Convert MOV to PNG

That is what separates this from taking a photograph out of a video to send to somebody. You are going to open the still in something — Photoshop, Affinity, Figma, an editor’s title tool — and put text over it, crop it, grade it, mask something out. The file has to survive being worked on rather than merely look acceptable.

PNG is the right container for that stage because nothing it stores degrades. Every subsequent save is exact, an alpha channel is available when you start cutting things out, and there is no generation of JPEG artefacts sitting under the work waiting to be sharpened into visibility. The compressed version is something you make at the end, from the finished image.

This is the fact worth the most to this reader. The PNG is written with no ICC profile in it. An application that manages colour therefore has to assume something, and what it assumes is sRGB.

Modern phone and camera video is not sRGB. An iPhone records Display P3, which is a noticeably wider gamut, and camera footage may be in a log or broadcast space entirely. So a frame that looked vivid on the phone can open on a desktop looking flat, with reds and greens pulled in — not because anything went wrong in the conversion but because nothing told the editor what the numbers mean. Assign the correct profile once, note which one your camera uses, and it stops being a mystery.

The still comes out at the size the video actually displays at. A 4K recording gives a 3840 by 2160 image; 1080p gives 1920 by 1080. There is no resize control on this conversion and none is needed, because one frame of 4K is a large photograph rather than a hundred of them.

Vertical footage is handled properly. A phone stores portrait video as landscape frames plus a rotation flag, and the flag is applied when the frame is read, so a portrait 4K clip produces a 2160 by 3840 PNG standing the right way up. Tools that read raw frames without honouring the flag are where sideways stills come from.

You set the timestamp in seconds and the frame displayed at that moment is what you get. A tenth of a second either way is the single most effective control on the page, and it is not close: between a frame taken mid-movement and one taken while a subject is briefly still there is a difference no amount of processing recovers.

The reason is how video compression works. Bits are spent where a viewer notices motion and economised everywhere else, so anything moving quickly is described roughly. Convert two or three candidate timestamps and compare them at full size before you start building anything on top of one.

A still from video is never as clean as a photograph, and PNG cannot change that. The blocking around high-contrast edges, the smearing in fast motion and the coarse detail in shadow were all encoded into the video before the frame was extracted; storing them losslessly preserves them exactly.

It also means sharpening has to be handled carefully. A frame from a compressed source responds to sharpening by making the artefacts crisper along with the detail, particularly on skin and on flat backgrounds. Where the still is the deliverable and the subject was static, a much better answer is to take a photograph — cameras and phones can shoot stills while recording for precisely this reason.

A video frame is a rectangle with no transparency in it and nowhere for any to come from. The decoder hands over four channels with the fourth opaque from edge to edge, and the optimiser that runs before you get the file removes it: dropping a wholly opaque alpha plane is one of the standard PNG reductions and it is on by default here. What you download stores three channels.

Nothing is lost by that, because a uniformly opaque channel carries no information, and the file is smaller for not writing it out. It is worth knowing before you count on the channel being there. If you need one to work in — masking, cutting out a subject, compositing over another layer — your editor adds it the moment you start, and it will hold what you paint into it rather than what a conversion guessed.

PNG compresses losslessly, so the size depends on the content rather than on the dimensions alone. A 4K frame of a plain wall may be two megabytes; the same frame of a crowd, foliage or fine texture can be fifteen. Grain and video noise are the expensive part, because noise is by definition unpredictable and lossless compression cannot predict it.

That is worth knowing before building a folder of stills. If they are working files, the size is the price of exactness and it is worth paying. If they are going anywhere near a web page, run the finished image through a lossy format afterwards — a WebP or a JPG at the end of the process, made from the edited PNG rather than instead of it.

Each run produces a single image. Getting five frames means five conversions with five timestamps, and each one reopens the file and decodes forward from the keyframe before the moment you asked for, so the cost is much the same wherever in the recording the frame sits.

For a handful of candidates that is faster than opening an editor. For an actual sequence — every tenth frame, or a contact sheet of an entire take — it is not, and a desktop tool that steps through frames is the right instrument. This page is for the two or three stills you have actually decided you want.

The MOV is read, decoded and encoded to PNG inside this browser tab, using the codecs the machine already has for playing video. Nothing carrying the footage is transmitted, and the network tab confirms it during a conversion.

For this audience that has a specific value. The thumbnail is for a video that has not gone out, the mock-up is for a client who has not approved it, the composite is for a campaign that is not announced. Uploading a gigabyte of the source to get one frame back would be slow, and it would also put unreleased work on a server chosen from a search result.