Chapter 3: Codecs, Bit Depth, and Color

Overview

Two shooters film the same scene, side by side: a person at a table by a window, warm afternoon light on one cheek. They use the same camera model, the same lens, the same settings you learned to lock in Chapter 2 — 4K, 24 fps, a 1/50 shutter. On the back of each camera, the two shots look identical. You could not tell them apart if your life depended on it.

Then the footage goes into the edit, and someone tries to grade it — to warm the whites, deepen the shadows, push the light toward that golden café feeling the piece is supposed to have. One clip takes the grade like a dream: the colors move where you push them, the gradients on the wall stay smooth, the skin stays believable. The other clip falls apart. The moment the colorist reaches for any real adjustment, the smooth wall behind the subject shatters into ugly stair-stepped bands of color, the shadows turn to muddy blocks, and the skin goes plasticky and strange. It cannot be graded. It cannot be saved. The look the piece needed is simply not available, and no amount of skill in post will put it back.

The two shooters saw the same image on set. They did not record the same image. The difference — invisible at capture, catastrophic in post — was the recording format: the set of choices, made silently the instant each pressed record, about how the image would be compressed and stored, how much color information would survive, and how much room the file would leave for the edit and the grade. That is this chapter's entire subject, and it is the least glamorous, most consequential technical decision you will make on any shoot.

Here is the idea to carry through everything that follows: the camera does not record what you see. It records a compressed, encoded approximation of it — and how good that approximation is, and how much you can push it later, is a choice you make at capture and can never undo. In Chapter 2 we followed light from the world onto the sensor and turned it into a grid of numbers. This chapter follows those numbers one step further — from the sensor to the file on the card — and answers the question every setting in Chapter 2 set up: once the sensor has seen the light, what does the camera actually keep?

This is the most technical chapter in the book, and we are going to be honest about that. But here is our promise, and we mean it literally: you can skip every 🔬 The Tech box in this chapter and still make every correct decision. The deep detail — how compression predicts frames, the math of chroma notation, the color science — lives in those optional boxes, walled off so a reader who never opens one still walks away able to choose a format, avoid the disasters, and set up a workflow that protects the work. Read for the practical decisions first. The theory is there when you want it, and only when you want it.

In this chapter you will learn to:

• Explain what a video file really is — the difference between a codec and a container, and why the file extension lies to you.
• Understand bitrate and compression well enough to trade quality against card space on purpose, and to recognize compression damage when you see it.
• Explain bit depth and chroma subsampling — the two numbers (like "8-bit 4:2:0") that quietly decide whether your footage survives a grade.
• Define dynamic range in stops, and understand what a Log picture profile keeps, what it costs, and why ungraded Log looks "wrong" on purpose.
• Know what RAW video is, when its enormous flexibility is worth its enormous cost, and — more often — when it isn't.
• Choose a capture format for a real project and build a card-and-offload workflow that never loses footage.

Learning Paths

This chapter serves every reader, but it is the one where the paths diverge most. The deepest codec and color math matters intensely to some readers and not at all to others. Read for your goal:

• 📱 Phone-first: you can skim the deepest codec math (the 🔬 The Tech boxes, the interframe/GOP detail, the chroma notation) with a clear conscience — your phone makes most of these choices for you. What you must take from this chapter is smaller and more useful: shoot the highest-quality setting your phone offers, get white balance right in camera, decide honestly whether you'll grade before you turn on any "Log" or "flat" mode, and back up your footage before you delete anything. §3.2 and §3.6 are your sections.
• 🎥 Creator: §3.3 (bit depth and chroma) and §3.4 (Log) are your money sections — they decide whether your footage holds up to the grade that makes work look "cinematic." Learn where 8-bit 4:2:0 breaks.
• 💼 Pro-track: all of it, cold. You will be handed unfamiliar cameras and asked "what should we shoot this in?" and the answer depends on the whole pipeline — grade, VFX, delivery, storage, deadline. The 🔬 The Tech boxes are worth your time eventually.
• 🎓 Student: read straight through and do the banding test in §3.3 and the Production Checkpoint. This is the vocabulary Chapters 5, 31, 32, and 36 assume you own.

3.1 Codec vs container: what the file actually is

Before we can talk about quality, we have to answer a question so basic that almost nobody asks it: when you press record and a file appears on your card called C0012.MP4, what is that thing? What is a video file made of?

The answer has two parts, and confusing them is the single most common source of "why won't this file work?" misery in all of video. The two parts are the codec and the container.

A codec is the method — the compression scheme — the camera uses to squeeze the raw torrent of image data into a manageable size when it records, and that your computer uses to unsqueeze it when it plays back. The word is a smashing-together of coder–decoder (or compressor–decompressor), and that tells you its whole job: on the way in, it codes (compresses) the image; on the way out, it decodes (reconstructs) it. Without a codec, one second of raw 4K video would be a firehose of data far too large to write to any consumer card in real time. The codec is what makes recording video onto a small card physically possible, and the choice of codec is one of the biggest levers over quality and editability you have. Codecs you will hear named: H.264 (also called AVC), the workhorse of consumer cameras, phones, and the web; H.265 (also called HEVC), its more efficient successor; Apple ProRes and Avid DNxHR, high-quality "editing" codecs; and newer ones like AV1. Do not memorize the list. Understand that a codec is how the picture was compressed.

A container is the file itself — the wrapper that holds the compressed video, the audio, and a pile of metadata (timecode, frame rate, date, camera notes) together in one package your computer can recognize. The container is what the file extension names: .MP4, .MOV, .MXF, .MKV, .AVI. Think of the container as a shipping box and the codec as the way the contents were packed inside it. The box has a label on the outside (the extension). But the label tells you about the box, not about how the thing inside was packed.

And here is the trap, the thing that trips up beginners and even working editors: the container is not the codec, and the extension does not reliably tell you the codec inside. A .MOV file might contain H.264, or ProRes, or several other codecs. A .MP4 usually contains H.264 or H.265 — but "usually" is not "always." Two files can both end in .MOV and be completely different animals: one a tiny, hard-to-edit, highly compressed H.264 clip, the other a huge, buttery-smooth ProRes clip. Same box. Utterly different contents. When someone says "the file won't import" or "it plays with no picture," the culprit is almost always a codec the software can't read — a contents problem — even though everyone keeps talking about the box.

FIGURE 3.1 — Codec vs container: the box and how it's packed

   THE CONTAINER (the file / the box)            .MOV   .MP4   .MXF   .MKV   .AVI
   ┌───────────────────────────────────────────┐   ↑ the extension names the BOX
   │  ┌─────────────────┐  ┌──────────────────┐ │
   │  │  VIDEO stream    │  │  AUDIO stream    │ │   Inside, the VIDEO was packed by a
   │  │  packed by a     │  │  packed by an    │ │   CODEC (H.264? H.265? ProRes?) —
   │  │  video CODEC     │  │  audio codec     │ │   and the box's label does NOT
   │  │  (H.264/H.265/   │  │  (AAC / PCM…)    │ │   reliably tell you which.
   │  │   ProRes/DNxHR)  │  │                  │ │
   │  └─────────────────┘  └──────────────────┘ │   Same .MOV box can hold a tiny
   │  ┌───────────────────────────────────────┐ │   H.264 clip OR a huge ProRes clip.
   │  │  METADATA (timecode, fps, date, notes) │ │   "Won't import" = a CODEC problem,
   │  └───────────────────────────────────────┘ │   almost never a container problem.
   └───────────────────────────────────────────┘

Why does this matter to you, practically, today? Two reasons. First, when a file misbehaves, you now know to ask "what codec is inside?" rather than staring at the extension — and the fix is usually to transcode it (re-wrap or re-compress it into a codec your software likes) or to install the right decoder. Second, and bigger: the codec you choose on your camera, when it offers a choice, is a decision about the entire life of that footage — how big the files are, how hard they are to edit, and how much quality survives. That choice splits into two families, and understanding the split is most of what you need.

Delivery codecs (H.264, H.265) are built to make files small. They are astonishingly good at it, which is why they run the whole consumer world — your phone, your camera's default mode, every video on the web. The price of that smallness is that they are hard to edit: to shrink files so far, they use a trick (we will open it up in the Tech box below) that makes any single frame expensive for your computer to reconstruct, so scrubbing and cutting can feel sluggish, especially in 4K on a modest laptop.

Editing codecs (ProRes, DNxHR) make the opposite trade. Each frame is compressed on its own, lightly, so the files are large but your computer can jump to any frame instantly and cut without breaking a sweat. These are the codecs professional edits often run on — sometimes captured in-camera on higher-end bodies, more often created in post as high-quality intermediates. You will meet them properly when we talk about proxies and media management in Chapter 27 and delivery in Chapter 36. For now, just hold the shape of the trade: small-and-hard-to-edit versus large-and-easy-to-edit. There is no "best" — there is only "right for this stage of this job."

🔬 The Tech: why delivery codecs are hard to edit (interframe compression). Here is the trick that makes H.264 files so small. A naïve codec would compress each frame on its own — a series of independent stills. That is called intraframe ("within a frame") compression, and it is what ProRes and DNxHR do. Delivery codecs go further with interframe ("between frames") compression, also called long-GOP (Group of Pictures). The insight: in most video, one frame looks almost exactly like the last one — only a little changed. So instead of storing every frame in full, the codec stores an occasional complete frame (a keyframe or I-frame) and then, for the frames in between, stores only the differences — where things moved, what changed — as predicted frames (P- and B-frames) that reference their neighbors. A whole second of video might contain just one or two full frames and dozens of "here's what changed" instructions. That is how a firehose becomes a trickle. The cost lands in the edit: to show you any single in-between frame, the computer must find the nearest keyframe and rebuild every frame from there to the one you want — real work, done constantly as you scrub. Intraframe codecs skip all that (every frame is already complete), which is exactly why they cut so smoothly and cost so much storage. You can skip this box forever; the practical takeaway lives above it. But now you know why your 4K H.264 clips stutter on the timeline and why editors transcode or use proxies.

⚠️ Common Mistake: blaming the box when the codec is the problem. A file won't import, or imports as sound-with-no-picture, and the beginner concludes the file is "corrupt" or the software is broken. Nine times out of ten the file is fine and the codec inside is one the software can't decode (a common H.265 variant, or an unusual flavor from a particular camera). The fix is not to panic — it's to transcode the file into a friendly codec (many free tools do this, and DaVinci Resolve, this book's free reference editor, will often ingest and convert it for you) or to install the decoder. Learn to ask "what codec?" before "what's wrong with my computer?"

🔄 Check Your Eye. 1. In one sentence each, what is a codec and what is a container? 2. A friend sends you a .MOV and says "it's ProRes." Your other friend's .MOV is H.264. How can two .MOV files be so different? 3. Why are delivery codecs (H.264) small but hard to edit, while editing codecs (ProRes) are large but smooth?

Check yourself

1. A codec is the compression method used to encode the image on record and decode it on playback (H.264, ProRes). A container is the file wrapper that holds the video, audio, and metadata together, named by the extension (.MOV, .MP4).
2. .MOV is the container (the box); it can hold many different codecs. One box was packed with ProRes (large, easy to edit), the other with H.264 (small, hard to edit) — same extension, different contents.
3. Delivery codecs use interframe/long-GOP compression (storing mostly the differences between frames), so files are tiny but any single frame must be rebuilt from neighbors — hard work in the edit. Editing codecs compress each frame independently (intraframe), so files are large but every frame is instantly available.

3.2 Bitrate and compression: quality vs card space

We have said codecs "compress" the image and that compression is a trade. Now let us put a number on the trade, because that number — the one you can actually see and set on many cameras — is bitrate.

Bitrate is the amount of data the camera records for each second of video, measured in megabits per second (Mb/s). It is the throttle on the firehose: a higher bitrate means the camera keeps more data per second, which generally means a higher-quality image and larger files; a lower bitrate means it throws more away, giving smaller files and, past a point, visible damage. If the codec is how the image is compressed, the bitrate is how hard. A clip at 400 Mb/s and a clip at 40 Mb/s can use the identical codec and resolution and look worlds apart, because the first kept ten times as much information about each second of the world.

Compression, remember, means throwing data away — and video compression is almost always lossy, meaning the discarded detail is gone for good, not merely folded up to be restored later. A good codec at a healthy bitrate throws away only what your eye was never going to miss. Starve the bitrate, or hand the codec a scene it finds hard, and it starts throwing away things you will miss. What does discarded detail look like? It has names, and once you can name them you will see them everywhere:

• Macroblocking — the image dissolves into visible blocky squares, worst in shadows and fast motion, as the codec runs out of bits and gives up on detail within each block.
• Banding — a smooth gradient (a clear sky, a lit wall, a fade to black) breaks into distinct stepped bands instead of a seamless transition. (Bitrate is one cause; bit depth, in §3.3, is the other — hold that thought.)
• Smearing / mosquito noise — fine, busy, moving detail (grass blowing, water, rain, confetti, foliage) turns mushy or shimmers with a haze of artifacts, because busy motion is the hardest thing for an interframe codec to predict.

The through-line: compression damage is worst exactly where a scene is hardest to compress — deep shadows, smooth gradients, and fine chaotic motion. A locked-off shot of a evenly lit wall compresses beautifully at a low bitrate. A handheld shot of wind in a field of grass at dusk will fall apart at the same bitrate. This is why "what bitrate do I need?" has no single answer — it depends on how hard your footage is to compress and how much you intend to push it later.

FIGURE 3.2 — Bitrate: the quality vs card-space (and card-speed) trade

  LOW bitrate  ┤████░░░░░░░░░░░░░░  small files, long record times, cheap cards
   (e.g. tens  │    ↑ but: banding in skies, blocky shadows, smeared motion —
    of Mb/s)   │      damage you CANNOT remove in post
               │
  HIGH bitrate ┤████████████████░░  big files, shorter record times, FAST cards needed
   (hundreds   │    ↑ clean gradients, solid shadows, detail survives the grade
    of Mb/s)   │
               └────────────────────────────────────────────────────────────────
                 The rule: for anything you'll GRADE or that has skies, shadows,
                 or busy motion, shoot the HIGHEST bitrate you can afford to store
                 and your card can keep up with. Bitrate is the cheapest quality
                 you will ever buy — and the one you can never buy back later.

There is a second, sneakier cost hiding in bitrate, and it bites beginners hard: your card has to be able to keep up. A high bitrate is a lot of data arriving every second, and if your memory card cannot write that fast, the camera will stutter, drop frames, or simply stop recording — sometimes mid-take, sometimes without warning. This is where a genuinely confusing units problem lives, so let us kill it now. Bitrate is quoted in megabits per second (Mb/s, lowercase "b"). Card write speeds are usually quoted in megabytes per second (MB/s, uppercase "B"). There are 8 bits in a byte, so a 200 Mb/s bitrate is 25 MB/s of actual writing — and you want a card rated comfortably above that, not exactly at it. Cards carry ratings for this: look for the Video Speed Class marks (V30, V60, V90 — the number is the guaranteed minimum MB/s) or the UHS Speed Class (U1, U3). Match the card to the bitrate, with headroom, and the camera stops surprising you.

🎒 Gear Note: cards are part of your image quality, not an afterthought. Beginners spend on cameras and lenses and buy the cheapest card that fits. This is backwards. A card too slow for your chosen bitrate will corrupt or refuse a take at the worst possible moment, and a fake or failing card (the counterfeit market is real) will eat footage you can never re-shoot. The principle, gear-agnostic: buy cards from reputable sellers, rated with headroom above your camera's top bitrate, and retire them before they're old. For a phone shooter, the equivalent is free space: high-bitrate 4K fills a phone fast, and a full phone stops recording mid-take just as surely as a slow card does. Clear space before the shoot, not during it. The card (or the free gigabytes) is the floor your whole image stands on — do not build on a cheap one.

Now, the practical decision. Most cameras that let you set bitrate offer a few tiers, sometimes labeled by quality (a "standard" and a "high" mode) or by codec (an efficient H.265 mode and a fatter H.264 or intra mode). The guidance is simple and rarely wrong: for anything you intend to color-grade, or that contains skies, shadow detail, or busy motion, shoot the highest bitrate you can store and your card can sustain. The extra file size is the cheapest insurance in this chapter. The only reasons to step down are real ones — a card that can't keep up, a storage budget you can't exceed, a very long record time (an all-day event) where you'd rather have the footage at a lower bitrate than run out of card. Those are legitimate trades, made on purpose. What you must not do is shoot a low bitrate by accident and discover the banding in the grade, because — unlike a wrong white balance or a soft focus — compression damage is baked in and permanent.

✂️ In the Edit. This is the first of several places this chapter forces the Chapter 1 rule you shoot for the edit to become concrete. Bitrate is a promise you make to your future self in the grade. Shoot a sky at a stingy bitrate and the codec bands it — steps it into visible stripes — and when the colorist tries to deepen that sky or add a gradient, the bands don't smooth out; they get worse, because grading amplifies whatever damage is already there. The editor cannot add back detail the camera never wrote. Every artifact you can name in this section is a door that closes at capture. Shoot enough bitrate and the door stays open; shoot too little and no amount of talent in post reopens it. The edit can only ever rearrange and reshape what you captured — it cannot invent.

🔄 Check Your Eye. 1. What does bitrate measure, and what happens to file size and quality as it rises? 2. Name two kinds of compression artifact and the scene conditions that trigger them. 3. Your camera shoots 200 Mb/s but your card is rated V30 (30 MB/s). Is that card fast enough? (Watch the units.)

Check yourself

1. Bitrate is the data recorded per second of video (Mb/s). Higher bitrate → larger files and generally higher quality (more detail kept); lower bitrate → smaller files and, past a point, permanent artifacts.
2. Macroblocking (blocky squares, worst in shadows/fast motion) and banding (stepped bands in smooth gradients); also smearing/mosquito noise in busy motion like grass or water. They appear where scenes are hardest to compress: shadows, gradients, and chaotic fine motion.
3. No. 200 Mb/s ÷ 8 = 25 MB/s of actual data; a V30 card guarantees only 30 MB/s, which is cutting it far too close with no headroom. You want a comfortably faster card (e.g., V60/V90) so the camera never drops a frame.

3.3 Bit depth and chroma subsampling: the two numbers that decide the grade

If you learn one section of this chapter cold, make it this one — because these two properties, expressed as a little string like "8-bit 4:2:0" or "10-bit 4:2:2," are what decide whether your footage survives being pushed around in post. They are invisible on set. They are invisible when you simply watch the footage. They become visible — sometimes ruinously — the instant you grade. And they connect straight to the warm café look this whole book is quietly building toward.

Bit depth: how many steps of tone the file can hold

Bit depth is the number of bits the file uses to record each color channel of each pixel — which sets how many distinct steps of brightness are available, per channel, between pure black and pure white. It is the resolution of tone, the way pixel count is the resolution of detail. The common depths:

• 8-bit gives 2⁸ = 256 steps per channel. That's the standard for delivery — your phone, most cameras' default, every streaming platform. 256 steps of red, green, and blue combine to about 16.7 million colors, which is plenty to display a beautiful image.
• 10-bit gives 2¹⁰ = 1,024 steps per channel — four times as many. That's over a billion possible colors, and, far more importantly, four times as many rungs on the ladder from dark to light.
• 12-bit gives 2¹² = 4,096 steps per channel, the realm of high-end and RAW capture.

Now, why does the number of steps matter, if 256 is already enough to display a lovely picture? Because grading stretches the ladder. When a colorist deepens shadows, lifts highlights, or pushes contrast, they are taking the steps the file recorded and spreading them apart. If you started with 1,024 steps, there's a fine step nearby wherever the grade lands, and gradients stay smooth. If you started with 256 and stretch hard, the steps pull apart until the gaps between them become visible — and a smooth wall, a clear sky, a soft-lit cheek breaks into banding, those stair-stepped stripes where there should be a seamless blend. Eight-bit footage is perfectly fine to watch. It is fragile to grade, especially in the smooth, subtle areas — skies, walls, skin, any gentle gradient — which are exactly the areas a nice grade lives in.

FIGURE 3.3 — Bit depth: why 8-bit bands and 10-bit doesn't (a gradient, dark → light)

  8-bit (256 steps/channel), pushed hard in the grade:
    black ▓▓▓▓▓▓▒▒▒▒▒▒░░░░░░      ← few steps; stretch them and GAPS open up →
          ███  ▒▒▒  ░░░  ▁▁▁       visible BANDS: stair-steps where the sky/wall should be smooth

  10-bit (1024 steps/channel), same push:
    black ▓▓▓▒▒▒░░░▁▁▁ ··· ░░▒▒██   ← 4× the steps; a fine rung always lands nearby →
          smooth, seamless gradient — no visible banding, even under an aggressive grade

  The image LOOKS the same on the camera. The difference only appears when you PUSH it in post.
  More bit depth = more room to grade before smooth areas break apart.

Chroma subsampling: the color detail the camera throws away on purpose

The second number is stranger, and it rests on a genuine fact about your own eyes. Chroma subsampling is the practice of recording the full brightness detail of an image but deliberately keeping only part of the color detail — because human vision is far sharper at seeing brightness (luma) than color (chroma), so the camera can discard a lot of color information and you literally cannot see that it's gone. It is one of the oldest and most effective compression tricks there is, and nearly every video you have ever watched used it.

It's written as a three-number ratio — 4:4:4, 4:2:2, 4:2:0 — and you can read the shorthand without the full theory:

• 4:4:4 — full color; nothing thrown away. Every pixel keeps its own complete color. Used for high-end work, VFX, and graphics.
• 4:2:2 — half the horizontal color detail is kept. Effectively invisible to the eye when watching, and a strong choice for grading and green screen. Common on better cameras' higher modes.
• 4:2:0 — a quarter of the color detail is kept (color is halved both horizontally and vertically). This is the standard for delivery and for most consumer capture — your phone, most cameras' default, and every streaming platform deliver 4:2:0. You do not see the missing color when you watch. You feel it when you grade heavily or pull a green-screen key.

FIGURE 3.4 — Chroma subsampling: full brightness, less color (a block of pixels)

  Every square = one pixel.  ▓ = full color kept here    · = color borrowed from a neighbor

   4:4:4 (full)         4:2:2 (half horizontal)     4:2:0 (quarter)
   ▓ ▓ ▓ ▓              ▓ · ▓ ·                     ▓ · ▓ ·
   ▓ ▓ ▓ ▓              ▓ · ▓ ·                     · · · ·
   ▓ ▓ ▓ ▓              ▓ · ▓ ·                     ▓ · ▓ ·
   ▓ ▓ ▓ ▓              ▓ · ▓ ·                     · · · ·
   every pixel its      color sampled every         color sampled every other
   own color            other column                column AND every other row

  BRIGHTNESS (luma) detail is FULL in all three — that's why 4:2:0 looks fine to watch.
  What differs is COLOR (chroma) detail. It matters for: heavy grading, green-screen keying,
  and fine colored edges (like colored text). For simply watching, you can't tell them apart.

💡 Why It Works: your eyes sold you out, and the codec knew. Chroma subsampling isn't a flaw or a corner cut to save a few pennies — it's a precise exploitation of how you see. Your retina has far more receptors for light and dark than for color, so brightness edges look sharp to you while color edges look soft. The codec keeps every scrap of the brightness detail your eye is picky about and quietly halves or quarters the color detail your eye is careless about. The result: 4:2:0 footage looks, to a watching human, essentially identical to full-color 4:4:4 — while being dramatically smaller. It's one of the most elegant tricks in imaging. It only stops being free when you ask the footage to do something a watcher never asks of it: to be pushed, stretched, and keyed in post, where that missing color detail suddenly has a job and isn't there to do it.

Reading the pair: "8-bit 4:2:0" vs "10-bit 4:2:2"

Put the two numbers together and you have the two most important quality descriptors your camera doesn't advertise on the box:

The practical reading is not "always shoot 10-bit 4:2:2." It's match the pair to what you'll do in post. A talking-head you'll correct lightly and cut? 8-bit 4:2:0 at a good bitrate is genuinely, honestly fine — do not lose sleep. A moody piece you'll grade hard, a product with a smooth gradient backdrop, or anything going in front of a green screen? Now the extra tonal steps and color detail earn their keep, and if your camera offers 10-bit — even 10-bit 4:2:0 — reach for it. The failure mode to avoid is committing to a big grade or a green-screen composite and then discovering you shot it 8-bit 4:2:0.

🎬 On Set: make banding with your own hands. You will believe bit depth forever once you've broken an image yourself. Find a smooth gradient — a clear patch of sky at dusk, a plain wall lit softly from one side, a white surface fading into shadow. Shoot ten seconds of it in the most compressed mode your camera or phone offers (lowest bitrate, standard 8-bit profile). Bring it into any free editor and push it hard: crank the contrast, deepen the shadows, add a strong color tint — grade it like you're trying to hurt it. Constraint: use a genuinely smooth subject; a busy, textured one hides the effect. Self-review: watch the smooth area shatter into bands and stair-steps. That is the ceiling of 8-bit heavily-compressed footage, seen with your own eyes. If you have a 10-bit mode, shoot the identical subject in it and push it the same way — the bands won't come, or won't come nearly as fast. Keep both clips; they're the most persuasive argument in this chapter.

♿ Accessibility & Inclusion. Bit depth and color have a quiet accessibility dimension worth planting now. First, the smooth gradients that banding destroys are often the backgrounds you'll later lay text over — titles, lower thirds, captions — and text needs a clean, high-contrast field to stay legible; banding and blocking under a caption make it harder to read for everyone, and worse for low-vision viewers. Second, when you reach grading (Chapter 32), remember that a meaningful slice of your audience is color-blind: never let color alone carry essential information (a red-vs-green distinction, a colored callout with no other cue). Capture clean, gradeable footage now, and you keep the freedom later to build looks that stay readable and inclusive rather than looks that fight your viewers.

🔄 Check Your Eye. 1. What does bit depth control, and why does 8-bit "band" under a heavy grade when 10-bit doesn't? 2. In "8-bit 4:2:0," what does the "4:2:0" part describe, and why can't you see its effect when simply watching? 3. You're about to shoot a subject against a green screen to key them out later. Which pair do you want if your camera offers it, and why?

Check yourself

1. Bit depth sets how many tonal steps per channel the file holds (8-bit = 256, 10-bit = 1,024). Grading stretches those steps apart; with only 256, the gaps become visible as banding in smooth areas, while 10-bit's 1,024 steps keep a fine rung nearby so gradients stay smooth.
2. "4:2:0" is the chroma subsampling — how much color detail is kept (here a quarter, versus full brightness detail). You can't see it while watching because human eyes are far sharper at brightness than color, so the discarded color detail is one your eye barely registers.
3. 10-bit 4:2:2 (or at least 10-bit) — keying needs clean color edges to separate subject from background, and 4:2:0's quarter-resolution color gives soft, blocky edges that key badly. More color detail (4:2:2) and more tonal steps (10-bit) make a cleaner matte. (Green screen and keying: Chapter 35.)

3.4 Dynamic range, and Log vs standard picture profiles

We now arrive at the choice the Production Checkpoint will ask you to make — standard profile or Log — and to make it well, you first need the idea underneath it: dynamic range.

Dynamic range: the span a camera can hold at once

Dynamic range is the span between the darkest and brightest parts of a scene that a camera can capture in a single exposure while still holding real detail at both ends — measured, like everything about light, in stops (recall from Chapter 2 and the exposure work ahead in Chapter 5 that one stop is a doubling or halving of light). A camera with a wide dynamic range can hold detail in a bright window and a shadowed face in the same shot. A camera with a narrow one is forced to choose: keep the face and let the window blow out to featureless white, or keep the window and let the face crush to black.

Your eye, by comparison, is astonishing — with its constant adaptation it handles a range often cited as roughly twenty stops, which is why you can look at a person against a sunny window and see both perfectly well. The camera cannot adapt shot to shot; it captures one fixed range at a time, and modern cameras land somewhere in the neighborhood of twelve to fifteen stops depending on the sensor (a figure worth treating as approximate, not gospel — makers measure it differently and marketing inflates it). Whatever the exact number, the practical truth is the one that will shape your shooting life: many real scenes contain more range than your camera can hold, so you will constantly be deciding what to keep and what to sacrifice. The classic version of this problem — a person indoors against a bright window — is the exact situation the Café Scene lives in, and we'll wrestle it directly when we light and expose that interior in Chapters 11 and 13.

FIGURE 3.5 — Dynamic range: the scene has more stops than the camera can hold

  THE SCENE (person by a bright window):   dark shadow ├────────────────────────┤ bright window
                                                        │  ~16+ stops, more than   │
                                                        │  the camera can hold     │

  CAMERA CAPTURE WINDOW (~12–15 stops):          ├──────────────────┤
    Slide it DOWN → keep the shadows,        shadows kept ┤        ├ window BLOWN white (lost)
    Slide it UP   → keep the window,             face BLACK ┤        ├ window kept
    A wider-range camera (or a Log profile)  ├────────────────────────┤  ← holds more of both ends

  You are always choosing which end to protect. The tools: light the scene to fit the camera's
  range (Part III), expose carefully (Ch.5), and record a profile that keeps as much range as
  possible (this section). You cannot recover detail the sensor never captured.

Picture profiles: standard vs Log

A picture profile is how the camera maps the light the sensor captured onto the values it records — how much contrast and color it "bakes in." You have essentially two philosophies to choose between.

A standard profile (often a Rec.709 or "standard/natural" setting) bakes in pleasing contrast and color right away. The image that lands on your card looks good immediately — it's ready to watch, ready to hand over, ready to cut with no color work at all. The cost is baked in with the beauty: to make that contrasty, saturated look, the profile clips the brightest highlights to white and crushes the deepest shadows to black at capture, discarding the range beyond its pretty middle. You get a ready image and less room to change your mind.

Log — short for logarithmic — is the opposite bargain. A Log picture profile records the sensor's full captured range as a deliberately flat, low-contrast, desaturated image, spreading the tones along a logarithmic curve so that far more of the camera's stops — especially in the highlights and shadows — survive into the file instead of being clipped away. The catch, and it is the whole point, is that ungraded Log looks wrong: grey, milky, washed-out, lifeless. That is not a defect. Log is not a look — it is a container for information, designed to be graded. The flatness is the sound of the file holding onto everything, waiting for you to decide the contrast and color later, in post, with maximum room to work.

FIGURE 3.6 — Standard profile vs Log: the same scene, two philosophies

  STANDARD profile (Rec.709):           LOG profile:
   contrast + color BAKED IN at capture   FLAT, grey, desaturated at capture
   highlights ██ clipped to white         highlights ▓▓ held (detail kept)
   shadows    ██ crushed to black         shadows    ▒▒ held (detail kept)
   ─────────────────────────────          ─────────────────────────────
   ✅ looks great straight away            ❌ looks WRONG until graded (grey/washed out)
   ✅ no grading needed                    ✅ maximum range + latitude survives to post
   ❌ less room to change the look         ✅ you decide contrast & color LATER
                                          ⚠️ REQUIRES grading; wants 10-bit or it bands

Here is the decision, stated plainly, because Log is the single most over-reached-for setting among ambitious beginners: shoot Log only if you will actually grade, and can grade it well. Log's flexibility is real and it is how the "cinematic" looks you admire are built — but that flexibility is a loan you repay in post. Three strings come attached, and ignoring any one of them turns Log from an upgrade into a downgrade:

1. Log must be graded. Deliver ungraded Log and you've delivered a flat, grey, broken-looking video. Grading is Chapters 31 and 32; if that's not in your plan or your skillset yet, a good standard profile will look better on delivery than mishandled Log.
2. Log wants bit depth. Grading Log means stretching a flat image back into contrast — the exact operation that causes banding (§3.3). In 8-bit, heavy Log grading bands badly. Log and 10-bit belong together; 8-bit Log is a real trap. If your camera only shoots 8-bit, a good standard profile is often the wiser choice.
3. Log needs careful exposure. Because Log lifts the shadows into a flat mid-grey, underexposing it drags noise up out of the shadows when you grade. Log is generally exposed a touch brighter than you'd think, to keep the shadows clean — a discipline we cover properly in Chapter 5, §5.5, "Exposing for Log and protecting highlights."

🔗 Connection: the Café Scene's warm grade is being decided right now. Later in this book — correction in Chapter 31, grading in Chapter 32 — we will take the Café Scene and build its signature look: warm whites, deep amber pooling from the practical lamps, the cool daylight from the window balanced against it, skin kept believable through all of it. That grade is one of the payoffs the whole book is building toward. And here is the sobering, motivating truth of this chapter: that grade lives or dies on the color information you decide to capture now. Warming a café, pushing amber, and protecting skin all at once is a heavy grade — exactly the kind that shatters thin footage into bands and mud. Shoot the Café Scene 8-bit 4:2:0 with a contrasty look already baked in, and Chapter 32's grade will have almost nothing to work with. Shoot it with enough bit depth and range — 10-bit, and Log or at least a flatter profile if you can grade it — and that warm café look has room to breathe. You will not touch the grade until Chapter 32. But you decide today whether it's even possible. That is what "you shoot for the edit" means, taken all the way down to the recording format.

🎬 On Set (software note): how Log becomes a look, in a free tool. Because Log needs grading, it's worth seeing the very first step once, concretely. In any modern editor, Log footage is brought back to a normal look by applying either a color space transform (telling the software "this is S-Log/C-Log/V-Log; show it as normal Rec.709") or a LUT — a look-up table, a preset that remaps the flat Log values into a contrasty, colored image (we cover LUTs properly in Chapter 32). In DaVinci Resolve — the free, professional color and edit suite this book anchors its hands-on steps to — you'd add a color-space-transform node or the camera maker's LUT, and the grey, flat image snaps into a normal-looking one you then grade from. (Every editor has its own name for these steps; see Appendix E for the translations.) The point for today isn't the button — it's the workflow reality: Log is step one of a grade, not a finished image. If that grade isn't happening, don't shoot Log.

🔬 The Tech: color spaces, gamut, and HLG (the middle path). Two related ideas sit under this section for the curious. First, color space / gamut — a picture profile also declares which range of colors the file describes. Rec.709 is the standard color space of ordinary HD video and most delivery; Rec.2020 is a much wider gamut used for HDR and high-end work. Log profiles usually carry a wide gamut so they can hold vivid colors the grade might want. Second, if standard-vs-Log feels too binary, there's a genuine middle path: HLG (Hybrid Log-Gamma), a standardized profile (part of the Rec.2100 HDR standard) that captures more range than a standard profile but, unlike Log, is designed to look acceptable even without a full grade. For a shooter who wants extra latitude but a lighter post workflow, HLG can be a smart compromise. None of this changes the practical rule above — match the profile to your grade — but it's the fuller map of the territory. Skippable, as promised.

🔄 Check Your Eye. 1. What is dynamic range, and why must a camera often "choose" between a bright window and a shadowed face? 2. Why does ungraded Log look grey and washed out — and why is that a feature, not a bug? 3. Name the three strings attached to Log. Which one makes 8-bit Log a trap?

Check yourself

1. Dynamic range is the span from darkest to brightest a camera can hold in one exposure, measured in stops. Many scenes contain more stops than the camera's ~12–15, so it must sacrifice one end — blowing the window to white to keep the face, or crushing the face to keep the window.
2. Log records the full captured range as a flat, low-contrast, desaturated curve so highlights and shadows survive into the file instead of being clipped. It looks wrong because the contrast and color haven't been applied yet — Log is an information container meant to be graded, not a finished look.
3. (a) Log must be graded (ungraded, it looks broken); (b) Log wants bit depth (stretching flat Log to contrast bands badly in 8-bit — that's why 8-bit Log is a trap); (c) Log needs careful, slightly brighter exposure to keep shadow noise down (Chapter 5).

3.5 RAW video: when it's worth it (and when it isn't)

There is one more step up the ladder of flexibility, and it is the top rung: RAW. It is where the most latitude lives, and also the most cost, and knowing where its line sits — the point past which flexibility stops being worth what it takes — is a genuinely professional skill.

RAW video records the sensor's near-unprocessed readings — the raw brightness values straight off the photosites, captured before the demosaicing step that reconstructs full color (recall Chapter 2's Bayer-filter Tech box: the sensor measures one color per photosite, and a demosaic step fills in the rest) and before a picture profile bakes in any contrast, color, or even a fixed white balance. Where a normal codec — even a Log one — hands you a processed, demosaiced image, RAW hands you the sensor data and lets you reconstruct the image in post. That is a profound difference in where the decisions live.

What that buys you is the deepest flexibility in all of capture:

• White balance becomes a post decision, essentially for free. Because the color hasn't been locked in, you can set or change white balance in the edit with no penalty — a safety net that has rescued countless mixed-light shoots.
• Maximum dynamic range and bit depth survive, because nothing has been clipped or crushed by a profile. RAW is typically 12-bit or more.
• The most latitude for heavy grading, VFX, and big-screen delivery — RAW is what high-end commercials, features, and effects work often demand, because every possible decision stays open.

And what it costs you is exactly proportional:

• Enormous files. Truly uncompressed RAW is gigantic; even the compressed RAW formats that make it practical — Blackmagic RAW, Apple ProRes RAW, REDCODE, Canon Cinema RAW Light, and others — produce files far larger than an ordinary codec.
• A demanding pipeline. RAW needs fast, capacious storage; a computer strong enough to handle it (often with proxies, Chapter 27); and a real grading workflow. It is slow to shoot with and slow to finish.
• Overkill for most work, and — this is the part beginners miss — it can make your video worse if you lack the pipeline, because footage you can't offload, back up, or grade properly is footage at risk, and time spent wrestling RAW is time not spent on light, sound, and story.

FIGURE 3.7 — The flexibility ladder: more latitude, more cost, at every rung

  STANDARD 8-bit 4:2:0  ──►  LOG 10-bit 4:2:2  ──►  RAW (12-bit+)
  ┌────────────────────┐    ┌──────────────────┐   ┌───────────────────────┐
  │ ready to watch      │    │ flat; grade it   │   │ sensor data; BUILD the │
  │ least latitude      │    │ lots of latitude │   │ image in post; MOST    │
  │ small files         │    │ bigger files     │   │ latitude; HUGE files   │
  │ any card/computer   │    │ fast card, 10-bit│   │ fast big storage +     │
  │ no grading needed   │    │ grading required │   │ strong computer + time │
  └────────────────────┘    └──────────────────┘   └───────────────────────┘
     talking-head, social      the "cinematic"        high-end / VFX / features
     run-and-gun, fast turn     graded look            — deliberate, not default

  Climb the ladder ONLY as far as your post pipeline (and the job) actually needs.
  Every rung up buys flexibility you must PAY FOR in storage, computer power, and time.

The honest guidance, then. RAW is a specialist's tool with a specialist's costs, and the mark of judgment is reaching for it deliberately — never by default, never "to be safe," never because a spec sheet made it sound like the pro choice. Reach for RAW when the job genuinely demands it: a high-end branded piece or commercial with a big grade (the territory of Project 3, later in this book), heavy VFX or green-screen compositing, a tricky-lighting situation where you must be able to fix white balance and exposure in post, or a big-screen deliverable where every scrap of quality shows. Do not reach for it for a talking-head, a social clip, a run-and-gun event, a fast turnaround, or any shoot where your storage, computer, or schedule can't absorb the load. For the overwhelming majority of what you'll make — and all three of your portfolio projects at the level we'll shoot them — a good standard profile or well-handled Log is not a compromise. It's the right tool.

⚠️ Common Mistake: shooting Log or RAW "to be professional." This is the exact cousin of Chapter 2's "shoot 8K to be safe," and it wrecks as many projects. The ambitious beginner reads that pros shoot Log and RAW, flips their camera to Log (in 8-bit, unknowingly) or RAW, and then: the files won't fit, the laptop can't play them, the footage looks grey and broken because it was never graded, and the whole piece ends up worse than if they'd shot a clean standard profile they knew how to handle. Format is not a status symbol — it's a link in a chain that runs from the card to the final export, and a chain fails at its weakest link. Match your capture format to your whole pipeline — your storage, your computer, your skills, your deadline, and what the job actually needs — not to what sounds impressive. The most professional choice is the one you can carry all the way to delivery.

🔄 Check Your Eye. 1. How is RAW different from a normal codec (even a Log one) — what does it record that they don't? 2. Name one thing RAW makes essentially free in post, and two costs it imposes. 3. Give one shoot where RAW is the right call and one where it clearly isn't.

Check yourself

1. RAW records the sensor's near-unprocessed readings — before demosaicing and before a picture profile bakes in contrast, color, or white balance — so you reconstruct the image in post. A normal codec (even Log) hands you an already-processed, demosaiced image.
2. Free: changing white balance in post with no penalty (also maximum dynamic range/bit depth). Costs: enormous files and a demanding pipeline (fast/big storage, a powerful computer, a real grading workflow, more time).
3. Right: a high-end commercial or VFX-heavy piece with a big grade and the storage/computer to match. Wrong: a talking-head, social clip, or run-and-gun event on a laptop with a tight deadline — where a standard profile or Log is smarter.

3.6 Choosing a capture format for your project

Everything so far has been parts. This section assembles them into a decision — the one you'll actually make on every shoot for the rest of your life: given this job, what do I record it in? The good news is that the decision reduces to a single honest question, asked before you ever touch the camera.

The question is: what will happen to this footage in post? Not "what's the best format?" — there is no best — but "what does this piece, on my pipeline, need?" Everything flows from the answer:

• Will you grade it hard, or lightly? Heavy grade → more bit depth (10-bit), maybe Log. Light correction → 8-bit standard is fine.
• Any green screen, VFX, or fine color work? → 10-bit 4:2:2 if you can.
• What's your storage and computer situation? → high bitrate and Log/RAW demand fast cards, big drives, and a capable computer; be honest about what you have.
• What's the turnaround? → tight deadline favors a ready-to-watch standard profile over a grade-dependent one.
• Can you actually grade Log well yet? → if not, a good standard profile beats mishandled Log every time.

FIGURE 3.8 — Choosing a capture format: one honest question, four common answers

  "What will happen to this footage in POST?"
       │
       ├─ Cut it and post it fast, light color only, tight deadline
       │     → STANDARD profile (Rec.709), highest bitrate you can store.  8-bit 4:2:0 is fine.
       │       (talking-head • social clip • run-and-gun event • news)
       │
       ├─ Build a real "look," moody grade, you CAN and WILL grade it
       │     → LOG (or a flat profile), 10-bit if at all possible, high bitrate, exposed carefully.
       │       (branded piece • music video • narrative • the Café Scene's warm grade)
       │
       ├─ Green screen / VFX / fine colored detail
       │     → 10-bit 4:2:2 minimum; clean, high bitrate.   (compositing • product on color)
       │
       └─ High-end grade, big screen, must fix WB/exposure in post, you have the pipeline
             → RAW (compressed: BRAW/ProRes RAW/etc.).   (commercial • VFX-heavy • feature)

  Match the format to the WHOLE pipeline — card, computer, skill, deadline, delivery — not to
  what sounds impressive. The right format is the one you can carry all the way to delivery.

Let us make it concrete with the four setups this book keeps returning to, because each one pulls the decision a different way:

• The talking-head at a table (your Project 1). A single subject, controlled light, a straightforward cut and light correction. Honest answer for most readers: a good standard profile at the highest bitrate your card sustains, 8-bit 4:2:0 if that's what you have, white balance set correctly in camera. Step up to 10-bit or Log only if you're confident you'll grade it and your kit and skills support it. This is the un-glamorous, correct choice, and it's the one the Production Checkpoint asks you to test.
• The walk-and-talk outdoors. Now dynamic range is the pressure — bright sky, shadowed faces, moving between light and shade. Here a wider-range profile (Log or HLG) if you can grade it, and the highest bitrate you can manage, earn their keep, because you're fighting a high-contrast world your standard profile would clip.
• The product on a table. Color-critical and often heavily graded, sometimes on smooth gradient backdrops that reveal every weakness. This is where 10-bit pays off most visibly — smooth backgrounds without banding, saturated product colors that survive the grade. (We'll shoot exactly this comparison in Case Study 2.)
• The event or room in available light. Run-and-gun, one take only, long record times, mixed and uncontrollable light. Here the priorities flip toward reliability: an efficient codec, a bitrate high enough to survive mixed light but low enough that you don't run out of card, and a format you can offload fast between segments. This is rarely the place for RAW — the pipeline can't absorb it and the moment won't wait.

The other half of the decision: the card-and-offload workflow

Choosing the format is only half of what the Production Checkpoint asks. The other half is what happens to the footage after the camera writes it — the card-and-offload workflow — because the most beautifully chosen format in the world is worthless if the file is lost, corrupted, or overwritten. This is the least exciting paragraph in the chapter and possibly the most important.

The discipline, which we'll formalize as media management in Chapter 27 and backup in Chapter 37, starts the moment you shoot:

1. Use a card fast and large enough for your chosen format (§3.2) — matched with headroom, from a seller you trust.
2. Offload as soon as you can — copy the footage off the card onto a real drive (a computer, an external SSD) before the day is over. Cards are the least safe place footage can live.
3. Back up before you erase anything. The rule that will save your career: never format or reuse a card until its footage exists in at least two other places. A single copy is not a backup; it's a single point of failure waiting to happen. (Chapter 37 makes this the "3-2-1" rule; for now, "two copies before you wipe the card" is the seed.)
4. Then test that it actually imports and plays. Don't assume — confirm. Open a clip in your editor, scrub it, watch it play. Codec problems (§3.1), a card that was silently dropping frames, a corrupt file — all surface here, on the ground, while you can still re-shoot, not three weeks later when you can't.

✂️ In the Edit. The bridge from this section to the timeline is blunt: the format you choose here decides how your edit feels and what it can do, before you've cut a single frame. Shoot a heavy interframe codec in 4K and your laptop may crawl on the timeline (fix: proxies, Chapter 27). Shoot 8-bit 4:2:0 and the grade hits a low ceiling (§3.3). Shoot Log and forget to grade and the whole thing looks broken. Shoot RAW with no pipeline and you can't even open the files. Every one of these is a set-day decision that lands in the edit room — usually on you, three weeks later, wishing you'd chosen differently. The editor you're shooting for is future-you, and future-you has no way to add back what past-you didn't record. Choose the format for the edit you want to have.

🔄 Check Your Eye. 1. What single question should drive every capture-format decision? 2. For a straightforward Project 1 talking-head you'll cut fast and correct lightly, what's the honest format choice — and when would you step up from it? 3. State the one card rule that protects footage you can never re-shoot.

Check yourself

1. "What will happen to this footage in post?" — the grade, the VFX, the deadline, the storage and computer you have, and your skill all follow from that, not from "what's the best format."
2. A good standard profile at the highest bitrate your card sustains, 8-bit 4:2:0 if that's what you have, with white balance set correctly in camera. Step up to 10-bit or Log only if you'll genuinely grade it and your kit and skills support it.
3. Never format or reuse a card until its footage exists in at least two other places (offload and back up before you erase — the seed of the 3-2-1 rule, Chapter 37).

Production Checkpoint

Your task: decide Project 1's capture format and card-and-offload workflow, then shoot a short test that you confirm imports and plays back cleanly. In Chapter 2 you set Project 1's technical baseline — resolution, frame rate, shutter — and shot a camera test. Now you decide how that image will be recorded, and you prove the whole capture-to-computer chain works before it counts.

Do four things, and write the first three in the project note where your Project 1 baseline already lives:

1. Choose a picture profile: standard or Log — and say why in one sentence. For most readers and a straightforward talking-head, the honest answer is a good standard profile at the highest bitrate your card sustains (8-bit 4:2:0 is genuinely fine here), with white balance set correctly in camera. Choose Log only if you're confident you'll grade it, your camera shoots 10-bit, and you'll expose it carefully (Chapter 5). Whichever you pick, tie it to what you'll do in post — not to what sounds impressive.
2. Confirm your quality settings. Note your codec/bitrate (shoot the highest bitrate your card can sustain) and your bit depth and chroma if your camera exposes them (8-bit 4:2:0, 10-bit 4:2:2 — whatever you have). If you're on a phone, note whether it offers a high-quality or 10-bit/HLG mode, and turn it on if it does.
3. Write your card-and-offload workflow. Where does the footage go the moment you stop shooting? Name the card, the drive you'll copy to, and the second place it'll live before you ever reuse the card. "Two copies before I wipe the card" is the minimum.
4. Shoot a short test and confirm it imports and plays cleanly. Record 15–30 seconds with your chosen format on your real Project 1 subject (or a stand-in) in your real location. Then offload it, open it in any editor (DaVinci Resolve is the book's free reference from Chapter 26 on), and actually watch it — does it import without codec errors? Does it scrub and play smoothly? If you shot Log, does it grade back to a normal look? If anything misbehaves, you've just caught it for free.

Why this matters: the recording format is the one Chapter 2–3 decision that is truly irreversible — you can re-light, re-frame, and re-cut, but you cannot un-compress footage or add bit depth that was never recorded. Testing the whole chain — format, card, offload, import, playback — before the day that counts is fix it in pre, not in post applied to the most permanent choice you make. A ten-minute test now prevents the three-week-later discovery that your files won't import, your card was dropping frames, or your 8-bit Log bands the instant you grade it. Keep the test clip; you'll add light to this project in Part III, and by Chapter 15 the whole 60-second piece is in the can.

Summary

A reference-grade recap of the recording pipeline and how to choose within it.

What a video file is:

The quality controls, and what each decides:

Standard vs Log vs RAW — the flexibility ladder:

Decision rules to carry:

• Ask one question: "what will happen to this footage in post?" Everything follows from the answer.
• Match the format to the whole pipeline — card, computer, skill, deadline, delivery — not to what sounds professional.
• Compression damage is permanent. Banding, macroblocking, and smearing baked in at capture cannot be removed in post — the grade only amplifies them.
• Shoot Log only if you'll grade it, in 10-bit, exposed carefully. 8-bit Log is a trap; a good standard profile beats mishandled Log.
• RAW is a deliberate specialist choice, not a safety net or a status symbol.
• Never format a card until its footage lives in two other places. Offload, back up, then test that it imports and plays.

Common mistakes → fixes:

This chapter's project step: choose Project 1's capture format (standard vs Log) and card-and-offload workflow, then shoot a short test and confirm it imports and plays back cleanly.

Spaced Review

Bringing back Chapter 2, because spaced retrieval is how the technical foundation sticks. Answer from memory, then check.

1. Chapter 2 said the sensor turns light into a grid of numbers, and its 🔬 The Tech box described the Bayer filter and demosaicing. Where in this chapter's pipeline does RAW sit relative to demosaicing — before it or after it?
2. Chapter 2 taught that shooting 4K to deliver 1080 buys reframe headroom — coverage for the edit. How is choosing a higher bitrate or 10-bit the same kind of thinking (a set-day decision made for the edit)?
3. In Chapter 2, noise came from starved photosites in the dark. Why does that Chapter 2 problem come back to bite you specifically when you underexpose Log and then grade it?
4. Chapter 2's walk-up order put "format / codec" as a step after resolution, frame rate, and shutter. Now that you've read this chapter, why does that ordering make sense — what does the format decision depend on?

What's Next

You now understand the whole capture pipeline — from the light landing on the sensor (Chapter 2) to the compressed, encoded file on your card (this chapter) — and, crucially, which of your post-production powers are won or lost the moment you choose a recording format. But we've been treating the image as if it arrives fully formed on the sensor, and it doesn't: everything about depth, focus, background separation, and the very "look" of a shot is shaped by the piece of glass in front of the sensor. In Chapter 4 we turn to lenses — focal length, aperture, depth of field — and discover the claim that surprises every beginner: the lens shapes the image far more than the camera body does, and focal length is a storytelling choice long before it's a zoom level.