The Colour of Cameras

Why an ALEXA, a RED, a Sony and a Canon see the same sunset differently, and what a colourist actually does with what they record.

Colour · Cinematography · 14 min

Open a clip straight from an ARRI ALEXA and it looks washed out, like a photo printed on dishwater. Open a RED file and it looks different again, slightly greener in the mids and darker overall. Sony and Canon footage each have their own particular grey. None of this is a mistake. Every cinema camera stores light in its own **colour space**, and the colourist’s first job is to translate each one into a common language before any creative grading can start. This post walks through that translation from photon to final look, and through what really sets the big camera makers apart.

Light becomes numbers

A camera sensor cannot see colour. Each photosite just counts photons. To get colour, nearly every cinema camera places a **Bayer colour filter array** on top: a mosaic with two green filters for every red and blue one, because our eyes are most sensitive to green. The exact dyes in those filters decide how each camera separates colours. ARRI, RED, Sony and Canon all tune their dyes differently, trading colour separation against light sensitivity. That is why two cameras pointed at the same orange jacket produce slightly different raw numbers before any software is involved. Turning that mosaic back into full-colour pixels is called **demosaicing** (or debayering), and the maker’s maths for doing it is part of its colour science.

What a colour space is

A colour space is a recipe with three ingredients. The first is a **gamut**: three primaries, red, green and blue, plotted as a triangle on the CIE chromaticity chart above. Any colour inside the triangle can be made by mixing the three. The second is a **white point**, which defines what “neutral” means; film and TV mostly use D65, roughly midday daylight. The third is a **transfer function** (the “gamma” or “log curve”), which decides how brightness values are spread across the available code values. When people say “shoot in S-Log3 / S-Gamut3.Cine”, they are naming a transfer function and a gamut. Display spaces such as Rec.709 are small because screens are limited. Camera spaces are huge, often bigger than human vision, because the camera wants to hold on to everything until the colourist decides what to keep.

The point of log

Light works linearly: each stop brighter is twice as much light. If a camera stored that directly, the brightest stop alone would eat half of all available code values, while the shadows would be starved into a few coarse steps that band and posterise as soon as you lift them. A **logarithmic curve** fixes this by giving every stop of exposure roughly the same number of code values. The price is that the image looks flat on a normal monitor, because log was never meant to be watched directly. It is a storage format, much like a scanned film negative. Kodak’s Cineon curve did exactly this for film scans in the 1990s, and every modern camera log curve is a descendant of that idea.

ARRI ALEXA

ARRI’s reputation rests on two things: **highlights** and **skin**. From the original ALEXA in 2010 through the ALEXA Mini and Mini LF, every camera recorded **LogC3** in **ARRI Wide Gamut 3 (AWG3)**. The sensor holds over 14 stops of latitude and the bright end of the range fades into white smoothly instead of hitting a hard ceiling. A blown-out window on an ALEXA tends to look like film. On many other cameras of that era it looked like video. ARRI also kept the colour science largely the same across a decade of cameras, so a Mini can be intercut with a large-format LF with little fuss. That consistency is a big reason ALEXAs dominated feature films and high-end TV.

In 2022 the **ALEXA 35** broke from that tradition. Its new Super 35 sensor claims 17 stops of dynamic range, too much to fit inside LogC3, so ARRI introduced **REVEAL Color Science** with a new curve, **LogC4**, and a new gamut, **AWG4**. LogC4 places middle grey much lower in the signal, which leaves room for highlights. It also uses one curve for every exposure index, whereas LogC3 subtly changed shape at each EI. For colourists this means ALEXA 35 footage needs its own transform and will not drop onto an old LogC3 grade without adjustment.

RED

RED’s approach has always been to record **raw**. Its R3D files hold compressed sensor data, not a baked picture, so white balance, ISO and even the colour science itself are applied later, in software. When the **RED Dragon** sensor arrived in 2013 (Epic Dragon, then Scarlet and Weapon Dragon), footage was decoded through RED’s own **DRAGONcolor** and **REDgamma** curves. Early Dragon footage had a reputation for red-shifted, sometimes magenta skin and fussy highlights, and a lot of colourists learned workarounds. Because the files were raw, every later improvement in RED’s decoding could be applied retroactively to old shoots.

That is exactly what happened in 2017 with **IPP2**, RED’s Image Processing Pipeline 2. IPP2 decodes every R3D, Dragon footage included, into **REDWideGamutRGB** with the **Log3G10** curve, which puts 18% grey neatly at one third of the signal. It also separated the creative “look” from the technical conversion and introduced smoother highlight roll-off and better-behaved saturated colours. Many colourists consider IPP2 the moment RED footage became easy to match with ALEXA. The lesson is that with raw, a camera’s colour science is software and can keep improving long after the shoot.

Sony

Sony’s cinema line, from the **FX3** and **FX6** up to the **VENICE 2**, records **S-Log3**. S-Log3 is deliberately close to Kodak’s Cineon curve, so colourists with film-scan habits feel at home, and it puts 18% grey at about 41% signal. Sony offers two gamuts. **S-Gamut3** is the widest, while **S-Gamut3.Cine** is narrower and closer to DCI-P3, which makes it easier to grade toward a cinema deliverable. Both share the same curve. Sony’s older S-Log2 and early S-Gamut footage earned a reputation for greenish skin and was hard to grade. S-Log3 with S-Gamut3.Cine largely fixed this, and the VENICE in particular is now regarded as a true peer to ALEXA for skin tones.

Sony’s other signature is **dual base ISO**. The sensor has two native analogue gain settings: 800 and 3200 on the VENICE 2, and 800 and 12,800 in S-Log3 on the FX3 and FX6. Switching to the high base changes the circuit itself rather than digitally brightening the image. Cinematographers can therefore shoot under streetlights or candlelight with surprisingly little noise. For a colourist this matters because cleaner shadows can be lifted further before grain and colour noise appear, and noisy shadows tend to drift towards magenta or green when pushed.

Canon

Canon builds its look around **Cinema Gamut** and a family of log curves. The original **Canon Log** (C300, 2011) was easy to grade but limited in range. **Canon Log 2** gave the widest latitude, and **Canon Log 3** is a middle ground that keeps most of the range while staying friendlier in the shadows. Canon footage has long had a reputation for pleasing, slightly warm **skin tones** straight out of the camera, partly because Canon’s video heritage tunes colour towards flattering faces rather than strict measurement. Documentary and corporate crews love it for that reason, since less grading is needed to make people look good.

Canon’s most interesting recent idea is the **Dual Gain Output (DGO)** sensor in the **C300 Mark III** and **C70**. It is a different solution to the same problem Sony tackles with dual base ISO. Instead of choosing one gain, DGO reads every pixel through two amplifiers at the same time, one tuned for shadows and one for highlights, and blends them. Canon claims more than 16 stops with very clean shadows. The trade-off is that DGO only runs at certain frame rates and in specific recording modes. It shows that colour science covers more than colour, since noise and latitude also limit how far a grade can go.

Comparing the colour spaces

Put the camera gamuts side by side and the first surprise is that they are **all bigger than any display**, and several reach past the edge of human vision. Primaries outside the horseshoe are “imaginary”. No light could ever be that colour, but placing the corners there means every real colour fits inside without clipping. The shapes differ because each maker positions its primaries around its own sensor’s filter response. A conversion matrix designed for one camera’s dyes would twist hues on another’s. This is why you cannot simply apply an “ARRI LUT” to Sony footage and expect it to look like an ALEXA.

The log curves differ just as much. Plotted together you can see each maker’s priorities. **S-Log3** and **LogC3** place middle grey high, around 40%, which gives generous shadow detail. **LogC4** and **Canon Log 3** sit lower to protect highlights. **Log3G10** puts grey at exactly one third. Because the curves are so different, the same correction can have very different effects. Lifting shadows by the same amount on LogC3 and on Log3G10 pushes different exposure stops. This is the strongest argument for converting everything to one common working space before you touch a single wheel.

The numbers make the differences concrete. Middle grey ranges from about 28% (LogC4) to 41% (S-Log3), so an exposure that looks “correct” on a waveform in one system would be more than a stop wrong in another. Claimed dynamic range now clusters between 15 and 17 stops for flagship cameras. Independent tests usually measure a stop or two less than the brochure and rarely agree with each other. In practice, the honest differences between modern flagships are less about raw latitude and more about **how** each camera behaves near its limits: the shape of highlight roll-off, the colour of the noise, and how skin and saturated lights are rendered.

Finding a common language

Modern grading workflows solve the mixed-camera problem with **colour management**. Each clip passes through an **input transform** that understands its specific gamut and curve, converting it into a single scene-referred working space. The industry standard is **ACES** from the Academy, whose 2.0 release in 2025 brought new output transforms. DaVinci Resolve’s own **DaVinci Wide Gamut / DaVinci Intermediate** is just as common. Once everything shares one space, a grade built on ALEXA footage behaves predictably on Sony or RED shots. At the end an **output transform** renders the result for each delivery format. The same master can produce a Rec.709 TV version, a P3 cinema DCP and a PQ HDR stream.

With the technical translation done, the creative grade begins, and it usually follows a fixed order. First come **primary corrections**: balancing exposure and white balance so shots match, using Lift, Gamma and Gain (or the more log-friendly Offset, Contrast and Pivot). Next are **secondary corrections**, which isolate parts of the image with keys and windows: a face, a sky, a neon sign. Last comes the **look**, often a film print emulation or a carefully built LUT applied on a late node. Colourists watch **scopes** more than the picture, the waveform for brightness and the RGB parade for balance, because their eyes adapt within minutes while scopes do not.

The look is where camera colour science and artistic taste finally meet. The famous **teal-and-orange** grade exists because skin sits along a narrow band of warm hues, the **skin tone line** on a vectorscope, and pushing backgrounds toward the opposite hue makes faces pop. This is where camera choice shows up again. A camera that separates subtle warm hues well, as ALEXA and Canon are known to, gives the colourist more distinct skin shades to work with. A camera with noisier or less separated colour forces broader, cruder adjustments. Good colour science does not grade the picture for you. It gives you more clean, distinct information to shape.

The short version

If you remember one thing, remember that **every camera speaks its own dialect**. ARRI records a wide gamut with a gentle curve and highlights that roll off like film. RED records raw, so its colour science keeps improving after the shoot, as IPP2 showed. Sony offers a Cineon-style curve, a grading-friendly Cine gamut and clean high-ISO images. Canon gives flattering skin and inventive sensors such as DGO. Colour management translates them all into one language, and the colourist turns that common language into a look. The camera decides how much clean information reaches the grade. The colourist decides what it means.

Sources for the maths

ARRI, “ALEXA Log C Curve: Usage in VFX” (LogC3) and “ARRI LogC4 Logarithmic Color Space Specification”
RED Digital Cinema, “REDWideGamutRGB and Log3G10” white paper (IPP2)
Sony, “Technical Summary for S-Gamut3.Cine/S-Log3 and S-Gamut3/S-Log3”
Canon, “Canon Log Gamma Curves” and Cinema Gamut white papers
CIE 1931 2° standard observer; ITU-R BT.709, BT.2020; SMPTE ST 431-2 (DCI-P3); Academy ACES documentation

*All figures are illustrations drawn in the browser from these published specifications. They are not photographs or product screenshots.*