Fact-checking the GIF format

The Politifact article on the White House’s video “evidence” against reporter Jim Acosta looked plausible enough to me, until I got to the explanation of GIF files. It got significant points wrong, following common misunderstandings.

The regular readers of this blog mostly know what GIF really is, but this article may be a useful reference if you need to explain to anyone. The Politifact article says:

A GIF is basically a video with fewer frames — televised video has 29.97 frames per second, while a GIF made online might be reduced to between 10 and 15 frames.

Rafael Shimunov, an activist and longtime video editor, whose tweet on the exchange went viral, overlayed the original C-SPAN video with Watson’s and found there were discrepancies, unlike with the GIF.

Dan Voshart, a cinematographer based in Toronto, said it boils down to bad sourcing. When Watson took the GIF, which had half the frames as a video, and then turned it into a video, the software likely blended the missing frames. The choppy images might make it look quicker.

GIF is “basically” not a video format at all. It’s a still image format which has some primitive animation capabilities. Compuserve devised it in 1987 and revised it in 1989. It was intended as a way to create small image files. It includes some simple animation or sequencing capabilities. The 1989 spec says, “The Graphics Interchange Format is not intended as a platform for animation, even though it can be done in a limited way.” Video of live events is even more demanding than animation.

The format is primitive and would be obsolete except that it provides an easy way to post animations. All that’s necessary to display a GIF animation is to render the image. There was a massive controversy over a patent claim on the LZW compression algorithm, but it’s long since expired. GIF is patent-free, which is a huge advantage. No video support is necessary in the browser beyond the ability to render a simple graphics format.

However, its limitations are severe. A pixel is 8 bits, so a file can have no more than 256 colors. The width and height in pixels are each 16 bits, so the image dimensions don’t have a practical limit.

A GIF file can contain more than one block, and each block can represent an image. The specification recommends specifying a delay time between blocks, which is expressed in hundredths of a second. Each block is complete in itself; the format doesn’t attempt to save space and rendering time by storing the difference between successive blocks. The original idea was much closer to a slide show than a video. Theoretically, a GIF could have a hundred frames per second. Politifact’s claim that GIF has “fewer frames” than televised video is wrong as a description of the format, though it’s correct as a description of common practice.

Since GIF frames are intended to display as a sequence rather than a continuous video, software to render them generally isn’t optimized to keep the frame rate consistent. If it can’t keep up, it will slow down rather than drop frames or reduce resolution.

To complicate matters more, the GIF file doesn’t seem to be available on the Web. Sanders’ file was uploaded and displayed on Twitter. Twitter doesn’t display GIFs natively but converts them to videos. The lack of access to the actual file adds a large amount of guesswork to any analysis. Sanders’ source is apparently a tweet by Daily Wire. Some sources attribute it to InfoWars. The GIF seems to have circulated privately and been posted in several places on Twitter.

Buzzfeed News attributes some of the steps to Paul Joseph Watson at Infowars. It quotes a direct message from him: “Fact is, Daily Wire put up a gif, I download a gif, zoomed in saved it again as an mt2 file – then converted it to an mp4.” That says he did the zooming operation, which makes it appear to some as if Acosta struck the White House aide, on a low-quality file format, then converted it twice more. If it was then converted back to GIF for the Twitter upload, how many conversions does that make?

The circumstances make it hard to tell whether the file is unintentionally confusing or intentionally “doctored.” Whichever is the case, Sanders chose to make her case against Acosta by using a low-quality, multi-generation conversion of a news video. I don’t expect her to be a file format expert, but as the White House press secretary she has access to many resources. She would have had no trouble finding and using a decent video. Anyone who uses an antiquated still-image format to present video evidence when better sources are available is being highly irresponsible. This is a file format blog, not a political blog, so I won’t speculate here on whether she was willfully dishonest or not.

Update: A statement by White House counsel Kellyanne Conway looks like an admission that the video was intentionally altered, though she rejects the word “doctored.” She claimed, “They do it [speeding up a video] all the time in sports to see if there’s actually a first down or a touchdown.” This is total nonsense. I don’t watch a lot of sports, but replays slow down the video to get a better view of what happened. Speeding video up and removing frames only makes it harder to identify what happened.

Comments are closed.