Adobe Max 2019 wrapped up yesterday, and over the past week the company revealed a bunch of new automated capabilities it’s currently developing for its various applications – both on desktop and mobile. These demos are always crowd-pleasers and tantalising teases of how users might soon be able to further streamline their workflows. But in recent years these sneak peeks have also provided a look at how artificial intelligence promises to radically change all the digital tools we use, as more often than not, Adobe’s latest and greatest leverage the company’s Sensei deep learning platform to pull off their seemingly magical feats.
Not to be mistaken with the classic children’s toy where plastic pegs were stabbed into a glowing board, Adobe LightRight might be the holy grail for photographers who incessantly tweak and adjust every aspect of their photos in apps like Adobe Lightroom. Using Adobe Sensei, LightRight can be used to radically adjust the lighting in a photo after it was taken, and not just in regards to the overall exposure or brightness. The tool can calculate the 3D dimensions of objects in a 2D image and even recreate the proper shadows as the user adjusts the intensity and position of a simulated sun. One day photographers might not even have to bother looking through their camera’s viewfinder, as everything wrong with a shot could be easily fixed by an AI after the shutter button is pressed.
As progress on leveraging AI and deep learning to manipulate photos and videos accelerates at an alarming pace, so does the development of tools designed to help humans recognise what’s real, and what’s been faked. ProjectAboutFace uses machine learning to analyse photos and look for the telltale signs that a digital manipulation has taken place, like patterns of pixels that have been copied from other areas, or interpolation that’s been used to fill in missing parts. It will not only provide an assessment of a photo’s authenticity and probable manipulation, but it can also be used to undo those edits, without having any prior knowledge of what the original photo looked like. Eventually Photoshop could not only be used to commit a crime, but uncover one too.
Selfies and front-facing cameras on our mobile devices have all but solved the challenge of fitting a bunching of friends into a group shot – including the photographer. But for those times when you’d rather use the nice camera you spent a thousand bucks on, but don’t want to fiddle with timers, or carry a tripod, Adobe has developed an AI-powered tool that analyses a series of photos, determines who is missing from a specific shot, and then automatically crops and composites them into the photo so that the person behind the camera isn’t missing.
Animation helps democratise filmmaking because you don’t need millions of pounds to pay for actors, sets, costumes, and post-production – just a solid script, a microphone, a computer, and some software. Assuming it’s ever made available to consumers, Adobe SweetTalk also means you don’t actually need to know how to animate a character. Using Adobe Sensei it can analyse pre-recorded audio and then automatically bring a 2D static image to life, including mouth movements and facial distortions as a character talks. The results aren’t quite Pixar-level, but every filmmaker has to start somewhere.
One of the most tantalising promises of AI and deep learning is that it can help even the most untalented artists produce masterpieces without having to spend a lifetime practising and perfecting their craft. ImageTango is a tool that intelligently applies the texture from one image to another, so what starts out as a barely recognisable bird doodle on the back of a receipt can be turned into a near photorealistic sketch by simply pointing the tool at photos of the type of bird you were trying to draw.
You can spend hours perfecting every aspect of a video you’re making, from the camera work, to the lighting, to post-production elements like colour correction and visual effects. But the final product will still seem amateurish if you don’t take as much care with the audio. With AwesomeAudio, Adobe wants creators to know they no longer have to worry about all the background noise going on when they’re capturing audio. Using Adobe Sensei, sounds muffled by unwanted noises can be isolated, cleaned up, and automatically tweaked so they sound as if they were captured in a professional recording studio – when in reality your voice-overs were recorded in the bathroom with the extractor fan and shower running.
Featured image: Adobe (YouTube)