Your files and privacy
What happens to a document
When you drop a file, it goes to Podcaster's server over an encrypted connection, where the text is read out of it and the episode is made. The document, and the text read out of it, are deleted from the server when the episode is done, and they are deleted just the same when the job fails or you cancel it, or when a preview is left a day without being made into an episode. What stays is the episode and its study material, in your library, until you remove them.
The picture drawn for an episode's cover, without its title or anything about you or the reading, may appear in the art on Podcaster's front page. It is Podcaster's own drawing from a scene it imagined; nothing of your document or your account is in it.
The text of the reading is sent to AI models to write the script, the notes and the questions, and the script is sent to a voice model to record it. Those models are run by their providers (Google, Anthropic, Mistral AI and Deepgram, reached through OpenRouter, and Google's voices at Google directly as well) under their API terms. Podcaster sends them the text of the reading and the name the hosts use for you, so that the voices can say it, and nothing else about you.
What Podcaster keeps
Your account, your library (the episodes and their files), and a ledger of what each episode was charged. Nothing else about you: no listening history beyond how far you got in each episode, no contacts, no browsing.
Fourteen days after you delete your account, any reading still being made is stopped, your library is removed, and the titles of your readings and everything that says which texts they were are blanked. The privacy policy lists all of it.
Podcaster also keeps what it made from a reading, so it makes it once: the script, the recorded voices and the study material, without your name, your choices or anything else about you. When somebody else brings the same reading and asks for the same kind of episode, theirs is made from that, with their own name in it. The reading itself is never kept.
Copyright
The reading is yours to study, and making a podcast of it for your own use is study. Two things to keep in mind: a subscribed article comes with your library's terms, which may not allow uploading it to a third-party service, so check them; and the episodes are for you, not for publishing. Podcaster does not publish your episodes; what it made is only used again for somebody who brings the same reading themselves.
Made by AI, and it says so
Every episode, page of notes and question is made by AI models, and each one says so. The last line of every episode, after the credit, is "All content has been AI generated. AI can make mistakes.", in the episode's language; the same two sentences close the study notes and the practice questions; and the MP3 carries a marking that software can read. Models make mistakes, and an episode is a companion to the reading, not the reading itself.
The full pages
The privacy policy lists exactly what is kept and where, and the terms of use say what you agree to when you make an account.