AI in Pursuit of the Perfect Image
AI can transform a static image into a living, breathing, infinitely flexible creative element.
Image source: ryzhi / Adobe Stock.
by Shambhavi Kadam
posted on 02-06-2019
Designers and creative pros will spend hours in pursuit of the “perfect” image. That’s because the right image can be the difference between winning or losing a client pitch. We rely on the perfect image to capture the emotional essence of a campaign, to engage an audience, and to represent a brand.
Time is the main sacrifice we make in pursuit of this perfection. A photography director might wait an entire day for light and composition to converge in a single moment. An editor might lose several afternoons tweaking curves to capture the perfect color and emotion of a shot. Or a creative director might have an entire team scouring stock sites for days, gathering the perfect campaign elements.
So, capturing, creating, or finding the perfect image can feel a little bit like waiting for creative lightning to strike. But what if instead of hoping for the perfect image to emerge, it was possible to make it emerge by leveraging artificial intelligence (AI)?
AI is poised to make this possible, and is already changing the way images are captured, edited, stored, managed, and applied — transforming the very essence of the static image itself into a living, breathing, infinitely flexible creative element.
Creative pros have a unique take on what makes for a great photo because they need to match a style guide or fit with other elements in a multichannel campaign. The near-infinite qualities of subject, framing, color, light, contrast, and more all need to converge in just the right way. That begins with image capture.
It’s now technically possible for AI to recognize the subject of a photo as it is captured and recommend filters or other camera settings for capturing a more stunning image. We’re already seeing AI-powered applications like this make their way to smartphones and DSLRs, assisting photographers with helpful recommendations and suggestions at the point of image capture.
Today, AI in cameras is focused on creating aesthetically pleasing images — how to take a great sunset photo, for example. Still, it’s not hard to imagine a future where AI assistants might analyze a brand’s imagery or style guide and make real-time adjustments to camera settings to achieve more brand-compatible photography.
Of course, AI is more easily applied to images beyond the point of capture during the editing process. Software today can recognize the subject of a photograph, automatically apply filters and adjust settings, or assist in more creative tasks like automatically selecting and masking subjects or combining elements of two or more images together.
The possibilities for AI-powered creative exploration and image repurposing are just getting started. It can be used to create consistency across many different visual assets — a time-saver when designing for a brand campaign. AI will also unlock new forms of experimentation, combination, and creative serendipity. Just imagine the untapped potential for bringing new life to existing brand assets and asset libraries that might otherwise be discarded as stale or out of date.
Image storage and management
Whether managing a personal collection of thousands of images, an asset library for a global brand, or millions of curated images from stock collections, AI is making it possible to more consistently categorize, label, group, search for, and find powerful images.
For large content libraries like Adobe Stock, it starts with the process of moderation — selecting the best, high-quality images. Before an image can be added to the library, moderators need to check each image for aesthetic quality, as well as factors like intellectual property, defects like artifacts and noise, and more.
But no matter the size of the asset library, every image needs to be appropriately and consistently tagged so it can be discovered and retrieved during subsequent image searches. Traditionally, this has been very difficult to achieve, especially across teams in the enterprise, where each organization uses the assets for different purposes, channels, or activities.
Adobe Sensei, our AI and machine learning technology, accelerates this workflow by automatically performing object and scene recognition so that each image can be automatically and consistently tagged across a large library, allowing photographers, designers, and marketers to focus more on the creation process.
Search and discovery
The search process for image libraries has historically been text based. It’s why proper and consistent metatagging is so critical. It’s also a time-consuming, hit-and-miss way of finding images. One search query might result in thousands of possible images, which then need to be reviewed and refined to make it more relevant to the project.
Searching for content is one of the longest parts of the creative process. Creative pros tell us they spend hours looking for the right image. That’s a lot of time spent on an activity that’s not really a core part of creative value, so we’re always looking for ways we can help people get to the right content more quickly, so they can spend more time doing creative work.
Adobe Sensei has already unlocked several new capabilities for finding the perfect image on Adobe Stock and other image libraries.
Drag-and-drop visual search is one popular example — the ability to find similar images by using an image as the search query. If a picture is worth a thousand words, it’s also worth a thousand metatags — and the best way to search through millions of assets to find aesthetically similar imagery.
The ability to search for images using aesthetic filters is another Adobe Sensei powered feature. Adobe Stock allows users to adjust their search results by using sliders to find images with more/less depth of field or color vibrancy, for example. These characteristics often play an important role in brand style but are notoriously difficult to describe via text alone.
We want to allow creatives to consider how objects are related to each other, where objects are in the frame, and what’s happening in the photo, and expose all those nuances to make them accessible to the user in new and interesting ways.
The intelligent evolution of the image
The goal of all of this is to leverage the power of AI to create end-to-end experiences and workflows that help creative pros and marketers tell visual stories. It’s clear we’re just starting to scratch the surface.
As technologies like voice are added to visual search capabilities, the process of finding the perfect image will become more efficient and seamless within the creative workflow. VR and AR displays may unlock new ways of organizing and visualizing image search results. Intelligent assistants might learn from your behavior and work, and suggest images that are the best fit for your current marketing campaign.
AI stands ready to make the static image obsolete. As these technologies emerge, every image will become a living, breathing, and evolving thing — and every photo gains the potential to become the “perfect” image, custom built for any brand, campaign, or other creative work.
Topics: Creativity, Digital Transformation