A New Vision of Color.
Imagine a color engine built from measurements of real film. Where each control is chemistry. Where every pixel is shaped by light and crystals. Imagine the best color you've ever seen.
The Photographic Grading Engine.
Color.io is a tool for natural color grading, driven by an engine that models the response behaviour of analog film. Create dense and vibrant colors with a streamlined interface, and tools that treat pixels like crystals in an emulsion for painterly, film-like colors.
Controls that respond like film.
Naturally.
Every control in Color.io is driven by an engine that models the complex, non-linear light response of analog color spaces. The result is better looking images with deep, painterly colors - completely shaped to match your unique taste and vision. Whether you want to use it to emulate a specific film stock or do something else entirely is completely up to you. Just as it should be.
Complete color control. Zero LUTs needed.
Color.io gives you nuanced tools to create natural color grades that rival the best film emulation plugins and LUTs. But instead of locking you in to a look, Color.io puts you in the driver seat of subtractive color modelling, physically accurate refraction and film density emulation. Powered by best-in-class cinema color science for maximum smoothness*.
* "Smoothness" refers to the gradual and even transition between different color values in an image. Most mathematical models for color manipulation make it hard or impossible to retain color smoothness during color manipulation due to their geometric limitations, perceptual non-uniformity and overall implementation features. Color.io uses custom color models, specifically developed for creative photographic color grading, that help retain smoothness even for extreme saturation increases, hue shifts and colorizations.
Cinema Raw™
for more than 600 cameras.
Color.io allows you to develop raw photos from over 600 cameras directly in your browser. A powerful VisionLog™ RAW interpreter paired with a custom wide-gamut ACES © IDT enables cinema color management for digital photography with a logarithmically encoded base image - an ideal starting point for grading in SDR and HDR.
Images developed from camera .dng originals and 16-bit .tiff composites using VisionLog ACES IDT workflow in Color.io.
Color.io debayers and interprets raw image data in a wide-gamut color space with a logarithmic gamma called VisionLog. The resulting image contains all shadow and highlight information and is an ideal starting point for creative color grading and look development. Cinema Raw describes the non-destructive conversion from VisionLog Raw through ACES to the output device transform.
Consistent colors. Absolutely anywhere.
Native ACES © HDR compliant color management enables cutting-edge color that is consistent across cameras, on-set monitors, during the edit and finally in the grade. Share your grades with anyone as fully color managed 3D LUTs or custom generated DCTLs for Davinci Resolve Studio.
But wait, there's more.
Use most features and capabilities of Color.io for free, directly in the web app. Upgrade to the pro plan when you need more.
Single user on up to 2 devices. Learn More
In case you missed anything.
Color.io is software for color grading and raw image development that runs directly in your web browser. You can use Color.io to develop raw images and to create your own 3D LUTs and DCTLs for color grading films, games and videos. Color grading in Color.io is done with an innovative UI that controls a mathematical color transformation framework that models how analog film responds to light. It offers several advantages over traditional color processing tools like Adobe Lightroom or Davinci Resolve and extends the concept of film emulation to a fully parametric approach.
Absolutely. Many Color.io users are full time photographers. You can use Color.io either to develop and color grade images directly in your browser, or create 3D LUTs for use in Adobe Photoshop, Darktable, Affinity Photo and more. Color.io is also a complete, LOG-based RAW processor that can develop RAW files from over 600 cameras. (Yes, a web app can do that!) The unique color manipulation tools and analog color grading engine make Color.io a powerful extension to any photo editing software that doesn't come with advanced 3D color visualization, film halation and deep gamut shaping tools.
Color.io is an online raw editor and can read raw data from over 600 cameras. Here's a non-exhaustive list of the image file formats you can open and edit with Color.io:
.jpg, .png, .gif, .bmp, .tiff, .dpx, .dng, .cr2, .cr3, .arw, .nef, .raf, .dc2, .rdc, .bay, .crw, .cap, .dcs, .dcr, .drf, .eip, .erf,.fff, .iiq, .k25, .kdc, .mdc, .mef, .mos, .mrw,.nrw, .obm, .orf, .pef, .ptx, .pxn, .raw, .rwl,.rw2, .rwz, .sr2, .srf, .srw, .x3f, .3fr
Color.io is end-to-end ACES © color managed.
ACES, or the Academy Color Encoding System, is a color management system that is quickly becoming the industry standard for photographers and filmmakers. This powerful system offers a wide range of benefits, including improved color accuracy, greater flexibility, improved archival capabilities, and better workflow efficiency.
One of the key advantages of ACES is its use of a wider color gamut than traditional color management systems and the fact that it works with a wide variety of cameras, displays, and output devices.
ACES also helps with workflow efficiency, it allows for consistent color throughout the entire workflow, from capture to final delivery, which can save time and effort. All images and 3D LUTs created with Color.io are fully ACES color managed. This is especially beneficial for professional photographers and filmmakers, who can trust that their images and color grades will be compatible with the equipment and software that they use on a daily basis.
You can create color grades for video by dropping multiple still frames into a scene and exporting the color you create with Color.io as a 3D LUT into your video editor of choice. Color.io accepts DPX and TIFF, as well as most other standard image formats commonly used in post production.
Color.io Pro can also output DCTL code that can be integrated into Davinci Resolve so that you can directly work with the deep color transformations created by Color.io in custom, high-end video and film workflows.
Color.io can generate 3D LUTs for most color grading and video editing applications like Davinci Resolve, Adobe Premiere, Final Cut Pro and many others. You can also generate color grading LUTs for Unity and Unreal Engine as well as Cinema4D RedShift and other 3D tools. It is also possible to export LUTs to Adobe Photoshop, Affinity Photo and other image editing applications.
You can find all supported 3D LUT formats directly in the export panel of the app.
Color.io is a progressive web-app so it self-updates automatically. Whenever you launch the app, the latest version is fetched and installed in the background. You might sometimes receive an in-app notification for major updates or announcements from Color.io.
When you subscribe to the changelog you will also receive a weekly or monthly digest email with all updates.
A single Color.io Account can be signed in on up to 2 devices at the same time. When you're switching to a third device, just sign out of your first or second device to unlock access. You can remotely mange your device sessions in your browser with your Color.io Account.
It's not as much an obsession as it is an inspiration. Film, when treated well, looks nice because it exhibits color and structure characteristics that have a much higher degree of stable color transform complexity than what can be easily achieved with the linear input controls in most interface driven software. But that's a shortcoming of the mathematical models embedded into color processing engines, not a unique characteristic of film. It's just that film has the advantage of being a physical medium that can leverage the naturally occurring framework of chemistry to do its magic, something that digital systems just don't get for free.
And this is where Color.io comes in. Color.io is not software that emulates film stocks by statically mapping digital colors to their film representation like Dehancer, FilmConvert or VisionColor.
Color.io simulates the underlying chemistry of film and its interaction with light. You can use Color.io to emulate film stocks and produce physically accurate and natural-looking film-like colors. But you're not locked into a specific film look. What you get is a streamlined and simple-to-use interface into a color engine that models the complex non-linearities of film while ensuring very high color stability and smoothness under the hood. The result is better looking images with deep, painterly colors - completely shaped to your unique taste and vision. And whether you want to use it to emulate any specific film stock or do something else entirely is completely up to you. Just as it should be.
Magic™. No seriously.
Ok, yes, you're right. Blackmagic Design has the exclusive patent on all color magic so here it is, the full trick revealed as a multi-step methodology for turning film response behaviour into mathematical color transformation models. Ready? Let's dive in!
We'll start with the problem of deriving stable, analog-to-digital color mappings from empirical film data sets. Not a trivial problem, for several reasons. Most importantly, when dealing with a physical medium, signal-to-noise-ratio has to be evaluated before mapping the signal into a vector space. If this is not done carefully, the geometric properties of the target space will be irreversibly corrupted. For static film emulation lookup tables, this is a problem that has been solved in several ways in the past by ILM, VisionColor, Adobe and later Koji, FilmConvert, Dehancer and others. If our goal however is to derive data not for direct color processing but for gaining insight into the behavioural properties of film and deriving enhanced transformation models, we need to take things a step further.
One way to robustly improve signal-to-noise ratio and increase spatial coherence in a single process is to generate importance metrics for every sampled vector (ie a three-component rgb color) in a film data set. These importance metrics are used to calculate a weight of sample noise and 3D vector location in relation to the bounds of a static target color space. The accumulated weights are used to select for the optimal data points in the final color mapping. This results in color vectors that exhibit low noise and high spatial coherence and are highly predictive of the aesthetic characteristics of a color space when applied to real world images. Let's call this metod Spatial Coherence Reconstruction because it sounds cool and is actually a pretty good description of what the algorithm does:
So now that all of our film data samples are spatially coherent and we've aggregated and positioned each and every data point in 3D space, let's superimpose these resulting vector spaces on top of each other to find common characteristics between them, which should give us clues about commonalities between different film stocks and processing techniques. At this point we can reach for some basic data science tools and machine learning magic™.... Sorry no magic, just unsupervised k-means clustering to compute some centroids. By analyzing the centroids of different film data clusters, we can gain insights into the commonalities and patterns within our film data. For example, if two data sets have similar centroids, it could indicate that they have similar patterns or characteristics. Additionally, if a centroid falls in a specific area of the feature space, it could indicate that the data points in that cluster share certain properties or characteristics. We can experiment with interpolation weights to generate different cluster set variations and then extrapolate those data to generate completely synthetic color spaces that exhibit the transform behaviour of analog film. Wonderful, we have just invented digital color spaces that look like film stocks that never existed.
Now, if we were doing actual science and not just having fun with scientific tools and methods, we would have to validate our centroids before the reconstruction and projection steps. But since we don't really have any ground truth and no one is going to get harmed if we take some poetic liberty here, let's assume that our centroids can be trusted.
Our resulting centroids, or better yet our glorious set of Magical Constants™ can be interpreted as the universal color characteristics of film, irrespective of manufacturer, stock and even, within sane limitations, processing time and solutions. That's the idea anyways. Now let's use these magical constants as the foundational building block of an algorithmic color transformation framework, reconstructed from the synthetic gamut pathways in our projected color spaces so we can actually see if our approach was any good.
...
Turns out it wasn't.
But we're not too far off. And we learned a lot.
Let's spend the next 5 years refining our methods, porting ACES to WebGL by hand, creating a color processing engine and an image renderer for the web, create a proprietary user interface framework (because Facebook and React can suck it!1!), then use the fifth iteration of that framework to create user interface components that talk to the processing engine through ACES and then finally arrange a few of those components on an HTML5 web page that is now known as:
Color.io. An unregistered trademark.
Someone's been around a while! Back in 2016 VisionColor announced a product called Spektrum and later Sekoia, neither of which would ever see the light of day.
Not because it was a bad idea to move beyond static film emulation. Certainly not because there was no demand. Simply because it took almost 7 years of research, insight and development on multiple fronts to arrive at a solution that lives up to the way the product was initially announced in 2016:
"Imagine a color engine built entirely from measurements of real film. Where each control is chemistry. Where every pixel is shaped by light and crystals. Imagine better cinematic color."
Damn, those VisionColor guys really knew how to sell stuff! I wonder who they were and what happened to them... 😏
Thank you for sticking around and may the force be with you!
#VisionColor #Spektrum #Sekoia #NeverGiveUp #XoXo
The beautiful images you see on color.io have been shot by these incredibly talented and generous photographers, cinematographers and filmmakers:
Maksim Isotomin, Spencer Sembrat, Ryan Ancill, Zeynep Guler, Robbie Herrera, Adi Constantin, Albert Vincent Wu, Aleksandr Popov, Anaya Katlego, Bruno Guerrero, Caleb George, Calicadoo Ttdio, Devin Justesen, Elia Pellegrini, Ergi Murra, Francisco Andreotti, Georgina Postlethwaite, Ihor Malytskyi, Iri Chernookaya, Ivana Cajina, Jakob Owens, Jeremy Bishop, Jed Villejo, Kai Pilger, Karina Maslina, Kate Olfans, Kevin Müller, Mantissa, Monokee, Mark Basarab, Maxim Tolchinsky, Michele Canciello, Ming Han Low, Scott Evans, Simon Berger, Simon Wilkes, Stephen Ellis, Kelly Stow, Tim Trad, Lilly Rum.
Most original images have been modified in the form of de-grading and color grading to showcase how Color.io works. To see the original images and discover more from these photographers, head over to the official Color.io Unsplash collection that highlights their phenomenal work.