Videos

Turn screenshots into a video

You have a folder of screenshots from documenting something, and they are already in roughly the right order. The work is deciding which of them deserves its own step, which are duplicates, which belong together, and what the narration says over each one.

Create a video with OneCraftWhat the vision pass reads out of each shot, and how the order is decided

The review step showing four uploaded screenshots numbered down the left, the first one open, with No zoom, Zoom in and Close-up choices and an Add 4 images button. — Uploaded screenshots are reviewed in order before they enter the video. This is where you drop the ones you regret and mark the spot each shot should move toward.

Every screenshot is read before anything is built

Each image goes through a vision pass that returns what kind of thing it is, a summary of what it shows, a few key points, whether it is usable as a scene asset, and a sequence hint. That last one is the interesting field: it is evidence about where the shot sits in an order, such as a step number visible in the interface or a stage in a wizard. Separately the real pixel size is read from the file header rather than guessed, so a tall mobile capture is cast into a tall slot.

The director arranges rather than accepts

It is told explicitly that the screenshots are the core of the tutorial, that it should arrange the ones that fit into a logical teaching order using the sequence evidence, and that it should leave out irrelevant, duplicate or low quality shots. So an upload of fifteen screenshots does not become fifteen steps. It becomes the six or eight that actually teach something, in the order somebody would perform them.

One screenshot per step, on purpose

The rule is one screenshot per step scene so the viewer sees each screen properly. Shots that genuinely belong together, three variations of a settings page for instance, may share a single recap scene at the end using a step cards, quad or slideshow layout, but never as steps in the middle. The same full screen layout is held for every step, so the badge and the instruction line stay in the same place and the eye stops relearning the frame.

Shots with nothing to show

Some steps have no screenshot worth taking: a decision, a wait, something happening outside the software. Sixteen step layouts draw the interface themselves, covering a terminal, a keyboard shortcut, a toggle, a form being filled, a menu being navigated, a warning, a tip and a result. So a gap in your screenshot set does not become a gap in the video.

How it works, in three steps

Step 1
Upload them in order
Display order is what the analysis is indexed against, and it is the strongest signal the director has after the sequence evidence in the images themselves.
Step 2
Describe the steps in words too
The brief is passed to the vision pass as context, so telling it what the software is makes every image description sharper.
Step 3
Drop the ones you regret in the review grid
Uploaded images land in a review dialog first. Remove the near duplicates yourself rather than hoping the analysis catches them.

The full walkthrough with screenshots is in the guide Make a how-to video with AI.

Limits worth knowing

Eight images per generation. A longer process has to be split into more than one video.
Screenshots are not cropped, redacted or retouched. Anything sensitive in the frame stays in the frame.
Uploaded pictures are never upscaled, except a hero slot filling most of the canvas, and even then only up to two and a half times.
Tutorial videos are a paid feature from the Starter plan. Screenshots in a promotional video are not gated.

Questions people ask

Does the order I upload them in matter?

Yes, it is the default order and the analysis is indexed to it. The director can rearrange when a screenshot carries clear evidence of its own position, but starting in the right order gives it much less to guess.

What if two screenshots show the same screen?

It is told to leave out duplicates, and it usually does. Where two shots genuinely differ in a way that matters, say so in the brief, otherwise one of them will be dropped.

Can I highlight part of a screenshot?

Yes. In the review grid, click the spot where the action happens and pick Zoom in or Close-up. The video glides a cursor to it, clicks with a ripple, then eases in over about a second.

Do phone screenshots work?

They do, and their shape is read correctly, so a tall capture lands in a tall slot instead of being cropped. There is a phone frame layout that suits them, and one that walks through mobile taps.

Make your own video

The button opens the generator with this use case already described. Change the wording to match yours, generate, then edit anything you like.

Create a video with OneCraft

Zoom into a screenshot

A full screen screenshot in a video has a problem: the thing you are talking about is a small button somewhere in it. Marking a focus point solves that by turning one static picture into a small piece of choreography, a pointer arriving, a click landing, and the frame closing in on it.

Screen recording to tutorial video

Be clear about what this is before you use it. You share your screen once, then press a button for each step and it snaps a still frame. There is no recording, no video file, and nothing is processed from footage afterwards. What you end up with is a set of PNG screenshots and a walkthrough built from them.

AI tutorial video maker

A walkthrough fails for boring reasons: it moves too fast to follow, it changes layout every step so the eye has to relearn the screen, or the narration says something different from the words on the picture. The tutorial director is a set of rules written specifically against those three failures.

More finished work of this kind is on the video examples hub.