Filmina vs ChatGPT
Uploading the PDF to the chat and asking it to explain works very well. Until there are sixty slides at eleven at night. How they differ, and why they compete less than it looks.
Almost everyone studying with AI today does the same thing: they drag the PDF into ChatGPT and type «explain this to me». And that's fine, because it works. No tool is going to clear up a specific doubt better than a chat that understands what you are asking.
The problem shows up when what has to be studied is not a specific doubt, but a whole file.
ChatGPT answers what you ask it. Filmina walks you through the whole document without you asking anything, in the order of the lecture and marking on the slide what it is talking about.
What ChatGPT does better, and isn't up for debate
- Asking. «I didn't get why the matrix is inverted in step 3» is something Filmina cannot answer and ChatGPT can.
- Pushing. You ask for another example, for it to explain as if you were fifteen, for a comparison with something you already know.
- Diagrams. It sees images. Filmina works on the text only: if there is a figure, it tells you and tells you to look at it.
- Testing you. It builds practice questions and marks your answers.
- It's everywhere and it has a free version.
Filmina does none of this. If your problem is «there is a concept I don't understand», the chat is the right tool and this page ends here.
What happens with sixty slides
It summarises, because that is what a chat does
You ask «explain this PDF to me» and you get fifteen paragraphs about a sixty-slide document. It is a summary of a summary: the slides were already the summary of the lecture. What was left out it doesn't tell you, and the one who has to notice what is missing is you, who by definition doesn't know the subject yet.
Filmina goes slide by slide, in the order the lecture was given, and covers every block of text on each page. Coverage is verified in code: if a block was left unexplained, the model is asked again for that block only. It doesn't depend on it remembering.
Asking one at a time is work
The way to get a chat to explain slide 22 properly is to paste it slide 22 and ask it to explain it. Sixty times. Copy, paste, read, copy, paste. By slide twelve you are tired and you switch to asking for summaries ten at a time, which is exactly going back to the previous point.
In Filmina you drag the file in once. While you listen to the first slide, the rest is generated behind you, so in the long run you don't wait on any of them.
You have to read it, and reading is what you were avoiding
The chat's answer is text on the screen. If at eleven at night you didn't feel like reading forty slides, you won't feel like reading forty answers either. ChatGPT has a voice mode, but it is a conversation, not a walk through the document: there is no place where the PDF stays open with someone pointing at the part they are talking about.
That is literally the only thing Filmina does: the page on the left, the voice explaining, and a frame that surrounds the block being talked about and slides to the next one.
It makes things up, and on class notes you notice late
A chat fills gaps with what it knows about the world. Sometimes that's a gift, because it brings context the slide didn't have. Sometimes it is a definition your course gives differently, and you find out in the exam.
Filmina explains what is on the slide, block by block, and when you ask for the extended level it tells you that there it is indeed adding outside context. The summary and full levels don't leave the document.
The file gets uploaded
For the chat to read your PDF, the PDF travels to a server. Filmina has no server: the model that writes and the voice that speaks are installed on your Mac. It works offline, and the file never leaves your folder.
Filmina vs ChatGPT, point by point
| What matters | Filmina | ChatGPT |
|---|---|---|
| What you get | The whole document explained out loud, slide by slide | An answer to what you asked |
| Format | Audio with the page in front of you and a frame that points | Text on the screen, or a voice conversation |
| PDF coverage | Every block, verified in code | Whatever fits in the answer |
| Work on your side | Dragging the file in once | Asking part by part, and deciding what to ask |
| Questions and doubts | There is no chat | It's what it does best |
| Diagrams and images | Tells you there is a figure and tells you to look at it | It interprets them |
| Where it runs | On your Mac, offline | In the cloud, with an account |
| Pricing | One payment of US$ 39 at launch, US$ 19 for students | Free with limits, or around US$ 20 a month |
The short way to decide
ChatGPT, when you have a doubt
You know what it is you don't understand and you want to ask it, argue about it, ask for another example or be tested on it.
Filmina, when you have a file
There are sixty slides, you don't even know where to start, and what you need is for someone to walk you through all of them while you look at them.
They go well together, and not out of diplomacy: you go through the PDF with Filmina, note the two or three things that stayed shaky and ask the chat about those. What doesn't work is the other way round, asking a chat to be a whole lecture.
And the same goes for Claude or Gemini
Everything on this page applies just the same to the other chats. The difference is not which model answers better: it is that a chat is a place where you ask and Filmina is a place where the document walks itself, with the page in front of you. They are two different things.
Frequently asked questions
Why not ask ChatGPT to read the PDF out loud to me?
You can, and it sounds good, but it is a conversation: there is no PDF left open with someone marking which block is being talked about, and what it reads you is what it summarised, not the whole document.
Does Filmina use ChatGPT behind the scenes?
No. It uses an open model running on your own Mac, so there is no call to any server and no file that gets uploaded.
So does it explain worse than ChatGPT?
For an open, hard question, a large model in the cloud reasons better. For the specific task of explaining the blocks of a slide without making things up and without skipping anything, a local model with the text in front of it and verified coverage is more than enough.
Can I ask Filmina questions?
No, there is no chat. What you can do is raise the level of detail to extended and listen to that same slide again, with examples, the why and the usual mistakes.
What does Filmina cost against a chat subscription?
A one-time payment of US$ 39 at launch, or US$ 19 with a student email, against around US$ 20 a month for a chat subscription. You can try it free with the first ten documents, up to slide 100 of each one. See the pricing.
Keep comparing
- Filmina vs SpeechifyThe best known text-to-speech reader, against an explanation.
- Filmina vs NotebookLMGoogle's two-host podcast, against the explained slide.
- Why I made FilminaWhere it came from: an exam, a hundred slides and no tool that did the job.
by @ignaprados