Let’s be honest; resizing is a pain. In the world of visual content, it’s like a recurring bad dream. Different social media platforms – Facebook, Instagram, Twitter, and the newcomer Threads – all have their own rules for image sizes. And it doesn’t stop there, regular websites, with their numerous display sizes and diverse pages, add to the complexity. A recent study reveals that 76% of graphic designers dedicate at least 20 hours each week just to resizing graphics.
That’s where Frameright steps in. With its Image Display Control, it aims to streamline this process. By combining the power of AI and smart coding, images are coded to understand how they should be cropped depending on their display environment. We had a chat with co-founder Marina Ekroos to delve deeper into how this works:
A little bit about you: What is your background?
Before founding Frameright, I worked as a photographer. I earned my master’s in Visual Journalism and, as part of my final thesis, I identified the problem we are now addressing with the Image Display Control (IDC). My co-founder, Ilkka, has a tech background, and he had experienced the same issue from a technology point of view. After several frustrated conversations, we finally came up with the solution, and at that point we dropped everything else we were working on and dedicated our lives to Frameright.
Prior to becoming a photographer, I studied business. So, jumping into the world of startups made sense in a way, and now I see it as my own way of being creative. Instead of expressing my creativity through photos, I am building a product, a team, a company, and hopefully, a better visual world.
What is Frameright, and what does it do?
At Frameright, we build Image Display Control technology. IDC allows everyone to control how images behave in different digital environments.
With the browser-based Frameright editor, anyone can define various regions and shapes for an image. This information is embedded inside the image file’s metadata, so the image knows how it should be cropped to different sizes and ratios, no matter where it is used. The editing process is guided by our own Frameright AI, which understands image composition and dynamics.
We also provide a collection of APIs and open-source libraries to make it easy and fast to use the IDC instructions in all types of environments.
How does Frameright ensure compatibility with various web channels and screen sizes to maintain image quality?
A significant benefit of IDC is the ability to keep the original file intact as far as possible through an image’s publishing process and life cycle instead of creating multiple image versions by hard cropping and resizing.
The IDC enriched image file contains crop instructions for the use cases known at production time. As new shape and size requirements come up, the original file will have enough pixel data for the new use cases. It contains a wealth of data from which to create versions and has cues on future needs, so new IDC instructions can always be added.
IDC is built on the open IPTC image metadata standards and works with all normal image files, like JPEGs, PNGs, GIFs and WebPs. Because the files are just regular files, it’s also fail-safe. If the channel doesn’t support IDC yet, the image behaves just like a file without any IDC metadata would.
Frameright also provides extensive presets for the most common channels like Facebook, Twitter, Open Graph images, etc. And users can, of course, create their own templates that exactly match their own channels.
Who is Frameright for? Photographers, agencies, brands, publishers?
We think everyone can benefit from Frameright. The browser-based Frameright editor gives photographers, image editors, content creators and cross-channel marketers finally a way to ensure that their material is used and shown as intended, no matter where and when the image is used. Our open-source components and developer tools free web developers from adding hacks and manual corrections for images that don’t show up correctly. Frameright also frees designers to create more creative and engaging formats, not limited by fixed image size and placement.
Our backgrounds are in photography and development, and correct image display always was an issue that we felt had to be solved by the “other side.” With Image Display Control, we want to bring these sides together and give everyone the appropriate tools to solve it.
Where IDC can bring tangible benefits and savings are companies that use a lot of image material. So far, we have mainly worked with companies in the media industry and journalism, where engagement and correctness of information are critical. However, the actual market is much bigger, from brands and e-commerce to any company that uses image assets in their communication. Businesses that understand that taking care of their image assets is crucial for their products’ success and sales.
As new channels, devices and platforms emerge, Image Display Control is required everywhere, from journalism to e-commerce and from branding to corporate communication. Eventually, our goal is to have Frameright running behind every image published digitally, even the everyday snapshots in your family’s WhatsApp group. Every image deserves to be seen correctly.
Can you provide examples of real-world scenarios where Frameright’s technology has significantly improved image display and user experience for businesses?
One of our first customers is a large media company with high-end online publications. With them, I was able to compare and analyze their old image publishing workflow and compare it to Frameright. They had been creating six different versions of their images to ensure correct image display across different channels, devices and image placement.
Using our technology, the process became 70% faster. In addition to the speed, other issues with images disappeared as well. Now their image editors have time to concentrate on the actual task they love doing.
I especially enjoy discussing this case because it was their tech team who wanted to give Frameright a try. Although we were still in the early stages, they put in some extra work to implement support for IDC. Now Frameright has been part of their daily image processes for years, and they say it has transformed their work. Daily discussions with image editors have vanished, and no one is complaining about images anymore. They can now focus on building new instead of fixing old.
Does Frameright offer any customization options or AI-driven recommendations based on user preferences and data analysis?
Frameright AI is our own in-house built machine learning model. It can understand image composition and dynamics and powers the automatic crop recommendations in the Frameright editor and our APIs. We have been fortunate to have had the opportunity to collaborate with a large image publisher, who gave us access to a dataset of over 1 million photos, cropped to different sizes by their professional image editors. This data was the starting point for our unique approach.
We also offer enterprise customers the option to join our AI program, where we extend the base model with customer-specific data. The program allows the model to learn the relevant styles and contents of the customer. It also improves trust and transparency, as it’s possible to review and audit the customer-specific training data to understand the results and mistakes better. In machine learning, it’s often too easy for implicit and explicit biases to creep in, which is unacceptable and carries significant reputational risk.
While we are very happy with the performance of our AI, images are very dependent on context. They also carry a lot of power in shaping how we perceive the real world. That is why we often emphasize our “augmented intelligence” approach and encourage human involvement in the process. With the Frameright tools, verifying the results and correcting any mistakes is fast and easy. Images are too important to be left just to the machines.
Any insight on the end-user experience? Is the image richer, faster to load, and more appealing? In other words, as a viewer of images processed by Frameright, what are the advantages?
The most evident advantage is that users see images in the way they were intended to be seen. I’m sure we’re all familiar and fed up with seeing pictures of people with their heads cropped out by some automated process or texts that are illegible. Image Display Control solves this.
The lack of IDC has also transformed how a large share of photography is done these days: a small subject in the middle with ample space around to be cut away automatically when trying to fit the image in differently sized containers. As everything has started to look the same, we’ve lost much of the storytelling power. We hope that IDC can bring back some of the power that images used to carry in capturing readers’ eyes.
A large part of the IDC magic happens already before the image reaches the end-user. However, it integrates well with all types of end-user optimizations, like the use of CDNs and newer file formats. In addition to experiencing images correctly and more appealingly, it also frees developers and designers to come up with new, creative ways of content storytelling for end-users to enjoy.
Does Frameright offer any specific features or tools tailored to enhance the visual representation of products on e-commerce websites?
The core benefit for e-commerce is the same as for all other users: Frameright and Image Display Control ensure that image material, including product images, shows up in a controlled, intended way no matter what size or ratio the image is shown in. It’s dizzying to think how many images we actually see every day. Your image must have the best possible presentation for us to remember it tomorrow.
IDC is vital for brands and manufacturers whose image material will travel and be used across a magnitude of channels: for sales, advertising and product info. The majority of these channels will be outside the control of the producer. However, as the image display instructions travel inside the file, they will be available wherever the image is displayed. Being able to stand out from the digital shelf is a struggle, and correct image display is one of the main ways to win that battle.
Frameright can also help with the production of product images, making the often repetitive task faster and more enjoyable. As a side effect, we recently realized that the IDC metadata can be used to add additional information into specific regions of the image, e.g., what the products are and where they are located in the picture. This data can then be used to zoom in on specific parts of the picture, create thumbnails or convey more information about the products. And anyone with the image file can use this data. We’re looking to add these features into our open-source libraries soon, but as Image Display Control is based on the open IPTC metadata format, retailers can easily build their own implementations as well.
Does Frameright provide any analytics or insights related to the performance of optimized images?
We cannot provide any direct analytics for the performance of the images, but we encourage our customers to track the performance on their side. In fact, the whole philosophy of IDC is to decentralize the image display process as the file travels across different systems. We don’t have direct ways to track the performance as the publishing platform doesn’t need to query us for any data when the image is online: all the needed information is already inside the file’s embedded metadata fields.
Of course, we are happy to get all types of qualitative and quantitative feedback from our customers, be it performance data or user experience. This information is used to improve our products further and is especially useful in improving the performance of Frameright AI. As styles and tastes change, this helps us keep up with current trends. And for the custom AI models, we can use performance data from the customers as a parameter for training a model that matches their goals and needs.
Moonshot question: What would you like to see Frameright do that is not yet possible because of technology constraints?
There’s still a lot we cannot yet do, but we’re getting there. I still have some reservations about machine learning and AI truly understanding visual creativity.
As a photographer, the question about dynamics and composition is fascinating. I mean, how to know what really affects the way humans experience an image? There is also constant change. What is cool today is cliché tomorrow.
With our development, we’ve made a lot of progress in better understanding style and content. However, one thing that bothers me is the small distractions in an image, such as random objects on the edges of the image that start grabbing our attention. Also in the same space is the art of cropping, for example, limbs in the correct spot. How can a machine learn to ensure something does not become a distraction?
And how do we go from understanding a single image to understanding a series of different but related images? Recently there’s been a lot of progress for temporality in generative AI, but understanding how different images relate to each other is even more difficult. They might have a completely different setting, content, and mood, but they are connected from a story point of view. Getting a machine to understand that connection is a real challenge.
Author: Paul Melcher
Paul Melcher is a highly influential and visionary leader in visual tech, with 20+ years of experience in licensing, tech innovation, and entrepreneurship. He is the Managing Director of MelcherSystem and has held executive roles at Corbis, Stipple, and more. Melcher received a Digital Media Licensing Association Award and is a board member of Plus Coalition, Clippn, and Anthology, and has been named among the “100 most influential individuals in American photography”