Convert MP4 to TIFF

The control asks for seconds from the start, and what you get is the frame being displayed at that moment — precisely, the last frame whose own start time falls at or before the number you typed. At 0 that is the first frame of the video, which is the default and is usually not the frame anybody wants.

Timestamps rather than frame numbers, because people know a moment is about eight seconds in and nobody knows they want frame 197. It also sidesteps the frame rate, which on real material is 25, 30, 29.97 or something a phone decided during a slow-motion segment.

The file is uncompressed and carries four channels of eight bits, so the picture data is exactly the pixel count times four bytes with a small header in front of it. A 1920 by 1080 frame is 8.3 MB. A 3840 by 2160 frame is 33.2 MB. A 1280 by 720 frame is 3.7 MB.

None of those figures depends on the video’s bitrate, its length or its file size. A 40 MB clip and a 4 GB clip of the same resolution produce identically sized TIFFs, because what is being written is the decoded frame rather than anything the codec did to it.

TIFF is a container with a long history of variants, and readers agree on less of it than the format’s reputation suggests. LZW is patent-free and universally read; Deflate is not universally read; JPEG-in-TIFF is a different image inside a TIFF wrapper and defeats the point of asking for TIFF at all.

Plain uncompressed baseline is the one thing every reader accepts, and the destinations that specify TIFF — a press, a scanner’s own software, an archive with a written policy — are exactly the ones where an unreadable file is the worst possible outcome. The cost is size, and it is paid deliberately. Compressing to LZW afterwards is one step in ImageMagick or Photoshop, and it is a decision you are in a position to make and this converter is not.

A lossless format preserves what it is given and adds nothing. The frame it is given came out of a video codec, which stored eight bits per channel, kept colour at a fraction of the resolution of brightness, and discarded fine detail wherever motion made it cheap to do so.

So the TIFF is an exact record of a frame that was already compressed. It is not a raw capture, it does not have the latitude of a photograph, and pushing it hard in a grade will show that quickly. The honest way to think about it is that TIFF stops any further damage — it does not undo the damage that arrived with the file.

The output is RGB with no ICC profile attached, because the frame is read out of the decoder as plain pixels and nothing on the way describes what those numbers mean. Video colour is conventionally Rec. 709 for HD material, and an editor opening the file will assume whatever its default is.

For a press, both facts need handling by somebody with the press profile: assign the source space, convert to CMYK, and check the result rather than trusting an automatic conversion. That is normal for supplied stills and it is the right place for the decision — a converter guessing at a CMYK profile would produce a file that looked plausible and printed wrong.

No EXIF, no capture date, no camera, no copyright field, no caption. The pixels come off a decoder and everything the video knew about itself stays in the video. That is a clean starting point and it is also a gap if the receiving archive requires those fields.

Write them afterwards in an image editor or a metadata tool, and take the timestamp from the video while you still have it open. A still labelled with the moment it came from is far more useful in an archive two years later than one labelled with the file name of the clip.

Motion is the enemy. A frame taken during a fast pan or a quick gesture carries motion blur that was invisible at 25 frames a second and is glaring at A4, and video encoders also spend fewer bits on frames where everything is moving. A still moment gets both a sharper capture and a better encode.

So it is worth stepping through several timestamps a tenth of a second apart and comparing, rather than taking the first frame that shows the right thing. The difference between two adjacent frames of the same action is frequently the difference between usable in print and not.

The still is written at the video’s own pixel dimensions, with no downscaling — one frame of 4K is a large photograph rather than a hundred of them, and somebody extracting a still usually wants every pixel that was there.

What that gives you at 300 dpi is the honest ceiling: a 1920 by 1080 frame is about 163 by 91 millimetres, and a 3840 by 2160 frame about twice that in each direction. Anything larger means printing at a lower resolution and deciding whether the material can take it. No file format and no conversion adds pixels that the camera did not record.

The decode and the encode both happen inside the browser tab, so the video is never uploaded and neither is the TIFF ever downloaded from anywhere — it is assembled in memory on your machine and saved from there.

That is a practical point as much as a privacy one on this pair. Sending a multi-gigabyte source file to a server so that it can send back a single frame is an absurd ratio of network to result, and the material stills get taken from — unreleased footage, evidence, medical and broadcast recordings — is frequently the sort that is not supposed to travel.