{"id":21814,"date":"2018-11-14T16:40:43","date_gmt":"2018-11-14T23:40:43","guid":{"rendered":"http:\/\/introducing_fuif_responsive_images_by_design"},"modified":"2025-11-06T17:07:47","modified_gmt":"2025-11-07T01:07:47","slug":"introducing_fuif_responsive_images_by_design","status":"publish","type":"post","link":"https:\/\/cloudinary.com\/blog\/introducing_fuif_responsive_images_by_design","title":{"rendered":"Introducing FUIF: Responsive Images by Design"},"content":{"rendered":"

I\u2019ve been working to create a new image format, which I\u2019m calling FUIF<\/strong>, or Free Universal Image Format<\/strong>. That\u2019s a rather pretentious name, I know. But I couldn\u2019t call it the Free Lossy<\/strong> Image Format (FLIF) because that acronym is not available any more (see below) and FUIF can do lossless, too, so it wouldn\u2019t be accurate either.<\/p>\n

This is not the first time I\u2019ve made a new image file format. Some years ago, I created FLIF<\/a>, the Free Lossless Image Format. For lossless compression, FLIF still performs pretty well against the available alternatives. But, in practice, lossless compression is not usually what you need for image delivery. Photographic images, in particular, benefit from lossy compression, whereas lossless compression wastes precious bits on sensor noise. FLIF was simply not designed for that.<\/p>\n

\"FLIF\"<\/p>\n

One of the key features of the name FUIF is that it has an ambiguous pronunciation (just like GIF<\/a> and HEIF). For the record, the correct pronunciation is \/f\u0153\u028f\u032ff\/, and that\u2019s actually a word<\/a> in my native language, Dutch. But, personally, I\u2019m calling it \/fw\u026af\/ when speaking English. One could also argue that it should be pronounced \/fju\u02d0\u026af\/ (as in \u201cvaluing\u201d) since the U stands for Universal. Furthermore, the sequence \u201cui\u201d can be pronounced in many ways in English so there are plenty of alternatives to choose from: what about \/fu\u02d0\u026af\/ (like \u201cfluid\u201d), \/fuf\/ (like \u201cfruit\u201d), or \/f\u026af\/ (like \u201cguitar\u201d)?<\/p>\n

\"GIF<\/p>\n

But, naming and pronunciation aside, there are a lot of image file formats available already. So, if you\u2019re wondering why we need another one, I completely understand and I hope that this post (along with the subsequent posts in this series) will convince you that we do.<\/p>\n

Responsive Web Design<\/h2>\n

There is an increasing variety in end-user viewing devices: from tiny smartwatch screens and a huge spectrum of screen sizes and display resolutions for phones, phablets, and tablets to numerous laptop and desktop computer screens, all the way up to 4K or 8K televisions or projectors. Then you need to consider network conditions, which can range from slow 2G or 3G with poor reception to ultrafast, always-on fiber connections. As a result, it\u2019s no longer acceptable to create websites or apps that assume a given viewport width. Responsive Web Design<\/a> (RWD) has become the norm.<\/p>\n

\"Image<\/p>\n

The current approach to flexible images within RWD entails creating a variety of downscaled images from each high-resolution master image and then sending different images to users according to their browsing context. That approach helps reduce unnecessary bandwidth consumption. Imagine sending a 4,000-pixel-wide image to a smartwatch! There are downsides, however. You must create those downscaled variants and store them somewhere. To web developers, RWD is a major challenge and they must make many significant decisions, such as how many variants are needed, what sizes, and what sort of infrastructure to create and manage them with.<\/p>\n

\"Image<\/p>\n

Obviously, a service like Cloudinary<\/a> helps reduce those pains by providing such an infrastructure, creating downscaled variants on the fly and by using techniques like Responsive Breakpoints<\/a> to help automate most of the tasks involved. However, an image-processing infrastructure can\u2019t resolve some of the inherent problems with a multifile approach. For example, the multiplication of files to be delivered has a negative impact on cache behavior at the CDN level. It can also be hard to avoid the possibility that individual end-users might end up downloading several variants of the same image, such as when their viewport width changes because of a browser window resize or phone rotation.<\/p>\n

One File to Rule Them All<\/h2>\n

It\u2019s possible and desirable to have an image format that is responsive by design<\/strong>, meaning that you can use just one single file instead of many downscaled variants. Truncating this single \u2018master<\/em>\u2019 file to various offsets produces downscaled or lower-quality versions of the image, requiring a format that is essentially pyramidal and encoding the image from low resolution to high resolution.<\/p>\n

This idea is not new. JPEG already can use progressive scan scripts<\/a> and you can produce reasonably scaled 1:8, 1:4, and 1:2 previews by truncating a progressive JPEG with a suitable scan script. Hierarchical JPEG and JPEG 2000 took this idea even further. And, while that practice hasn\u2019t seen wide adoption, browsers already render progressive JPEGs progressively so feel free to take advantage of pyramidal image resolution right now.<\/p>\n

However, most of the recent image file formats have abandoned the progressive-decoding capability. WebP, BPG, HEIC (HEIF with a HEVC payload), and AVIF all cannot be decoded progressively; truncating a file would just produce the top part of the image. That\u2019s an unfortunate side effect of the fact that those image formats were derived from video codec intraframe bitstreams. Hence, in the context of video, progressive decoding of an individual frame isn\u2019t really useful.<\/p>\n

Progressive Versus Responsive by Design<\/h2>\n

In addition to RWD requirements, there is also increasing attention to the time it takes to load a web page and accord users the perception of a snappy experience. Low-quality image placeholders (LQIP)<\/a>, thumbnail galleries, and progressive rendering are ways to deliver aa preview on the screen before the entire image has been downloaded.<\/p>\n

\u201cProgressive decoding\u201d and \u201cresponsive by design\u201d are similar, though not quite identical, features:<\/p>\n