Not For The Faint-Hearted

Always topical this column. Last month Martin Christie was talking about the increasing use of artificial intelligence in all aspects of life, and the importance of recognising and addressing it. Since then a number of news items have popped up on my feeds, no doubt partly due to that very process of machine learning that directs me to things I might be interested in, and vice versa. Whatever the case, resistance to the relentless march of the appliance has spread through art and also photography, with a number of galleries and exhibitions refusing to acknowledge computer generated images as anything honestly creative, and refusing to display them. That’s not going to change the world, but it may make a few more people think, which was part of my recommendation here. We can’t avoid using AI as it has become so much a part of everything we do, but the key is to be aware of it and what it is doing, and question whether it is the best way of creating solutions, and whether there are alternatives. I remember watching a documentary on the artist David Hockney over ten years ago when he had been smitten by the appearance of the then new iPad, and was using it to make his art on the tablet. I did wonder then whether it could be considered proper art compared to his better known painted work. But then one of Hockney’s main considerations seemed to be that because it was backlit he could work while he was sitting in bed! Little did he know the can of worms he was opening with his patronage. My concern was that, as a photographer, I had just survived the first decade of digital imaging, when the very worst of Photoshop had had full reign. Not Adobe’s fault, they had simply provided the platform and the tools. But a whole new generation of camera users had arrived who had no previous experience or skills, empowered by the fact that you could just point and shoot and if you didn’t like the results, just press delete. Worse still, if you didn’t like the first results you could try and get creative on the computer. Having worked professionally with film for twenty years, the concept of just snapping away till you came up with something that worked was totally alien. Getting it right - or as close as possible - first time was everything. To make matters worse, because it was still something of a novelty - and no one had seen anything better - a good many of the computer-generated creations looked like a plasticine nightmare, and people just plastic puppets. It was a time when people still thought Windows Word Art was pretty neat too! Fortunately creativity has moved on since then, but then so has the machine. The justification of increasing AI is not that it will replace the human imagination, but free it from the more mundane, repetitive tasks to concentrate on the important ones. But I am old enough, and cynical enough to remember that we were told computers would take so much of the drudgery out of our working lives, the biggest worry would be what to do with all our leisure time! Whatever happened to that promise? The machines may be doing most of the leg work, but you can’t walk away and put your feet up. You need to let the electronics do what it does best and let human judgement and experience do the rest. Problem is the latter are not things easily taught, more gained in years of example, so the temptation is to simply rely on the former. If you read last month’s column you may recall the example of an old Victorian photo restored patiently by an expert, and one fixed in seconds by Photoshop’s new neural colour filters. That was never going to be a fair contest but the limitations of instant AI are clearly visible. Since then I’ve been able to install and try out more of these features with a Beta version of Photoshop, titled Strawberry Letter to identify the version rather than sequential numbers. This is a chance to try out future updates and provide feedback to the Adobe community about any issues - a bit of basic human learning. One of the new filters is Photo Restoration - something I do a lot of, the way and which is one of the biggest challenges for any automated action that relies on easily identifiable features often disguised in faded areas and ancient chemical stains. Not unexpectedly, as with the colouriser, some things it does remarkably well, and others equally badly. The biggest problem with all of the new Neural Filters - and an expanded list is shown here in the Beta workspace - is that there is little if any option for manual adjustment. It’s entirely AI controlled so you get what it wants. You can save it as an adjustment layer to apply edits in the normal way afterwards, but that is not much in the way of time saving for a good many images. For some it will work efficiently but it does seem a bit random unless you have a binary brain. Then I’m sure it all makes sense. Reviews on line seem to come up with the same conclusion: the good, the bad and the ugly, so I do recommend you try it and see if you can find a pattern that works so you will know when to call on it and when not to bother. In the shown example, with very clearly defined cuts and scratches, it does a pretty good job, while inevitably softening facial features. You can send your feedback directly to Adobe and at least play some small part in directing a process which will be the future whether we like it or not. As commercial printers we represent a small but important customer base as we have been using Photoshop for a long time and are likely to do so in the future long after others have migrated away to more simple apps. We need its complex and comprehensive image manipulation qualities because we deal with more complicated tasks. Popping a filter on a phone image to share on social media is easy, but preparing a file for print at any size is a whole different situation. I find that while a lot of advice from pundits on line is very useful, they are generally working with their own or submitted images which are usually fairly good sources in the first place. I have yet to see them tackle the dog’s dinner of customer files we have to tackle daily. And of course, if it doesn’t look good in print, it’s our fault, not their incompetence. It was a very expensive phone after all! A good example of combining advanced computer intelligence with an adjustable human interface is Nik Software which has been around a long time as a plug-in for Photoshop or as a standalone. I’ve been using the original version for over ten years and finally was convinced to upgrade to the latest version recently as it’s a significant improvement, and a single payment covers any future updates. The Nik suite is actually ten individual filters, each suited to particular tasks. Color Efex is the most comprehensive for colour images, and there is a more refined filter Viveza for even more precise tonal control. There is also Silver Efex for monochrome, Analogue for simulated traditional film camera styles. Define copes with noise reduction, and there are two sharpening filters, one for input and one for output particularly relevant for printing where there are presets for printer resolution, viewing distance etc. HDR and Perspective filters are fairly self- explanatory. So the obvious question is why pay for an extra piece of software when you can already have most of these functions in Photoshop or Lightroom? After all these are already heavily laden with features, many of which I have tried to reveal in these columns. Nik isn’t an alternative, although it can be used on its own, it’s designed to work complimentary to expand potential and because it uses lots of pre-installed presets with thumbnail previews you can easily see if one suits the direction in which you want to go. Once selected that preset is then open to a good range of adjustments to increase or decrease its effect. That can be saved and applied quickly to other images if you are editing several similar. Presets are available in Lightroom and in Camera Raw in Photoshop, and there are a good many plug ins available to download that you can add to this library. What’s different with Nik is that it uses its own intelligent software to do the fine tuning rather than Adobe’s Sensei. Long before the latter was introduced Nik was using what it called U point technology with control points that work a little like a mask, or a selection tool, isolating particular pixels and adjusting them in a variety of ways. If you do selections in PS or Lightroom you will know that often you have to try different tools to find which works best or even resort to manual sculpting of areas you want to work on. More recently these selections have got much more refined as a result of AI, and automated selections of sky and subject for example are pretty impressive. But there are times when a slip of the mouse or pen just gathers up a lot more than expected and you have to work backwards step by step. Even with automated selections there are often little isolated pockets that are easily missed on the screen and have to be removed by hand. Nik has been using their control point selection for over a decade and it is now even better and more precise. But a word of caution, this is not for the faint hearted. If you are already struggling to get your head round everything Adobe has to offer, this is probably not for you. Even as a user of ten years I am still learning all the new features and they are complicated simply because the task is complicated. Remember that modern digital images files are packed with pixels that have an integrated relationship with each other. This is not painting by numbers where you can just add a drop of cyan or magenta. As a printer if you are outputting files that contain many thousands of colour points it’s important to have as much control as possible before you hand it over to the machine that actually does the printing. So I’m not going to show any examples of how Nik works, there’s not enough space here, but you can get a thirty day free trial which includes lots of on line help and there’s a well established community of the faithful users who can also share experience and tips, as well as lots and lots of examples of all types. Like most photo editing software it is primarily aimed at photography rather than print, but control of the image is the crucial thing in either discipline. Try it and let me know what you think. https://nikcollection.dxo.com/download martin@colourfast.co.uk