Convert MOV to AVIF

The first wants a small file: a still from a video, going on a page, as few kilobytes as possible. AVIF is the correct answer for them and the rest of this page is about how well it does that job.

The second shoots in 10-bit, in log, or in HDR, and chose AVIF because it is the still format that can carry twelve bits per channel and a wide gamut. That is a true statement about AVIF and it is not available through this conversion. Saying so early is more useful than a page that quietly hands them an 8-bit file and lets them find out during a grade.

The frame is read out of the video through a canvas, and a canvas holds eight bits per channel. By the time the picture reaches the AVIF encoder it is ordinary 8-bit RGB, so the encoder writes eight bits — not because AVIF cannot do better but because there is nothing better to give it.

The effect is the one you would expect from any 8-bit conversion of high-bit-depth material: banding in gradients that were smooth in the original, and highlight and shadow detail that was there in the log file and is not there in the still. If the frame is a reference or a social post, none of that matters. If it is going through a grade, the extraction belongs in Resolve, Premiere or whatever holds the footage, where a 16-bit TIFF or an EXR is one menu away.

Nothing in this path attaches an ICC profile. An untagged image is treated as sRGB by everything that opens it, and Rec.2020 HDR or even Display P3 footage is not sRGB by a wide margin.

So the still is both narrower than the source and unlabelled about it, which is the combination that produces the "why does this look wrong" moment on a calibrated monitor. There is a defensible answer for the web — sRGB is what a browser assumes anyway — and no good answer for anything colour-critical. Know which of the two you are doing before you commit to a hundred frames.

Size. AVIF compresses a photographic still with AV1, the same codec family that would be compressing the video, and it does so with intra-frame tools far more sophisticated than JPEG’s. On a frame from real footage that typically means around half the bytes of a comparable JPG and a small fraction of a lossless PNG.

On a 4K frame that is the difference between a few hundred kilobytes and several megabytes, which is why the format earns its place on a page that carries a grid of video stills. It also handles the flat, gradient-heavy areas that video frames are full of better than JPEG does, so a sky or an out-of-focus background survives without the blocking a JPEG shows at the same size.

A 4K frame can take tens of seconds. AV1 encoding is computationally expensive by design — it searches a very large space to find the smallest representation — and doing it in a browser tab on a single still is the least favourable case there is.

That makes AVIF a publishing format rather than a working one. Take PNGs while you are choosing frames and editing them, and make the AVIF once at the end from the final image. Converting fifteen candidate timestamps to AVIF to pick one is a slow way to spend an afternoon for a file you are going to throw away.

The only control is the moment, in seconds. The still comes out at the resolution the video displays at — 3840 by 2160 from a 4K recording, and 2160 by 3840 from a vertical one, because the rotation a phone records is applied when the frame is read.

There is no resize here, which for AVIF is worth planning around. A 3840-pixel AVIF takes several times as long to encode as a 1280-pixel one and is far larger than any page needs. If the destination is a card or a thumbnail, taking a PNG, resizing it, and encoding the AVIF from that is both faster and produces a better file.

Whatever the video threw away is gone before the encoder sees anything. Motion blur, blocking around hard edges, coarse shadow detail and the softness of a fast pan were all decided by the video encoder, and AVIF preserves them efficiently rather than repairing them.

So the timestamp remains the most powerful control on the page by a distance. A frame taken while the subject is briefly still is dramatically better than one taken mid-movement, from the same recording, at the same settings. Choose the moment first and think about the format second.

Anything that has to be opened by other people’s software. Support in browsers is good and support elsewhere is patchy: a lot of desktop applications, content management systems, upload forms and email clients still do not know what an AVIF is, and a still that cannot be opened is worth nothing however small it is.

It is also wrong as a working file, because it is lossy and slow. Use PNG while you are editing, WebP when the destination is a page and you want no argument, JPG when somebody else has to open it, and AVIF when you control the delivery and the bytes are being counted.

Decoding the MOV and encoding the AVIF both happen in this browser tab. Nothing carrying the footage is transmitted, and the network tab will show that during a conversion.

It is worth noting what the alternative involves for this particular format. AVIF encoding is expensive enough that a hosted converter has to queue it, which means the master sits on somebody else’s disc for as long as the queue is — and for unreleased footage that is a longer exposure than a quick round trip would have been.