How to drive this thing.

Everything DeckDriver does, in the order you’ll need it. Nothing here is a preview of something unbuilt.

How DeckDriver works

You talk. Chrome turns your speech into text. DeckDriver compares what you just said against what each of your slides is about, and either moves the deck, offers to, or stays put — whichever you told it to do before you started.

The honest version of the sentence above, in three parts:

  1. Recognition is Chrome’s. The browser’s own speech API produces the words. That’s the one part of the loop that isn’t ours, and the reason DeckDriver is Chrome-only.

  2. Matching is local. Your slides were summarised once, at upload, by an AI pass you can read and correct. During the talk, the comparison between your words and those summaries runs in your browser — no server, no round trip, no rate limit.

  3. Your rules win. Anything you’ve written down — a phrase that must jump to a slide, a slide that must hold for 30 seconds, never going backwards — is applied before the matching gets a vote. And your key press beats all of it.

Getting your deck in

PDF is the format to bring. It’s what gives you your real slides, pixel for pixel. A genuine .pptx works on its own too; bring both and you get the PDF’s images with the PowerPoint’s text and speaker notes.

Made inExportWorth knowing
CanvaShare → Download → PDF StandardPDF Standard is enough; Print quality just makes a bigger file.
Google SlidesFile → Download → PDF Document (.pdf)Speaker notes don’t come through a PDF — add the .pptx too if you want them.
KeynoteFile → Export To → PDFTick “Print each stage of builds” so an animated slide doesn’t collapse into one page. That does mean one slide per build step for now — grouping build steps back into a single slide is coming.
PowerPointSave as PDF — or just upload the .pptxA .pptx alone works. Pairing it with the PDF export gives you both real slide images and your notes.
PowerPoint 97–2003 (.ppt)Open it, then export as PDFThe old binary format can’t be read directly, even when it’s been renamed to .pptx.

Upload rejects anything that isn’t really a PDF or a real .pptx — it checks the file itself, not its name — and tells you what it actually found. Nothing is saved for a file that can’t be read, so a rejection costs you nothing.

Presenting

Pick how much it drives

You choose a mode on the way in, and you can change it mid-talk. First time out, pick Assisted.

Automatic
It changes slides for you — for when you trust it. DeckDriver advances the deck itself when it's confident. Take over any time — your key press always wins.
Assisted
It suggests, you press Enter — best your first time. DeckDriver proposes the next slide and waits. Nothing moves until you accept.
Manual
You drive, it watches. Arrow keys and the jump grid only. No AI navigation, no voice commands.
Rehearsal
Practice run — records your timing and gives coaching. The engine watches and suggests without driving, so you can see what it would have done.

Keys

← / →
Previous slide / next slide
Enter
Accept the suggested slide
Space or A
Pause the AI / let it drive again
L
Lock the deck — only you can move it
G
Jump grid — or type a slide number, then Enter
Q
Q&A mode — the deck holds still
H
Hide or show the presenter overlay
T
Hide or show the live transcript
?
Open this card again
Esc
End the presentation

Things you can just say

These are matched as commands, not as content — so ordinary speech about “the next quarter” never moves the deck by accident.

go to slide 7
Jumps straight there, whatever you were talking about
deckdriver next / deckdriver back
One slide, on command
deckdriver pause / deckdriver resume
Stops and restarts AI navigation
let's take some questions
Enters Q&A — the deck stops moving on its own
back to the deck
Leaves Q&A
let's go back to the pricing
Finds the slide about that topic

The audience window

Press Audience in the presenter overlay. A second window opens with nothing in it but your slide — no transcript, no confidence bar, no controls, no cursor. Drag that window to the projector or share it in your call, and keep the presenter view on your own screen. It follows every slide change automatically.

Rules cookbook

Rules are the parts of the talk you won’t leave to judgement. They’re written per deck, they always apply, and when two could apply the one higher in your list wins. Most decks never need one.

Jump phrase

When I say “and now the demo” → go to slide 12

For the one transition you refuse to leave to inference. Exact words, exact slide, every time.

Never backwards

Never go backwards automatically

You can still go back with ← or the jump grid. The AI can’t.

No skip range

Don’t auto-advance past unvisited slides between 4 and 9

The compliance slides everyone has. Nothing in that range gets skipped, however you tell the story.

Minimum dwell

Stay at least 30s on slide 4

The slide you always rush. Nothing moves off it before the time is up.

Q&A behaviour

During Q&A: fully paused (or suggest only)

Answers wander across the whole deck. This decides whether the deck follows them or holds still.

Rehearsal & coaching

Rehearsal mode drives nothing. It listens, suggests, and records — so you can see what it would have done without it happening to you mid-sentence.

When you end the run you get a report about that run: total time, words per minute, filler words per thousand, how many times you paused and for how long, how many slides you actually covered, and how long you spent on each one. Then a written coaching pass over those numbers — about your session, not about public speaking in general. You can regenerate it if you want a second read.

Troubleshooting

Chrome isn’t hearing me

Chrome asks for the microphone the first time you present. If you dismissed it, click the icon at the left of the address bar → Site settings → Microphone → Allow, then reload the page. If the presenter view says “Microphone stopped”, press Restart listening — recognition sometimes gives up after a long silence and won’t retry on its own.

If it says the mic is unavailable, you’re not in Chrome. Everything except the listening works; the keyboard still drives the deck.

It didn’t switch when I moved on

By design. DeckDriver won’t move on a maybe. Before it switches, the new slide has to be a clearly better match than the one you’re on — not a marginally better one — and it has to stay the better match for a couple of seconds running. It also won’t jump off a slide you only just arrived at, so a switch that’s “due” can wait a few seconds.

That trade is deliberate: a deck that sits still one sentence too long is survivable on stage, and one that flickers between two slides while you talk is not. If it waits when you need it now, say “deckdriver next”, or press →.

It switched to the wrong slide

Two fixes, in order. First, open that deck’s mapping and read what the AI thinks the slide is about — a wrong title or summary is usually the whole story, and correcting it fixes every future run. Second, if two slides genuinely cover the same ground, write a jump-phrase rule for the one sentence you always use to introduce the right one.

In the moment: press ← to go back, and the deck stays where you put it. Your override always wins over the engine.

My deck uploaded but has no slides

The upload didn’t finish, and there’s nothing to map or present. Delete it and upload again. The usual cause is a PowerPoint 97–2003 file saved with a .pptx name — open it in PowerPoint, Keynote or Google Slides and export it as PDF first.

FAQ

What files can I upload?
PDF and PowerPoint (.pptx). You can upload both together — slide images come from the PDF, text and speaker notes from the .pptx, as long as they have the same slide count. Keynote, Google Slides, Canva and the old .ppt format all export to PDF; see the table above.
Do I need an account?
No. Try it free gives you a guest workspace with one deck, no sign-up. If you want to keep it — and upload more — you can turn that guest workspace into a real account without losing the deck you already made.
How much does it cost?
One deck is free. Pro, with unlimited decks, is coming — there’s no checkout to walk into yet, and nothing asks you for a card.
Who can see my decks?
Only you. Decks are private to your account and are never used to train a model. Speech recognition is Chrome’s own; the matching, the rules and the decision all run on your machine.
Does it work offline?
The decision loop does, once a deck is open — it’s running locally. Chrome’s speech recognition itself needs a connection.
Why Chrome?
The live loop depends on the browser’s speech recognition API, and Chrome is the only browser that implements it well enough to trust on stage. Other browsers can open DeckDriver and show your deck; the microphone just won’t drive it.
Try it free — no sign-upChrome required · PDF or PPTX · your decks stay yours