What Filmina is and how it works
Filmina is a desktop teacher for your class PDFs. You drag the deck in and it explains it out loud, slide by slide, marking on the page which part it is talking about.
It is not a text-to-speech reader. A reader says what is written; Filmina reformulates what the slide says, connects it and explains why it matters, which is what the lecturer did in class and what is missing when you study alone.
Everything runs on your Mac: the model that writes the explanation and the one that provides the voice are installed on your machine. The PDF is not uploaded anywhere, there is no account to create and you pay once.
How it works, in one sentence
Filmina reads each slide's text blocks, writes an explanation split into ideas, says it with a local voice and frames the block it is talking about; when the slide ends, it asks you a question.
Step by step, that is:
- The deck goes in. A PDF, or a .pptx, .ppt, .key or .odp that gets converted to PDF first. For each page, the render and the text blocks with their coordinates are stored.
- The explanation gets written. The model receives the slide's blocks and returns between three and seven segments, each with one or two ideas and the blocks it covers.
- It gets checked for gaps. On the levels with total coverage, the code verifies that no block was left unexplained and, if one is missing, asks again for that one alone.
- It gets said out loud. Each segment is synthesized with the voice you chose and plays while the frame surrounds the block it belongs to.
- And it asks. At the end of the slide a question appears about what was just explained, in one of four ways that take turns.
None of that happens on a server. While Filmina is explaining, transcribing or speaking, there is no request leaving your Mac: the models are on disk and work against your own processor.
What you need
- A Mac with Apple Silicon -M1 or later- and macOS 13 or newer.
- About 10 GB of disk for the model that explains. It downloads once, the first time, and it is the only thing internet is needed for.
- A class PDF. With a text layer: if the page is a scanned photo, Filmina tells you instead of making things up.
It explains in Spanish or English, whichever you pick, whatever language the PDF is written in.
by @ignaprados