagine comparing environmental impact assessments from three different countries about the same multinational corporation. One report in Portuguese says everything's fine, another in Thai mentions "minor concerns," and a third in French basically describes an ecological apocalypse. It's yours to find out. Upload all three PDFs and get a conversational breakdown highlighting the contradictions that might have taken hours longer to discover manually. Again, don’t forget to tell what focus you want with the Customize button.
3. Finding patterns across global propaganda
For disinformation researchers tracking campaigns across Chinese, Russian, and Hungarian media simultaneously, this tool is also super helpful.
Upload up to hundreds of articles and get a podcast identifying common talking points, timing patterns, and narrative structures. It is possible to use it for identifying a coordinated campaign across several languages within hours instead of days.
4. Legal documents
Legal documents are deliberately written to make you feel stupid even in your native tongue. Now imagine them in Gujarati or Albanian! You can transform indecipherable legal jargon into conversational summaries. I quickly put it to the test by comparing asylum laws across 12 European countries in a fraction of the usual time.
5. Comparing political promises across borders
Want to compare how politicians in Japan, Brazil, and Germany all promise to fix healthcare? Upload their campaign materials and get a podcast that identifies common promises, unique approaches, and cultural differences in framing. Track how policy promises evolved across 15 countries during election seasons.
6. Academic papers
That 40-page academic paper about neural networks written in Korean doesn't have to be intimidating anymore. Upload it and let the AI hosts break it down conversationally. As always, if you are excited about the result, do study the original document to avoid any AI hallucinating - although I never had this problem yet.
7. Historical archives
Found 19th-century letters in Dutch or ancient trade records in Portuguese? Now you don't have to spend years learning dead dialects. An archivist recently used this to process colonial-era correspondence in three different languages to identify previously unknown connections between trading companies.
8. Social media analysis across languages
Tracking social movements across Twitter, Weibo, and European platforms? Upload thousands of posts and get a podcast identifying how hashtags and movements spread geographically and linguistically. Track how certain rhetoric spread from Persian to Arabic within 48 hours.
9. Leaked documents dump
When facing thousands of leaked documents in multiple languages, this tool becomes handy for initial triage. Upload everything and get conversational overviews that help identify which documents deserve deeper investigation.
10. Research preparation and literature reviews
Before diving into in-depth research, upload key papers across multiple languages to identify gaps, trends, and methodological differences. This can save days of preliminary work and help focus your research questions more effectively from the start.
The bottom line for web researchers
For professionals drowning in multilingual information overload it’s a practical, time-saving tool that changes the fundamentals of workflow. Ten years ago, we needed a team of translators on retainer. Five years ago, we relied on automated translation with human verification. Now, we can grasp the gist of complex documents in dozens of languages in minutes, freeing us to focus on analysis rather than basic comprehension.
Is it perfect? No. There are legitimate concerns about accuracy and nuance. But when you’re facing terabytes of global information under tight deadlines, having AI hosts discuss complex regulations in a digestible format isn’t just convenient—it can actually make you WANT to dive into the full PDF when you uncover a hidden gem.