Skip to main content

Read

Lecture slides, with a teaching assistant for the parts you missed

Slides are what is left of a talk once the talking is over. Open the deck and each page gets explained; box the part you did not follow and ask about it; keep asking until it makes sense — with every answer pointing back to the slide it came from.

Opens
.pptx, laid out as slides
Includes
Speaker notes, in the text
Answers cite
The slide number
Runs on
macOS (Apple silicon) · Windows
Price
Free app · AI plans from $10/mo, or your own key
AI
Included, or your own API key

A deck is a strange document. Half of what it means is in the arrangement — this box under that heading, these three things in a row — and the other half was never written down at all, because the person who made it was going to say it. Tools that turn a .pptx into a stream of text throw away both halves and hand you the least useful part: the bullet points, which were already a compression of the real content. That is exactly the position you are in after a lecture you did not quite follow: the slides, and a gap where the explanation was.

Rare

A card per page, for the things nobody wrote down

The useful thing an AI can add to a deck is not another summary of the bullets. It is the half of the presentation that lived in the presenter's head.

  • Each page gets a card: what the speaker would have said out loud, how this page follows from the one before it, what is worth asking about it, and whether you can move on quickly.
  • Cards arrive a few pages at a time while you read, rather than after a wait for the whole file, and they are cached so a second visit costs nothing.
  • You start it deliberately. Opening a deck does not quietly spend credits on a hundred pages you were only skimming.
  • Speaker notes, where they exist, are part of the text — so the card builds on what the author actually said rather than guessing at it.

Box the part you didn't get, and ask about just that

On a slide, the thing you are stuck on is rarely a sentence. It is the diagram, the equation, the chart, the arrow from one box to another — and describing it in words to a chat box is harder than the question itself.

  • Drag a box around it — one formula, one quadrant, one step of the architecture — and ask. "Why does this arrow go back?" is a complete question when the arrow is what you boxed.
  • It is not a screenshot. The words and numbers underneath the box travel with it from the file itself, so a value in the answer is read from the slide rather than guessed from a picture.
  • The question knows which slide it is on, and the rest of the deck comes with it — so "how does this connect to what came before?" has something to answer from.
  • For a passage of text, select it and press Explain for a plain reading of just that.
Why a region is not a screenshot →

Keep asking, with the whole course in reach

One explanation is rarely where understanding happens. It happens on the third follow-up — and a follow-up is only useful if whoever answers it has read what you have read.

  • The slide you are on goes with every question, so "what does this mean?" never needs a page number or a description of the picture.
  • The whole deck is in reach, speaker notes included — so it can connect this page to the definition twenty slides back, or to where the idea is finally used.
  • So is the rest of the folder. Put the lecture notes, the textbook chapter and last week's slides next to this deck, and a question about one of them can be answered from the others.
  • Every answer that uses a slide cites it by number. One click opens that page with the passage highlighted, so you check the claim against the slide instead of taking it on trust.
Chat with your documents →

The rest of it

Review a deck that needs an answer

A pitch, a proposal, a plan: start a review and it leaves notes on the slides — unsupported claims, alternatives not considered, unclear steps, what works — then writes a memo with the questions to put to the author.

Laid out the way it was built

Shapes are drawn where the author put them, with their styling and layout inheritance. Tables stay tables, text in grouped shapes comes through, dark pages stay dark, and a two-column page is read in the right order.

A plain text view beside it

For the moment you want to select a long passage, or check what was actually extracted from a page that looks wrong.

Your own marks, per page

Highlights and comments belong to the page they are on and stay there, with ticks in the navigator showing where they sit.

Searched and asked with everything else

Slides sit in the same search as your PDFs, books and notes — speaker notes included — and a deck someone sent you is read alongside the reports it arrived with.

Nothing is converted

The .pptx in your folder is untouched. Pictures in it are shown in place and can be asked about like any other image.

One part of a loop: read, think, write

Your notes, and the two hundred PDFs you read to write them, sit in one folder — where a single AI works across both. Tools you would otherwise stitch together each hold one half of that.

  1. ReadThis pageRead the deck here — a card per page for what went unsaid, and questions about any part of a slide until it makes sense.
  2. ThinkSend a passage from a slide to a canvas card and work out what you make of it, beside the deck.
  3. WriteWrite your notes, or the reply, in Markdown in the same folder, and quotes from the deck keep their slide number.
All three work on the same folder of files — the one you already have →

Where it gets hard

  • It was not in the room

    The AI sees the slides and the speaker notes, not the lecture. What a card says the speaker would have said is a reading of the deck, not a record of the talk — if the lecturer made a point that is on no slide and in no note, it cannot know. Ask about what confused you; check it against your own notes of what was actually said.

  • It is a reader, not PowerPoint

    Animations, transitions and embedded video are not reproduced, fonts you do not have are substituted, and nothing is edited. The aim is a page you can read, quote and annotate — not a pixel-exact rehearsal of the presentation.

  • Generated decks can confuse the reading order

    A lot of automatically produced files use decorative boxes where a human would use a layout. Most of the time this is handled; when a page still comes out in a strange order, the text view is the fallback, and only that page is affected.

  • A deck of pictures has no text

    Slides exported as images — or a PDF of a deck that was scanned — contain nothing to read until they are recognised. That is the same deliberate, auditable step as any other scanned document.

  • The old binary formats are not supported

    .ppt files from before 2007 cannot be opened. Saving them as .pptx in any office suite is enough.

  • Explanations are only as good as the deck

    A page with three words on it and no notes gives the model very little to work with. On that kind of file the honest card is a short one — and putting the lecture notes or the textbook in the same folder helps more than asking again.

Common questions

Can I read a PowerPoint file without PowerPoint?

Yes. A .pptx opens laid out as slides — shapes where the author put them, tables as tables, images in place — with a plain text view beside it when a page looks wrong. Nothing is converted and no second copy is made.

Can the AI explain a slide to me?

It can write a card for each page covering what the speaker would have said out loud, how the page relates to the one before it, what is worth asking about it, and whether it can be skipped. It deliberately does not re-summarise the bullet points — those are already a summary. For one passage, select it and press Explain.

Can it help me catch up on a lecture I did not follow?

That is what it is best at. Each slide gets a card for what the speaker would have said; box any part of a slide you still do not get and ask about just that; then keep asking, with the whole deck — and any notes or readings in the same folder — in reach. Answers cite the slide they rely on. It reads the slides and notes, not the lecture itself, so it cannot know what was only said out loud.

Can it review a pitch deck or a proposal?

Yes. Start a review and it leaves notes on the slides — unsupported claims, alternatives not considered, unclear steps, what works — and ends with a memo: overall verdict, issues to address, what to keep and questions for the author.

Are speaker notes included?

Yes. The notes are part of the text of the document, so they are searched, quoted and available to the AI along with what is printed on the slide. On decks people actually gave, the notes are often where the argument is.

Open the deck someone sent you three weeks ago.

Reading and annotating it is local work. The per-page cards and reviews are the part that spends credits, and only when you ask for them.