After three decades of championing server-based document and imaging software, it is deeply satisfying to finally address the latest advance from our longstanding partner: Adobe Firefly Services (“FFS”). This is something that is not yet well-known, as initially Adobe is rolling it out to enterprise customers. But in my estimation, it (or something like it) represents the inevitable future of how content will be generated globally.
Consider this video, and the diagram below taken from it, while keeping in mind the power of the referenced Adobe products:
We’ve been working with this new technology intensely over the past year, on over 30 enterprise projects. While it has two major flaws (see the last section), the promise it offers is phenomenal.
Adobe has been completely dominant in the document and imaging software space from their earliest days. They pioneered PostScript, enabling the very first “desktop publishing” workflow. Later, they expanded into Photoshop, Illustrator, PDF, early web/video authoring tools, and more. Their “Creative Suite” circa 2005 represented the culmination of software tools for creative professionals, and much of that technology remains best in class to this day.
As a prepress technologist in the 1990s, I worked in a side of creative tech that they did not cover: server-side imaging. In Adobe’s early years, almost all of their products were based on desktop workflows. I remember asking for a Photoshop server in 1998, and in response I was asked “what’s a server?”
In similar fashion, when we first encountered InDesign in 2000, I begged them for an InDesign Server. We watched their progress in this direction in the very early 2000s. Beginning with AlterCast and Document Server (both short-lived prototypes), Adobe introduced FrameMaker Server, which was nothing more than the same install CD with the right to run it in batch mode. Finally in 2005, my request was honored (no thanks to me, it was just inevitable) when they released Adobe InDesign Server, software upon which our business is based to this day. Since that time, they’ve dabbled in similar software, acquiring Scene7 in 2007, and finally building Photoshop APIs starting around 2017.
Adobe expanded their Marketing Cloud considerably during the second decade of the 2000s, getting into DAM, Web Analytics, and Workflow, as well as data, commerce, and advertising. They had always been into Artificial Intelligence: Aldus, (acquired for PageMaker), had pioneered AI in the late 1990s in a serious way. They were well-prepared for that boom, once it took off.
By 2010, we could cobble together some rather amazing solutions based primarily on Adobe InDesign Server, the first successful server-based rendition tool from Adobe. These solutions integrated well with their other cloud-based technologies. We announced Silicon Designer (a web-based document editor) in 2009, while Canva was founded a few years later to bring a very similar solution to the consumer market. Meanwhile, Cloudinary built a brilliant business around server-side imaging in 2012. Figma, InVision, and Sketch also came to be at the same time. All were web-based, and somewhat competitive with Adobe offerings.
We expected Adobe to rise to the occasion, but we also knew that the depth and breadth of their stack was something of an impediment to their velocity. In contrast, their competitors chose one battle each, and as a result, they could move faster. By the late 2010s it was obvious that what the market really needed was full-out multi-channel. Adobe pretty much owned the best software in each category (Photoshop for raster imaging, Premiere/After Effects for video, Acrobat and InDesign for documents, etc.), but the aggregation of multiple channels remained a daunting challenge.
As AI exploded over the past 10 years, Adobe was perfectly positioned, due to their role as the leader in creative software. Their strengths included having some of the deepest early AI experience, as well as having a fully-owned library of stock images. This library was the result of Adobe having acquired Fotolia in 2011, which later became Adobe Stock. Adobe, unlike OpenAI or Midjourney, understood the nuance and criticality of respecting digital rights. When they released Sensei in 2016 and Firefly in 2023, their offerings were creative-friendly, unlike the majority of rushed-to-market AI imaging tools, tools which now face massive legal risk due to their disrespect for IP.
Firefly offered powerful AI graphic functionality, yet it demanded integration with a larger solution. Adobe Express (second generation released in June 2023), offered a browser-based “unified” tool for the broader market. Adobe Firefly services arrayed five core Adobe rendition technologies (Photoshop, Lightroom, InDesign, Premiere, Acrobat) in combination with Firefly’s state-of-the-art AI.
FFS is an enterprise graphic engineer’s dream come true. In theory, organizations can attain multi-channel publishing at full scale, with the highest quality and in complete harmony with the tools of choice used most widely by creative professionals. Coupled with Adobe’s numerous tangential offerings (DAM, workflow, site management, analytics, commerce, etc.), organizations have a content powerhouse that had never existed previously.
Does it magically deploy itself? Does it solve all of your problems right away? Adobe’s image, along with the amazing power of AI, could easily lead one to expect instant gratification. But the incredible breadth and depth of the solution is daunting, and best practices will take time to emerge, as the software is refined. Expect to see many wrinkles being worked out, as organizations blaze new trails in design and content workflows across all channels.
Does this involve Adobe Express? Not directly, yet. Express is like a streamlined, client-centric tool, similar to Firefly’s amalgamation of document, image, video, and 3D technology. It would seem inevitable that the two play nicely with each other, as Express already integrates well with the core Adobe creative tools. But Express is authoring-centric, targeting designers from entry level to professional, while Firefly Services is about dynamic real-time content, aimed at professionals and large organizations who require really custom, (sometimes even unique) workflows with maximal power, scalability and flexibility.
From my perspective (having been in this game for three decades), Adobe has an extremely bright future thanks to Express and Firefly Services. I love seeing Canva, Figma, Cloudinary, and the exploding plethora of competing AI-based design and imaging tools out there, but I also appreciate the phenomenal, extremely mature capability of Photoshop’s imaging algorithms, Premiere’s color correction and scalability, InDesign’s optical kerning and perfect print production, etc. It is just wonderful to see these all in one place. And as a web service!
While Firefly Services is a 30-year-old dream of mine finally come true, there are two big things holding it back, and that also dramatically limit its viability for real solutions. I am confident that these will be resolved, but it leaves us in an awkward state with InDesign Server: our solutions built on raw IDS are not only exponentially faster, they are also exponentially less expensive.
For some reason, Adobe chose not to tackle asset storage/persistence with FFS. This means that all assets need to be referenced via pre-signed URLs. With each request, assets must be loaded, creating latency that would be intolerable in many real-world situations. For FFS to serve a broad range of use cases, this limitation will need to be rectified.
As with other new Enterprise products, pricing does not offer a simple, predictable model that scales, but seems to require negotiation. Default pricing is extremely uneven and illogical, as the cost for minimalist operations is the same as the cost for extremely server-intensive operations. This can be resolved in the short term in the context of one-off deals, but must be figured out if Adobe is to compete in the real world against players such as Cloudinary.
Still, these two challenges are quite surmountable. I am thankful to have lived to see the time when server-based content generation has (almost) no limits.
