How is AI revolutionising Large Format Print?

Nessan Clearly discusses how AI in print relies on data pattern matching, already enhancing software for large format providers. He predicts that this will result in increased AI integration in workflow planning, job queue management, colour correction, image upscaling, and predictive maintenance via sensors and vision systems, ultimately streamlining operations and offering greater flexibility.

Judging by newspaper headlines from around the world, there’s a great deal of interest now in Artificial Intelligence, or AI, as multinationals and governments talk openly about using this technology to optimise their output and generate more income. But how will this filter down to the print industry and in particular the many large format print service providers that serve the sign and display market?

Ironically, there isn’t really much intelligence involved in AI. Intelligence normally implies some level of cognitive reasoning to solve problems. But AI is simply hoovering up large amounts of data, including past examples of reasoning, and matching this to patterns that emerge from that data. In any given situation, most AI systems are just taking the best guess based on the available data (and sometimes this generates falsehoods). We’ve previously used algorithms to do something similar, but what really sets AI apart is the huge datasets of raw information that these systems ingest. This itself has been made possible by advances in raw computing power.

Much of the current hype is based around generative AI, which can create solutions based on prompts fed into the system. However, such systems need to be pre-trained on existing data, so that they are more accurately known as Generative Pre-Trained, or GPT. The best known of these is probably ChatGPT.

At a general level, these systems can be used in practical ways, such as to create initial drafts of letters accompanying quotes or even whole sales presentations. Since AI is not perfect, this is best used for initial drafts rather than final copies, but nonetheless, this is enough to let users do more in less time.

Some AI variations are more focussed on creating images and are already good enough that they can allow people with some level of knowledge to produce work that might otherwise be beyond them. This greatly lowers the costs involved in producing artwork, even just to create quick mock-ups before bringing in a professional designer. It can include things such as the graphics for an exhibition stand, or patterns for digitally-printed clothing.

However, don’t think that AI will just get on and do all the work for you. The quality of the results depend on the quality of the prompts you feed into the system, and this requires some training in its own right. It means that someone who has been trained to use AI will be able to produce work in other areas where they have not had the same degree of training. This type of multi-skilling is reflected in the phrase: “AI won’t take your job but someone using AI probably will”.

AI and print

There are a number of areas where AI can have a profound impact on printing, some of which we can already see. AI technology is not quite as new as some of the newspaper headlines suggest. It’s been in use for over a decade now and is already used within some of the software available to the print industry. The bit that is new is that where previously it’s been used in individual isolated products, we can now expect to see those solutions being more integrated, with more automation powered by AI right through the production workflows.

Caption: Esko Phoenix uses AI for automated imposition and nesting.

This will include planning the most efficient way to produce jobs, which can be used both to generate a quote and to plan the route through production. AI is also increasingly being used to manage the job queue. An intrinsic part of that is the way that the different jobs are ganged together or nested around each other on the sheet or roll to minimise the amount of media wasted. This has evolved from imposition, that is, arranging pages on a press sheet to optimise the layout. There are several existing programs, such as Esko Phoenix, that already use AI for imposition, and this technology has now carried over into large format job management. There are a number of different parameters to consider, including how those different jobs will be finished and the agreed delivery times for each of those jobs. And of course all of these factors will constantly change throughout a day as more orders come in. But the ability to cope with late orders is an essential component of today’s fast turnaround on-demand ordering, which customers have now come to expect.

Another area that is already benefiting from AI technology is colour management. Many customer files include colour management mistakes, some of which such as the wrong colour space, can be easily corrected. A lot of colour management programs now also use AI to help tweak the colours in these files to produce a more natural look, which can cut down on customers rejecting jobs because of perceptions around the colours printed. Often this is just labelled in the software as an Automatic enhancement and not flagged up as being based on AI.

Caption: HP’s latest Latex printer, the 830-series, has a built-in spectrophotometer for automated AI-powered colour management.

Colour management also covers other areas from linearisation and calibration of printers, through to making profiles for all the different media and choosing the right settings for every job. Gradually we are starting to see more spectrophotometers integrated into large format printers to enable all this colour management to be automated, with minimal operator input. Again, this also is being driven by AI.

AI is also being used to improve other aspects of image quality. That could include upscaling images, essentially by using AI to add additional pixels to an image based on its best guess after analysing the existing pixels. That’s an obvious benefit to anyone looking to create large billboards. It also helps overcome issues with customers sending in low resolution files.

One area that is really benefiting from greater use of AI is that of predictive maintenance. The pandemic, with its various lockdowns, forced most vendors to rethink their service offering as it became more difficult to send engineers out on site visits. But AI has allowed those vendors to analyse the data they already have around the reasons for service call-outs to anticipate when each component might fail.

In order to add more automation, large format printers will have to generate more data. So we will almost certainly see more sensors added to large format printers over the next few years to detect potential problems, and to pick up on parts that are failing. The bigger printers are bound to gain more vision systems, to check that what has been printed matches the customer’s expectations.

Another area that AI could lend itself to is inventory management, including ordering in more consumables and spare parts as needed. The trick is not to over-order - which will tie up capital in stock - but at the same time not to risk running out of something that will hold up production.

AI also underpins many of the analytics services that some printer vendors have introduced, as well as some MIS. It’s good at spotting patterns, such as a drop in sales in some areas, and coming up with solutions based on ideas that others have successfully tried but tailored to an individual user’s specific situation based on their data.

In conclusion, AI is far from a future technology but one that is already embedded into much of the automation software that is already in use. And we can expect more hardware - in the form of sensors, spectrophotometers and vision systems - to be added to printers to facilitate that automation. AI should make it easier to integrate solutions between different vendors. That in turn should give large format users more freedom to build production workflows that better suit their individual requirements.