To create a podcast based on your own documents, use the NotebookLM tool, which acts as a digital organizer for your sources and includes built-in features for generating multimedia content.
NOTE!Starting in March 2026, students and staff will have access to AI tools, including NotebookLM and Gemini, as part of the Google Workspace for Education suite. All you need to do is log in with your UW domain account. |
Preparing the source materials
The first step is to create a notebook and add the materials on which the podcast will be based.
You can upload:
- text documents (PDF, .doc, .txt files),
- links to websites or YouTube videos,
- pasted text or your own notes created within the application.
NotebookLM analyzes this data and generates responses based solely on these sources, which helps reduce the risk of hallucinations.
Configuration in the Studio section
After gathering your source materials, go to the section called Studio, which is located in the notebook interface. You will find the “Audio Overviews” option there.
Customizing the podcast
Before generating the recording, it is worth using the pencil icon to customize the final result.
You can specify:
- podcast format: You can choose between a detailed discussion (a conversation between two people connecting topics), a short summary, a critical analysis (including source evaluation), or a debate between two people presenting different points of view.
- language: You can select Polish, even if your documents are in another language.
- instructions for the bot: It is recommended to provide specific guidelines, such as asking the bot to focus on a particular aspect or to pronounce foreign terms correctly, as these might otherwise be read phonetically by default.
Generation and playback
After setting the parameters, request that the material be generated. NotebookLM will generate a conversation between two people who naturally discuss the content from your documents.
- Length: Podcasts usually last from a few to several minutes (often around 7–8 minutes).
- Limitations: The generated podcast cannot be edited directly within the program (e.g. you cannot modify the spoken content afterward) – if there are errors, you need to adjust the instructions in “draft” mode and generate it again.
The finished podcast can be listened to directly in the application, downloaded as an audio file, or shared with others.
If you would like to learn more about NotebookLM, watch dr Małgorzata Al-Khatib’s lecture “Create, Visualize, Teach – Advanced AI Tools”, delivered during the 25th anniversary celebration of e-learning at the University of Warsaw: