Q: How is AI technology being used in imaging workflows?
A: AI technology is becoming a fundamental component of imaging workflows across the newspaper and magazine publishing industry, streamlining production tasks and boosting overall image quality and consistency.
The technology can appear as a complementary feature within established software, extending the functionality of existing tools, or as a self-contained system, capable of automating a wide range of previously manual processes.
One of the most established use cases is automated image enhancement. AI tools now routinely adjust tone, contrast, sharpness, and colour balance to meet the demands of both print and digital outputs. This ensures consistency across titles and channels, while reducing the need for manual correction. The result is faster processing and greater visual uniformity.
Background removal, traditionally a labour-intensive task, has also been transformed by AI. Cut-outs that previously took several minutes, even for experienced imaging professionals, can now be completed in seconds with minimal need for oversight. This has been particularly impactful in environments where large numbers of images need to be processed in a short time frame, particularly in daily newspaper production.
Another notable area is metadata automation. AI powered keywording and image recognition not only improve asset management within content systems but also support improved discoverability for archived content and SEO performance in digital publishing.
A particularly useful tool now being adopted more widely is Adobe Firefly’s Generative Fill feature. More of a semi-automatic process, this capability allows operators to extend the edges of an image in a visually coherent way, making it fit a required frame or layout.
Some publishers are also experimenting with generative AI to create entirely new images, particularly for digital formats where illustration or conceptual imagery is needed but resources are limited. While still in relatively early stages, these tools offer flexibility when stock photography or original content isn’t available. Interestingly, even when readers recognise it is not “real”, they still respond positively and engage with it.
As AI tools continue to mature, their integration into imaging workflows can only deepen, enabling even greater efficiency and creative possibility.
Q: In which areas of imaging workflows is AI having the most impact?
A: There are two distinct forms of AI technology utilised in today’s imaging workflows. The first is a manual, image-by-image approach, most notably seen in the beautifully striking images created by text prompt driven generators like Firefly, Ideogram, Midjourney and many others.
The ‘one-off’ pieces of AI created art have captured the attention of both the public and the press. While they beautifully demonstrate the imagination and rendering potential of AI imaging, it is the second form — batch processing automation — that has truly unlocked the transformative power of AI within publishing imaging workflows. Often overlooked, these solutions are powerful in their own right — quietly handling much of the heavy lifting in the imaging process.
Capable of delivering equally impressive results, at least from a professional imaging perspective, batch processing functions dramatically cut turnaround times, transforming tasks that once took skilled hands many minutes or even hours, into actions completed in mere seconds. Currently there are three main toolsets that offer batch process automation in imaging workflows:
- Image enhancement: The technology for generating enhanced images from originals has been available for 25+ years; the current solutions offer a multitude of settings and operations that can create some stunning results from terrible looking images, for both print and digital channels. By taking into account the limitations of press, paper, and digital screens, image enhancement engines produce consistent, high-impact results that reveal both the beauty and subtle details of each photo. Implementing an automated image enhancement strategy enables imaging teams to batch process the majority of their images, delivering multiple variants directly to layout teams. This gives time back to focus on more critical visuals that require expert attention and subjective retouching.
- Background removal: Background removal is perhaps the most visually striking example of image optimisation. While skilled operators can produce accurate cut-outs relatively quickly, complex images — such as a Wimbledon tennis player holding a racquet with a crowd visible through the strings — can take 10 minutes or more to complete. The latest AI engines reduce this process to just seconds, delivering highly accurate, ready-to-use results without any manual intervention. Traditionally, publishers have offshored background removal, with turnaround times ranging from 12 to 24 hours. With mature AI tools now available, this process can be brought in-house via the cloud, producing results in under 30 seconds. The accuracy is so high that two major UK newspaper groups report a consistent monthly success rate of over 95%. When scaled across multiple titles, the efficiency gains in themselves make for a compelling argument for further investigation regardless of a publisher’s size. Coupled with a minimal cost per cut-out, often £0.25 or less, the value of integrating an AI imaging platform becomes even more appealing.
- Image understanding: A very different, yet equally powerful, batch automaton process — image understanding — also minimises the time taken to identify what the content of an image is and create relevant metadata. As an example of image understanding, we all know when we look at a photo that it’s a child holding an ice cream on Westminster Bridge on a sunny day with Big Ben in the background, but to a computer its simply an array of pixels. Image understanding aims to identify and add tags that would include ‘child’, ‘bridge’, ‘summer’, ‘sun’ and a lot more. Even identifying Big Ben and adding GPS coordinates. Integrated seamlessly into the imaging workflow — yet extending its reach throughout the publisher’s broader ecosystem — this technology blurs the lines between roles and departments. The addition of enriched keywords significantly boosts discoverability within the content management system, allowing images to be easily found and reused time and again.
Q: In which areas of imaging workflows does more AI development need to take place?
A: While today’s powerful toolsets can deliver almost instant efficiency gains for publishers, especially in imaging and layout departments, there are still two key areas that stand to benefit even more as AI technology continues to evolve at rapid pace.
- Image upscaling: These tools create pixels to increase the details, sharpness and reduce noise in low resolution or poorly shot images. While this process has been getting some impressive results, it is not 100% ready for deploying in a fully automated workflow. Arguably perfect in a manual workflow, where skilled operators can edit the images further if required, the current results are not consistent enough to deploy into a full automated imaging workflow. There are also potential issues with introducing new details into a photographer’s original image, changing the creator’s vision for example, to consider. However, this form of AI imaging is incredibly exciting, being able to realise sharp, clear and punchy images from social media shots, screen grabs, thumbnails and compressed files, is a huge milestone for AI in the imaging world — and it is not that far off.
- Anatomical realism: AI engines have no concept of anatomy; they don’t know that a typical hand has four fingers and a thumb. Rather, the engines have learnt from the vast datasets, identifying patterns rather than any structural rules. Different to faces, which are typically captured with clear symmetry, hands and fingers are particularly tricky because they appear in countless variations and gestures within the dataset; pointing fingers, clenched fists, partially visible, holding hands or objects, distorted by perspective, in motion etc. All of which can lead the engines to believe that hands are highly variable in their shape, finger count and size. This issue is further compounded by the method AI image generators refine images from noise. Utilising diffusion techniques (this can be likened to a sculptor chiselling away at a block of marble; at first, it is just a rough shape, but over time, the details emerge) to remove the noise step by step, the engine guesses at what should be there based on patterns it has learned from the dataset. This will of course change as the engines evolve with higher-quality training datasets and better anatomical modelling. Until then, occasional extra fingers or distorted hands will likely remain a ‘tell’ of AI generated imagery.
Q: For publishers that get it right, what efficiency savings are they seeing?
A: A common response heard from multiple publishers around the world is the significant time savings — and, logically, cost reductions — achieved through the implementation of automated imaging workflows, all while improving image quality across their titles.
These imaging workflows bring a range of efficiencies. Some are immediate, others emerge through a broader commitment to automation, and many continue to grow over time as more departments and channels adopt the technology.
At the heart of this process is an image enhancement platform capable of meeting the high-quality standards set by art directors, while efficiently handling large volumes of images — even during peak production times. This includes all types of images such as cut-outs, black and white conversions, punchy hero shots, even thumbnails.
Imaging automation quickly delivers major gains: it saves considerable time in both imaging and layout, standardises image quality, and improves reproduction for both print and digital formats. Skilled imaging teams benefit too, freed from repetitive tasks, they can focus on high-priority images, helping to elevate overall quality while increasing capacity to manage additional titles, special editions, and more.
With highly quality enhanced images and precise cut-outs generated in just seconds, the time savings can quickly add up to several hours each day. Extrapolated out across a month, it is a very compelling financial discussion.
Another key efficiency boost applies especially to online titles. Often, images are uploaded without any enhancement because content creators lack the tools or training to improve them. With an AI imaging platform in place, images are automatically optimised — not just visually to make them more eye-catching on mobile and digital screens — but also technically, with the right file size, format, and SEO-friendly attributes. All of this plays a crucial role in improving search performance and overall visibility.
Key efficiencies:
- Faster turnaround: Automates editing tasks to help meet tight deadlines.
- Lower costs: Cuts down on manual work, freeing up budget and design resources.
- Better image quality: Delivers consistent, polished visuals that boost engagement.
- More output: Speeds up workflows so more content can be produced with less effort.
- Reuse old content: Makes archived images searchable and reusable, saving on new shoots.
Q: When using AI to create images, as opposed to enhance them, how does AI perform?
A: AI based image generation, where images are created entirely from textual prompts, has been the poster child of AI and has seen incredible development since 2023. Despite this, the results can still be inconsistent and need a watchful eye cast over them prior to publication.
That being said, AI systems such as Adobe Firefly, Midjourney, or Stable Diffusion can quickly produce striking visuals suitable for editorial illustrations, social media graphics or conceptual artwork providing publishers a convenient and effective method to generate one-off illustrative content. One newspaper group in the UK utilises Firefly’s capability for their daily infographic amongst other requirements. The learning curve for the users was short and, with an understanding of the text prompts, the process became a natural extension to their daily workflow, resulting in unique, attention-grabbing images on the pages.
Adobe’s Firefly has a standout tool that is arguably the most beneficial and practical interpretation of the manual form of AI imaging technology. The ‘generative fill’ feature extends the background of an image with content that blends naturally with the existing scene. Perfect for when the image doesn’t meet the aspect ratio requirements of its frame or the cover image needs a few more millimetres to wrap around the spine.
Imagine a portrait-oriented photo of the king and queen walking through Windsor Park however, it is published within a landscape-oriented frame. In seconds, photo realistic trees, grass, and sky fill the extra white space and, because the reader’s eye is naturally drawn to the focal point of the image rather than the edges, there is a more forgiving approach to realism. It is this functionality in particular that imaging departments gain most benefit from in Firefly.
It is worth mentioning that there are some notable limitations, particularly regarding anatomical accuracy and detail realism as mentioned above. When looking at images of people, we can inherently tell when something is visibly odd with the image; whether it be skin that is just too perfect, too many wrinkles around the eyes, or even that the irises aren’t perfectly round, there is always a ‘tell’.
Q: What do you expect to be the next phase of AI development for imaging?
A: One of the more exciting and tangible developments in AI imaging will be the ability to generate thematic, repeatable images. Rather than producing one-off, stand-alone visuals, AI will increasingly be able to create consistent sets of images that follow a specific visual style, brand aesthetic, or editorial theme. This has huge potential for publishers and content creators who need a steady stream of images that look cohesive across articles, campaigns, or issues.
Another critical area of development is the advancement of tools that can verify the authenticity of images, specifically whether they are AI generated or real. As generative AI continues to improve in creating highly realistic visuals, the line between real and synthetic imagery is becoming increasingly blurred. This presents major challenges, particularly for newsrooms, publishers, and any platform where visual trust is essential.
As these tools evolve, we’ll likely see deeper integration with content management systems, enabling AI to not only generate images but also anticipate visual needs based on the content pipeline, past usage, or current trends. That’s where things get really interesting!
Q: What's in the pipeline from Pixometry?
A: Pixometry’s image optimisation platform, developed over a number of years, is now in use globally across publishing and other imaging-intensive sectors, including creative agencies, retail, and supermarkets.
While the foundation of the platform remains focused on high quality automated image enhancement, it has naturally evolved to integrate third-party AI technologies provided by the world’s leading tech companies.
Now entering the sixth year of leveraging these incredible engines, Pixometry has an eye on new and emerging AI engines that will expand the scope and effectiveness of image optimisation.
As discussed above, there are AI imaging tools that are maturing to the point of consistently producing high-quality results in batch processing environments. At the same time, more specialised AI tools are emerging, each with varying degrees of suitability depending on the publisher's specific needs. Using the Pixometry cloud-based infrastructure will enable these solutions to be simply turned on for those users who are interested — think of it as an image optimisation pick ‘n’ mix if you will.
Alongside these developments, newer image file formats, such as JPEG XL (.jxl) and AVIF (.avif), are gaining interest for their ability to deliver higher visual quality with improved compression. These formats will become increasingly important for digital publishing, where image performance and loading speed are critical considerations.
Finally, Pixometry is collaborating directly with editorial and CMS software providers to build integrated apps and workflows. This allows content creators to generate the exact image they need, fully optimised and ready for use, with just a click, streamlining the entire imaging process for everyone involved.
Conclusion
As AI technology continues to evolve, its role in imaging workflows is shifting from optional enhancement to essential infrastructure. From automating time-consuming tasks like background removal and keyword tagging to enabling entirely new creative possibilities through generative tools, AI is proving its value across every stage of the publishing process.
For publishers, the opportunity lies not just in embracing these technologies, but in strategically integrating them, balancing automation with editorial control to ensure quality, efficiency, and consistency across print and digital platforms. As the tools mature, those who take a measured, informed approach to adoption will be best positioned to unlock long-term gains in productivity, creativity, and reader engagement.
About us
Pixometry: Intelligent Image Enhancement for Modern Publishing. Pixometry is an automated image enhancement platform built for publishers who demand clarity, consistency and speed. It refines contrast, sharpness, and colour balance while delivering precise local adjustments for every output; print, digital, or mobile.
AI powered features include highly accurate background removal, auto-keywording, and deep image understanding to boost SEO and workflow efficiency. Seamlessly integrating with editorial systems, our platform reduces repetitive tasks, enhances visual impact, and ensures professional results across high volumes.
Trusted by publishers around the world, Pixometry empowers imaging teams to deliver standout images, faster, smarter, and tailored for today’s multi-channel publishing environments.
