2015: Jon Sneyers introduces FLIF
The story of FLIF begins in Leuven, Belgium, with a single researcher. Jon Sneyers was working at Cloudinary at the time on image-processing algorithms and had a theoretical idea he wanted to try out in practice: a lossless image format that would beat every existing lossless codec in compression efficiency.
In October 2015, Sneyers released FLIF (Free Lossless Image Format)as an open-source project. The first benchmarks kicked off a small web storm: FLIF delivered lossless images typically 14–35 % smaller than optimized PNG, often better than WebP lossless, and even better than BPG lossless (see our BPG history). For some image types, FLIF also outpaced JPEG 2000 in lossless mode.
The technical innovation: MA-trees
FLIF's core innovation was a compression algorithm with the acronym MANIAC (Meta-Adaptive Near-zero Integer Arithmetic Coder). At its heart it built on MA-trees (Meta-Adaptive Trees), a variant decision-tree structure that learns, at encoder runtime, which prediction methods work best for each image region.
Instead of a fixed prediction (like PNG's Sub/Up/Average/Paeth filters), MANIAC decides dynamically per pixel region: here there is a gradient, so use gradient-based prediction; over there is a hard edge, so use edge-preserving prediction. This adaptive choice is what delivers the efficiency gains.
Progressive coding as the killer feature
A particular strength of FLIF was its progressive decoding. Unlike PNG (sequential) or progressive JPEG (with scan-layer steps), FLIF could be decoded at any ratio. A web server could start a FLIF stream, the browser first shows a low-resolution preview, and the client can decide at any moment, "okay, I have enough, stop the download." The server can end the connection at that point, and the image looks correct at the quality reached.
This property was especially attractive for bandwidth-conscious applications: mobile apps, slow-connection optimization, image-streaming CDNs. A 500 KB FLIF could already be recognizable as a rough preview at 50 KB downloaded, and grow sharper with every additional KB.
Universality: lossless for everything
FLIF's design goal was universal lossless. Instead of a separate lossless and lossy pipeline (as in WebP), a single algorithm was meant to compress all content types optimally: photos, graphics, UI sprites, animation frames, even medical imaging. Sneyers' benchmarks showed consistent advantages across all tested image categories.
In addition, FLIF supported animation, alpha transparency and 16-bit color depth — all lossless. Conceptually it was a universal container for any kind of pixel data, with better compression than all established competitors.
Adoption: enthusiastic and marginal
FLIF got a lot of attention in the tech community. Hacker News, Reddit and image-compression forums discussed it at length. Some open-source tools like ImageMagick added FLIF support. But browser adoption never came. Mozilla, Chrome and Safari did not implement FLIF natively.
Three reasons for the missing mainstream success: first, FLIF was lossless only, while web applications mostly serve lossy content; second, the encoder performance was poor (lossless compression with MANIAC is computationally expensive); third, a commercial driver was missing — no major tech company had a strategic interest in pushing FLIF.
2017: FLIF becomes FUIF
Sneyers himself recognized that FLIF in its pure form would not find mass adoption. In 2017 he developed an evolved variant called FUIF (Free Universal Image Format), which additionally supported lossy compression. FUIF was positioned as a universal image format — lossless and lossy in the same container, with progressive streaming and all the FLIF advantages.
FUIF had adoption hurdles of its own, but it resonated at one decisive point: in 2017 the JPEG committee was looking for proposals for a next-generation JPEG standard. Sneyers submitted FUIF as a contribution.
2018–2021: FUIF becomes JPEG XL
In the JPEG competition, FUIF met PIK, a parallel research project by Google employee Jyrki Alakuijala (see our JPEG XL history). Both codecs reached the final round. Instead of choosing one, the committee merged both. From PIK and FUIF came JPEG XL, which was ratified in 2021 as ISO/IEC 18181.
JPEG XL adopted central FUIF concepts: the MA-tree-based coding (for the lossless mode), the progressive decoding, the multi-mode pipeline. Sneyers became a co-editor of the JPEG XL specification. FLIF and FUIF thereby effectively merged into JPEG XL — no longer a standalone format, but the technical predecessors of an established ISO standard.
FLIF today: archive status
FLIF's official project is archived on GitHub. Sneyers himself stopped standalone development in 2019 and focused on JPEG XL. The FLIF file extension still exists, but no one is seriously producing new FLIF files. Anyone who comes across an old FLIF file can convert it to PNG or JPEG XL with ImageMagick or the original Sneyers decoder.
Sneyers' career is a success story. He still works at Cloudinary, is one of the key people behind JPEG XL and influential in the JPEG committee. FLIF was, for him, the door-opener to an industrial format standard.
What we learn from FLIF
- Research codecs are important forerunners. FLIF and FUIF existed only briefly as standalone formats, but their concepts live on in JPEG XL.
- Standalone formats have a hard time. Even technically superior formats need industrial adoption. Solo-developer formats rarely find their way into browsers.
- Committee standards are sometimes the right destination. FUIF disappeared as a standalone format, but through the merger with PIK it became part of an ISO-ratified standard with long-term maintenance.
When FLIF is the right choice
Practically never as an active delivery format. The only relevant use today is migrating old FLIF archives to modern formats — typically to JPEG XL, which inherits most FLIF properties. For modern lossless web delivery, the right choice is PNG (see our PNG history) or WebP lossless. For professional photo archives, JPEG XL offers the FLIF concepts with ISO maintenance.
Sources
FLIF — original source code (archived) · flif.info — original project page · FUIF — Cloudinary source code · JPEG XL community portal · Cloudinary — JPEG XL adoption report · Wikipedia — FLIF · Sneyers, J. & Wuille, P., "FLIF: Free Lossless Image Format," IEEE International Conference on Image Processing 2016.