Why I made Filmina

I'm Igna, a final-year computer engineering student in Buenos Aires, Argentina. I made Filmina myself, and I made it for myself first.

It didn't start out as a product. It started out as a tool I made one night during term, because I had an exam coming and I couldn't get through the material.

I'm telling it because it explains almost everything the app does, and also what it doesn't do.

A hundred slides and me

I was studying from a deck of more than a hundred slides. That is the format almost everything we see in class comes in: not a book, not lecture notes, a deck. I opened it one night, got to slide twenty and realised I hadn't been reading anything for a while.

I got bored. I fell asleep. I picked up my phone without having decided to, came back, looked for where I had left off, reread two slides to pick up the thread, and there went another ten minutes. It wasn't that the subject was hard: it was that I never finished it.

The strange part is that the very same deck, in class, takes two hours and flies by. The difference isn't in the content. It's that in class there is someone talking.

I looked for it first

Before writing a line of code I looked for it. What I wanted was very specific: I have an exam, I have the material we go through in class (slides, most of the time), and I want to learn it. I found two families of tools, and neither of them did that.

Text-to-speech readers. I tried Speechify, the one that comes up first when you search. It does very well what it sets out to do, which is turn text into audio. But a slide read out loud is still a slide: three bullets of six words, written as a reminder by someone who already knew the answer, for someone who was there listening. Hearing them doesn't bring what is missing, because what is missing was never written down.

Language models. You paste the PDF into any of them and it interprets it for you, and it does it well. The problem showed up when I tried to study from that: the summary they give back is never complete. They leave things out. They pick three ideas, tell them beautifully and drop half the deck, and from where you are sitting there is no way to know what got dropped. They are good for asking, which is the other thing they do well, and there they are still the best thing around. But to ask you have to know what you didn't understand, and I hadn't looked at slide 12 long enough even for that. None of them explains the material to you as you go through it.

Getting the class back

Put that way it was clear what had to be built. Filmina is the class again: the slide up front, a voice that explains it whole and in order, and a frame that tells you which part it is talking about.

With one difference worth saying out loud: the examples and the experiences a teacher brings from their own life are not there. That is the best part of being in the room and nothing replaces it. But when the exam is a week away, what I need is to learn the material and understand the concepts, and that you can have again at one in the morning, with the file they uploaded to the course page.

My university records its lectures, so for a while that is what I did: watch them again. It works, and it is the only thing that gives those examples back. What it costs is time. Each class is a four-hour mp4, and inside those four hours there are silences, dead minutes, the question from someone who came in late and remarks that are not needed to understand the concept. Watching all four to rescue forty minutes, a week before the exam, was time I did not have.

Filmina does that same job without the filler: the material explained out loud again, in a fraction of the time. Of everything it does, that is the part that changed my days the most.

The other twist: no credits

There is something else I wanted to change, and it isn't about how you study but about how you pay. Almost every study app works the same way: it sells you AI credits. You upload a PDF and it gets deducted, you ask something and it gets deducted, and next month, if it was any good, you pay again.

As a student I know how hard that is. At the start you have no money, and if you work you don't want it all going on AI subscriptions, which are never just one either: the chat on one side, the summariser on another, the reader on another. Exactly when you need the tool most is when you can least pay for it.

That is why Filmina is a Mac app and the model runs on your own machine. There are no credits and no tokens to spend: nothing runs out halfway through the deck, there is no counter watching you while you study, and using it all night before the exam costs the same as using it for ten minutes. The program is paid for once and that's it.

What is left of that night

The first version was for me alone and it showed. But I finished that deck, and the next one, and the ones for the subject after that. Everything the app has today came out of there, out of fixing one by one the things that annoyed me while actually using it to study:

• That it explains everything, even what looks obvious, because what you don't understand is never what the summary chose to tell you.
• Three levels of detail, because studying something for the first time and going over it two hours before are not the same reading.
• A frame that surrounds and doesn't cover, because any layer on top of the paper gets in the way exactly where you have to read.

I don't know whether it will work for you. I know I made it because I needed it, that I used it for my own exams before it was something you could buy, and that I still use it every time I have to study. It is the most honest thing I can offer you.

Today is September 21, which in Argentina is Student's Day, and this was written by a student, for other students. Happy Student's Day.