Video Production curriculum

Before choosing gear or locations, settle what the video is for and who will watch it. Those answers shape many later choices, from whether you need video at all to budget, time and skills, and naming the audience early tells you what they will want to know.

A treatment or outline tells the whole story briefly before any scene is written in shooting detail, so everyone agrees on which film they are making. Early screenwriters split the job the same way: the synopsis outlined the story, and the scenario then worked it out scene by scene in action.

A shot list numbers every setup you plan to film, close-ups and inserts included, so nothing essential depends on improvising. Silent-era scripts worked the same way: each close-up got its own scene number, and a scene plot showed at a glance which sets were needed and how many scenes each held.

Visit a location before the shoot and check what the camera and microphone will meet there: where the light falls at your shooting time, which outlets you can reach, and what hums, ticks or echoes. A big room with hard furniture sounds hollow, and a refrigerator or air conditioner you barely notice will still be on the track.

Shoot in the order that saves setups, not the order of the script: finish everything in one set or location before moving on, because every move to a new setting costs time while people wait. Listing the scenes each minor character appears in also shows where one performer can cover two small parts.

Where you film is often regulated, and the rules change. In U.S. national parks, a 2025 law dropped the permit and fee for most shoots of eight or fewer people who use only hand-carried gear in areas open to the public, commercial or not. A bigger crew, a closed area or exclusive use of a site still needs a permit, fees and insurance.

Shot size decides what the viewer can read. A wide shot holds people and setting but loses faces, which is why early crews hid a stunt double in one; a close-up enlarges a face or hand where an expression would not register at normal range; an extreme close-up isolates one detail, such as a letter or a ring.

Open a sequence with a shot wide enough to show the whole area of action, its location, people and objects, so the closer shots that follow have a place to sit. Frame it loosely so people can move, but frame it so the subject is still recognizable, and use it sparingly on small screens, where a wide view loses much of its detail.

Coverage means filming a scene from several setups, a wide master plus closer shots and inserts, so the editor can choose and trim. Shoot all the wide shots in a setting first and then move in for the close-ups, which saves repositioning the camera; then cut between them, as in going from an insert of a hand back to the wide view of the room.

A cut-in shows a detail of the main action and a cutaway leaves it for something related, such as a bystander or a sign. Both let the editor drop dull stretches, bridging time and distance, and both mask a jump in action, such as a chair that moved between takes, because viewers notice a jump most in the very next shot.

Camera height and tilt change how a subject reads. A low angle pointed up tends to lend strength and dominance, while a high viewpoint helps orient the viewer and makes the subject seem smaller and weaker. Treat these as tendencies that context and story can overturn, and walk around the subject to compare heights before you commit.

A point-of-view shot puts the viewer in a character's place by showing what that character sees. Rather than watching a banker open a telegram at his desk, the audience sees the message itself fill the screen and reads it over his shoulder, so the shot carries his attention instead of showing him.

Holding one shot size for a whole sequence leaves the editor nothing to cut to and the viewer nothing new to read. Vary size and angle with purpose: move closer when an expression matters, go back wide to re-establish the space, and use each close-up for a reason rather than out of habit, since overused close-ups were criticized even in early cinema.

Divide the frame into thirds both ways and place the subject on a line or an intersection rather than dead center. A centered subject splits the frame into two equal halves competing for attention, which is why off-center placement usually feels more dynamic. Treat it as a habit, not a law: however you arrange things, fill the frame with what tells the story.

Headroom is the space between the top of a head and the top of the frame. Leave some above the head in most shots, since a head pressed to the edge looks cramped, but let it shrink as you move in: in a tight close-up you may crop the top of the head as long as part of the shoulders stays in frame as a reference.

Leave open space on the side a person or object faces, points or moves toward. Without it a moving subject looks as if it is bumping into the edge of the frame, and with it the frame shows where the subject is going rather than where it has been, which is why you lead lateral movement with the camera.

In a medium or close shot, put the eyes near the upper third line rather than the middle of the frame. The face then lands on one of the thirds intersections, the points the rule singles out, and the space below holds shoulders and body instead of empty air above the head.

Arrange foreground, midground and background so the eye travels through the frame instead of across a flat wall. Lines such as a road or a rope can lead the eye from the front of the frame to the subject, and nearby objects can frame the subject from the edges, so plan what sits in front of the subject as carefully as what sits behind.

Match the frame's shape to the subject and to where it will be shown. HD is a wide 16:9 frame that suits subjects spread horizontally, while a frame turned upright suits a standing figure or a corridor and puts more pixels on it. Planning the shape matters, because cropping a widescreen frame to vertical later keeps less than a third of its width.

Keep titles, lower thirds and important action away from the very edges of the frame and from the zones where captions and platform buttons appear. U.S. television rules say captions should not block important visual content or run off the edge of the screen, so action crammed into the bottom of the frame invites a collision.

Look past the subject before you roll: a tree, pole or streetlight that seems to grow out of someone's head, or a bright car behind them, pulls attention from what matters. Move the camera or the subject until the background supports the story, but avoid a featureless one by default, since a plain backdrop detaches a person from the scene.

Exposure comes from three controls: aperture (how wide the lens opens), shutter speed (how long light is collected) and ISO (how sensitive the sensor is). In video the shutter is tied to the frame cycle and changing it changes how motion looks, so brightness is usually set with aperture, ISO or a filter, each with its own cost in focus depth or grain.

ISO, or gain, amplifies the signal the sensor already collected; it adds no light, so a higher setting brightens the picture along with its grain. Some sensors offer two gain modes: high gain amplifies each unit of charge more for dim scenes, while low gain lets each pixel hold several times more charge before it fills, which protects highlights.

A neutral density filter darkens the whole image evenly, cutting light at the lens while leaving colors as they were. That matters in video because the shutter is tied to the frame rate for natural motion, so in bright sun an ND filter is what lets you open the aperture for shallow focus without overexposing or forcing the shutter faster.

Dynamic range is the ratio between the brightest and darkest values a sensor or image can hold. When a scene exceeds it, detail is lost at one end, so decide which end to protect; raw recording keeps more of the sensor's range than a standard HD display can show, which leaves room to make that choice later in the grade.

Exposure instruments measure light; they do not judge it. A reflected light meter reads whatever it is pointed at as if it were middle gray, so a reading off a white wall underexposes the scene and one off a dark coat overexposes it. A waveform monitor instead graphs the video signal itself, so you can check levels rather than guess.

When the bright and dark parts of a scene are too far apart for both to keep detail, exposure decides which end you lose. Expose for the shadows and the highlights wash out completely; expose for the highlights and the shadows go dark without detail. Decide which part of the frame matters most, or add fill light to narrow the gap.

Focal length sets magnification: from the same spot, the image of a subject grows in direct proportion to the focal length, so a long lens narrows the view and a short one widens it. A telephoto only seems to compress distance; perspective depends on where the camera stands, so the flattened look comes from shooting from farther away.

Field of view depends on both the focal length and the size of the sensor, so the same lens frames tighter on a smaller sensor than on a larger one. That is why a focal length quoted for one camera only predicts the framing on another once the two sensor sizes are taken into account.

Three things set how much of the scene looks sharp: aperture, focal length and distance. Depth of field grows as you stop down, use a shorter lens or focus farther away, and shrinks as you open up, use a longer lens or move closer, so a fast long lens near a face can hold the eye sharp while the ear goes soft.

The closer the subject, the thinner the zone that stays sharp, so a small movement that goes unnoticed in a wide shot ruins a close-up. A fixed focus setting that covers people at a moderate distance stops working once you approach much closer, and focus then has to be set, and held, precisely.

A subject that is sharp while the rest of the frame blurs stands out, so shifting the sharp plane between subjects during a shot moves the viewer's emphasis with it. Because depth of field is thinnest with long lenses, wide apertures and close subjects, set focus by measured distances before the take rather than hunting during it.

Total pixel count says little about the detail you get on a subject; what matters is how many pixels land on it. One way to measure this is pixels per foot, the sensor's pixel width divided by the width of the scene in view, so the same camera framed twice as wide puts half as many pixels on each foot, and small subjects need more.

A rolling-shutter sensor reads its rows one after another, so each row is exposed a moment later than the row above it. Anything that moves fast during that sweep, a quick pan or a passing car, is recorded leaning or skewed. A global shutter exposes every pixel at once and avoids it, at some cost in noise and dynamic range.

Bit depth sets how many brightness steps each color channel can store: 8 bits give 256, 10 bits give 1,024. The extra steps matter most when you grade, because stretching tones apart spreads the available steps further. Converting an 8-bit source into a 10-bit file only upsamples it and adds no detail that was never recorded.

Chroma subsampling stores color more coarsely than brightness. The ratio describes a block 4 pixels wide and 2 rows high: the second number counts color samples in the first row, the third counts additional ones in the second. So 4:4:4 keeps full color, 4:2:2 halves it across each row, and 4:2:0 adds none in the second row.

A raw file keeps the sensor's data with very little processing, so choices such as white balance and color rendering are made later on a computer; a baked file arrives with the camera's own algorithms having made those decisions. Raw keeps more color and range to work with, at the cost of special software and an extra processing step.

A pan swings the camera left or right and a tilt swings it up or down while the tripod stays put, so the view changes without the camera traveling. Move smoothly and for a reason, such as keeping a moving subject in frame; a jumpy pan worries the eye and can smear the picture into a blur.

A dolly moves the camera toward or away from the subject, so a wide view can become a close-up without breaking the action, while a truck moves it sideways, keeping the same distance to follow lateral movement. Because the camera itself travels, nearer and farther objects shift against each other and the background changes continuously.

A zoom and a dolly can both make a subject grow on screen, but they are not the same move. During a zoom the camera stays put, so perspective does not change and the background simply enlarges with the subject; during a dolly the camera travels, so the relationship between near and far objects changes as it moves.

A stabilizer isolates the camera from the operator's small shakes so a moving shot stays smooth. The idea echoes the gimbal mounting of a gyroscope, which lets its axis keep pointing one way while the support turns, and hand cameras carried gyroscopic steadiers more than a century ago, trading extra weight for steadiness.

Move the camera because something in the scene asks for it: swing to follow riders crossing the frame, or travel toward a setting worth revealing. When nothing calls for movement, cut instead; early studios saved the slow traveling approach for an imposing set and used an ordinary close-up the rest of the time.

Three-point lighting gives each lamp one job. The key is the main, strongest light, placed to one side so that side of the face is lit and the other holds some shadow; the fill softens that shadow without erasing it; the back light, opposite the key and behind the subject, separates the subject from the background.

A shadow's edge is soft where only part of the light source is blocked, so the size of the source relative to its distance from the subject decides how hard the light looks. Midday sun is small in the sky and cuts harsh shadows under the nose and eyes; thin cloud, a softbox or a big window spreads the source and softens them.

Light from a lamp falls off with the square of the distance, so a small step back costs far more light than it seems to. A lamp close to a face therefore lights it unevenly from front to back and darkens the background quickly, while a distant source such as the sun lights the whole scene almost evenly.

The balance between key and fill sets the mood, and the simplest control is the fill's distance: move it closer or farther to change the ratio without touching the key. Strong light and shade, with part of the face in deep shadow, suits a dark background, while ordinary lighting that keeps the face full of delicate half-tones looks wrong against one.

Motivated lighting makes every light in the scene seem to originate in something the viewer can believe in, such as a window, a lamp or a fire. Mismatches give the illusion away: a lantern cannot produce the brightness of gaslight, and a north window cannot throw sunlight onto a figure, so match each light's direction, strength and quality to its apparent source.

Shadows are rarely black because light bounces off nearby surfaces back onto the subject, so a white wall acts as fill whether you want it or not. Negative fill reverses that: a dark flag or cloth beside the face absorbs light instead of returning it, which deepens the shadow side and gives the face more shape.

Existing light can be the best light if you choose the time and place. Low sun early and late in the day is warm and soft and gives depth; high noon is harsh and casts shadows under the eyes; light overcast is soft and easy on faces. Indoors, turn a subject toward a large window for a gentle key.

Color temperature names a light's color by the temperature, in kelvins, of an ideal black body that would glow that color. The scale runs against everyday intuition about heat: a higher number is bluer, as with daylight, and a lower number is redder, as with tungsten video lights, so the warm look of sunrise sits low on the scale.

White balance tells the camera what should read as neutral white under the light you are using, which removes the color cast that light would leave. Auto modes usually cope but can fail under mixed, colored or changing light, so set it manually on a white card in the shooting light, and reset it whenever the light changes.

When window daylight and indoor lights fall on the same face, white balance can only favor one, because daylight is bluer than typical indoor and video lights. Fix it on set: move away from the window, put blue gels on your lights to match the daylight, or cover the window with amber gel to match the lights.

Color temperature describes only a light's overall tint, so two lamps with the same number can still render colors differently; what matters is the shape of the spectrum. The Color Rendering Index scores only eight test colors of low to medium saturation, so lamps with peaky spectra, such as some LEDs, can do well there yet render saturated colors poorly.

A light whose output pulses over time, such as an intensity-modulated LED, can band on camera. A rolling-shutter sensor samples its rows one after another, so variation in time is recorded as a pattern in space: bright and dark stripes running across the frame. Test each light on camera before you roll, at the frame rate and shutter you plan to use.

A polar pattern maps how sensitive a microphone is from each direction. An omni hears all around it; a cardioid is most sensitive in front and fully rejects sound arriving from directly behind; a supercardioid is tuned for the best ratio of front to back pickup. Directional patterns help because they pick up less room noise and reverberation.

A dynamic mic makes its signal with a moving coil: it is low in noise and needs no built-in preamplifier, but it is less sensitive because the coil and diaphragm are heavy. A condenser senses a light diaphragm moving against a plate held at a fixed voltage, so it needs power from the recorder or a battery; in voice-measurement studies condensers have tracked the voice more accurately.

A shotgun narrows its pickup toward the front, which helps against noise from the sides, but it does not reach across a room the way a zoom lens does. Directional designs extend the usable distance somewhat; distance still decides quality, because a mic close to the mouth captures far less background noise and room echo. Get it as close as the frame allows.

A lavalier clips to clothing a few inches from the mouth, so it stays close however the speaker turns and is especially useful for a soft voice. Its weak point is the clothing itself: a lav rubbing against rough fabric spoils the track, so place and secure it with care, and expect some shirts and tops to make that awkward.

A built-in camera mic sits wherever the camera is, usually far from the speaker and aimed at the scene rather than the mouth, so room echo and background noise compete with the speech. An external microphone placed close to the speaker usually sounds better and gives you far more freedom in where the camera goes.

Moving the microphone closer is the cheapest improvement there is: the voice gets louder while the room's noise and echo stay the same, so speech stands out far more clearly against them. Voice research has found that a short mouth-to-mic distance cuts the effects of background noise and reverberation even with a low-cost headset mic.

Bring a directional or cardioid microphone very close to a voice and its bass response rises; this is the proximity effect. Used on purpose it adds warmth, but it also makes a voice swell and thin out as the speaker leans in and back. An omnidirectional mic largely avoids it, and in one test an omni headset showed none at all.

Hard, bare rooms bounce sound around, and those reflections blur speech. A big room with wood or metal furniture sounds hollow next to a small room with upholstered chairs and curtains, so pick the softer space when you can, add blankets or absorbing panels when you cannot, and keep the mic close so the direct voice outweighs the echo.

Wind flowing over a microphone creates turbulence that the mic records as low rumble and, in strong gusts, outright distortion, and a thin foam cover is not enough in strong wind. In one field study a furry windshield over a shotgun lowered the low-frequency noise floor compared with foam alone, while a parabolic dish proved vulnerable to gusts.

No room is silent: even a quiet one carries a steady background of air handling, appliance hum and distant traffic, and every take records it. Room tone is a stretch of that background recorded on its own, with everyone still, so the editor has a matching bed to fill gaps and smooth cuts instead of dropping into dead silence.

Recording sound on its own device frees the microphone from the camera, but picture and sound then exist as separate files that must be lined up afterward. Give the edit a shared reference on every take, a slate clap or matching timecode, because independently running devices make synchronization a real problem.

Check at a glance that the recorder is actually recording and the level is healthy, then keep listening on headphones as you go. A peak light warns when the input is too high, but problems such as clothing rustle, a humming appliance or a ticking clock are things you hear, and catching them during the take is what lets you fix them.

Set level once, at the first stage, the microphone preamp: enough gain that normal speech sits well above the chain's own noise, not so much that the loudest moment saturates the preamp. A preamp driven into saturation stops tracking the signal faithfully, and no later fader can put back what it flattened.

Record so that a speaker's usual level sits comfortably below the maximum and the loudest laugh or shout still has room above it. Voices swing widely between quiet and loud, so a level set for normal talk with no margin will clip on the first burst of emotion.

Digital audio has a hard ceiling called full scale, 0 dBFS. A sample that would go higher is simply cut off at the limit, flattening the top of the waveform into a hard-limited peak, and nothing above that ceiling was stored to recover later. Careful recordists check every take for samples sitting at full scale.

Uncompressed PCM audio gains fidelity as its sample rate (samples per second) and sample size (bits per sample) rise. Archival spoken-word work often asks for 24-bit, 96 kHz WAV, and highly compressed formats such as MP3 hold up less well over time, so record uncompressed and compress only for delivery.

A limiter is a safety net: it pulls the recording level down automatically when a peak would clip, which can save a take with an unexpected shout. Because it changes the level on its own, work that needs true levels, such as acoustic measurement, turns it off, and good practice is to set gain properly first so it catches only the rare peak.

Every recording chain adds hiss of its own, from the microphone's self-noise and the preamp, and what matters is how far the wanted sound sits above it. Raise the signal at the source by moving the mic closer and setting gain well; boosting a quiet track afterward lifts the hiss right along with the voice.

Frame rate is how many pictures per second you record and play. Too few and motion stops reading as continuous; early cinema treated twelve to sixteen per second as the minimum and found faster subjects needed more. The rate is part of a video format's name, as in 720p60, so match it to the format your delivery and display expect before you shoot.

Film cameras exposed each frame through a spinning disc with an opaque sector, so shutter angle describes how much of every frame interval the shutter is open. At 180 degrees it is open half the time, and the motion during the closed half is never recorded, yet the eye fills the gap. The 180-degree look is a convention, not physics.

Shutter time sets how much motion each frame smears: a slow shutter blurs moving things, and a fast one renders each frame crisply. The camera records nothing while the shutter is closed and the eye fills in the gap, so both choices change how motion looks; for ordinary-looking motion, keep the shutter near half the frame interval.

Slow motion comes from recording more frames per second than you play back. Shoot at 60 frames per second and play at 24, and each second of action lasts two and a half; early scientific filmmakers did the same at extreme rates, slowing thousands of pictures a second to normal projection speed so a bullet's flight could be followed.

Cameras and recorders running on their own clocks do not line up by themselves. Professional cameras can be timecode-synchronized with a cable, and any shared reference, such as jammed timecode or a common sync pulse, gives the edit one timeline; files carrying separate timecodes that were never synchronized must be aligned some other way.

A container such as MP4, MOV or MXF is the file that wraps video, audio and other streams together; a codec such as H.264 or ProRes is the method that compresses the picture inside it. One container can hold many codecs, so two .mov files can behave differently, and whether one plays depends on the software reading the codec, not the extension.

An intraframe (All-I) codec compresses every frame on its own, while a long-GOP codec stores occasional complete I-frames and predicts the frames between them from their neighbors, which saves far more space. Decoding can start only at an I-frame, so an edit point between them must first be rebuilt from its references, the extra work that makes long-GOP footage heavier to edit.

An intermediate, or mezzanine, codec such as ProRes 422 sits between the camera and the final export. It is a high-quality, 10-bit, intraframe-only format built for editing and finishing rather than small files: at 1080p and 29.97 fps it targets about 147 Mbps, roughly 65 GB an hour, still far smaller than uncompressed video.

H.264 is the safe default because fixed and mobile devices typically decode it. HEVC (H.265) compresses further and handles higher bit depths, but its patent licensing has held back some large web publishers; AV1 is open and royalty-free, created partly in response to that licensing. Choose by what your audience's devices decode, not by efficiency alone.

Compressed frames vary widely in size: complex, fast-moving shots need more bits than still ones. Variable bitrate lets the encoder follow that need around a target average, so file size depends on the content; constant bitrate smooths the stream to a steady rate, easier to send over a network but needing a buffer to absorb the variation.

Every time a lossy file is decoded and compressed again, new compression errors are added to the old ones, so quality drops a little with each generation. Edit from the best original you have, render the final export once, and use lossless or high-quality intermediates when footage must pass through several steps.

A proxy is a small, light copy of each camera file that a modest computer can play smoothly. You cut with the proxies and then relink to the original files for grading and export, so the finished video uses full quality; ProRes 422 Proxy, for example, targets about 45 Mbps, under a third of standard ProRes 422.

A gamut is the range of colors a system can show, fixed by the exact chromaticity of its red, green and blue primaries; displays are rated by how much of a standard gamut such as Rec. 2020 they cover. Rec. 709 sets the image parameters for HD, and a camera sensor sees more than that, so raw keeps more colors than a Rec. 709 display can show.

A transfer function, often called gamma, is the curve linking stored code values to light. Images are usually encoded through a compressive curve and the display applies the inverse, so the picture looks right only when the encoding and the display curves match; capture also needs a tone curve to fit a scene's wide brightness range into a display's narrower one.

Log recording applies a strongly compressive tone curve at capture, squeezing the sensor's wide brightness range into the available code values so highlights and shadows both survive. It is a container for range, not a finished look: the footage is not meant to be watched as recorded, and a conversion or grade must map it to the display afterward.

HDR is about the range of brightness, not richer color. Screens built for standard dynamic range have traditionally shown about 1 to 100 cd/m2, while HDR displays can go below 0.01 and up to 10,000. Wider color is a separate property, the gamut, with its own standards. In an HDR grade the white of a lit card is not peak brightness; the range above it is room for highlights.

PQ and HLG are the two transfer functions standardized for HDR video, and they treat brightness differently. Guidelines set a ceiling of 10,000 cd/m2 for PQ content and 1,000 for HLG, and with HLG the level of diffuse white depends on the display showing it. A file has to be decoded with the function it was graded in, so confirm which one each destination asks for.

HDR10+ adds dynamic metadata to HDR10 files, using a SMPTE method for color volume transforms, so the information can change over the course of a program. Codecs such as HEVC carry supplemental data that is not needed to decode the pictures but helps with display, and an application that does not recognize it cannot use the feature. A conversion that drops it still plays, without that help.

Tone mapping converts HDR material into the narrower range a standard screen can show. It is a squeeze, not a copy: common methods split the image into base and detail layers, and they can overshoot into halos, over-enhanced detail or oversaturated color, so judge the SDR version as a grade of its own. Expanding SDR toward HDR can instead magnify compression blocking in the source.

Make a checksum where a file is first copied, then confirm it after each move; a match shows the file arrived unaltered, and archives check again when it reaches long-term storage. Until a verified copy exists elsewhere, the card is the only copy, and removable media can be overwritten by accident, so formatting it is the last step, never the first.

The rule spreads risk across separate failures. Two kinds of storage, such as a drive and the cloud, guard against a hazard that hits one medium, and the off-site copy survives a disaster at your location. A backup counts only if it restores, so run scheduled recovery tests instead of trusting that the copy exists.

Give each file a descriptive name and put it in a folder structure you write down, so the layout explains itself later. Keep the folder tree a camera writes to its card intact, because that original order records how the files relate. A name can also carry an identifier that links the file to a fuller description kept elsewhere.

Label footage while you shoot. A slate recorded at the head of a scene, plus a data sheet, tells the editor what each shot is, which matters most when scenes are filmed out of script order. Footage with no record can be useless however good it looks, so collect metadata from the shoot onward rather than rebuilding it from memory.

Angles line up when every camera shares a time reference. Timecode records an exact address for each frame, so one continuous master timecode becomes the reference point for everything that depends on time, audio sync included. A slate at the head of each take identifies it, so you can find the matching take from every camera.

The line of action is an imaginary line created by the subject's movement. Keep every camera setup on one side of it and screen direction holds from shot to shot; cross it and the subject seems to turn around. To change sides legitimately, show the subject actually changing direction, or cut through a neutral head-on or tail-away shot.

Screen direction is the way a subject moves or looks as the camera sees it, not as it happened. Once a trip runs left to right, keep it that way, and the return can run right to left; two subjects moving in from opposite sides read as closing on each other. An unexplained flip makes the audience think the subject turned back.

When you control the action, have the same movement play out in both setups, so it overlaps the two shots and the cut can fall partway through it. In uncontrolled action that overlap is often impossible to film, which is what cut-ins and cutaways are for: they bridge a gap where the movement does not match.

In a sequence on one subject, change the camera angle between shots; two nearly identical framings cut together leave a jump in the action. A cutaway to related, already established material, or a cut-in to a detail, covers that jump and can hide removed time. Change too much, though, and the shots seem to come from different places.

Where a subject looks is screen direction too. The direction of a gaze should match across consecutive shots, or continuity breaks, and keeping the camera on one side of the action axis holds it. A close-up of an observer turning to follow action off screen tells the audience where that action is, and lets direction change.

The standing instruction is to change the camera angle between shots within a sequence, so each new shot shows the subject from a fresh position. The opposite mistake is real too: an angle change so large that the shots seem to come from different locations leaves the viewer disoriented.

Continuity is the match of details between successive shots of one piece of action: props, lighting, sound level, where things sit in the frame, and the direction of movement. Takes are filmed at different times, so anything that can change between them, such as clothing or a person's position, needs checking before you roll again.

A rough cut is the first edited pass; as it is refined it becomes the cutting copy that decisions are made on. In film, that copy then guided conforming the original negative, with an edit decision list naming every source shot. The working cut carries the decisions, and the final is rebuilt from full-quality originals.

Most changes from shot to shot should be plain cuts, with dissolves, wipes and fades saved for a reason. A program crowded with transitions looks unprofessional rather than polished. In film editing a fade or dissolve was an instruction marked on the work print, a reminder that each one is a deliberate choice.

A dissolve is an effect for moving from one sequence to another, and in film it was made during printing as a planned effect marked in advance. That places it at the boundary between sequences. Between ordinary shots inside one scene, the plain cut stays the default.

When you cut to new pictures, the sound from the clip you left can keep running under them. B-roll works this way: a speaker's audio continues over shots of what they describe. The sound edit and the picture edit then fall on different frames, so the voice carries the viewer across the change of image.

Cutting between angles lets you show action and reaction in quick or slow succession, which is how an edit adds or takes away emphasis. Holding one camera angle for a long stretch tends to lose interest, which is a reason to change shots when the content calls for it, not to cut fast everywhere.

A few chosen shots can stand in for a long stretch of action. Show a scoreboard, a few seconds of play, then the scoreboard again, and the audience fills in the time that passed. Highlights joined by cutaways to the crowd can feel like the whole event, as long as the cutaways hide what was removed.

Speech is the track a viewer most needs and, for many, the hardest to follow: about one in three people in the U.S. aged 65 to 74 has hearing loss, and aging hearing struggles with speech in noise. Keep dialogue on its own track, as film soundtracks did, so it can be balanced against music and effects.

Noise is unwanted sound in a track, heard as hiss or hum. Prevent what you can on location; listening with your eyes closed shows what the microphone will hear. Then treat processing like any repair: archives keep the original capture unaltered and apply improvements only to copies made from it, so a pass that goes too far can be redone.

Ambient sound is the background that surrounds a location, picked up by the microphone along with everything else. Recorded on its own, without picture, it becomes wild sound, which can come from the same event yet sit under shots filmed separately. That lets a scene keep the feel of its place across cuts.

Keep music on its own track, apart from dialogue and effects; film soundtracks were built from separate pre-mix elements for exactly this. The separation lets you cut or replace music without touching speech, and lets a music-and-effects mix be reused under a different narration. Music carries mood, and it needs permission like any other work.

Background sound competes with speech, which is why hearing-health advice tells listeners to turn down the TV to follow a conversation. A viewer cannot turn down your music alone, so lower it under dialogue and let it come up in the gaps. Separate music, effects and speech tracks before the final mix are what make that balance adjustable.

Integrated loudness expresses a whole file as one value in LUFS, and broadcast audio metadata stores it apart from the file's maximum true peak, measured in dBTP. They are different quantities, so a peak meter cannot tell you integrated loudness, and a loudness reading says nothing about how high the peaks go.

In the U.S., FCC rules under the CALM Act require television stations to comply with the ATSC A/85 practice for commercials, and a station shows that for spots it inserts by measuring their loudness and keeping the dialnorm metadata matched to it. That is a broadcast duty; a file headed anywhere else must meet the spec that destination publishes, so read each one.

Integrated loudness is measured over the entire file, so meter the whole program rather than judging by its loudest moment. Broadcast audio metadata also records the highest momentary and short-term loudness and the loudness range as separate values, which flag a passage that spikes while the program as a whole measures fine.

Correction repairs problems. A timed film print carried scene-to-scene light changes to fix color and density, and the first such print went to the director for approval. Archives leave the master unaltered and allow minimal color correction only on copies made from it, and they set video levels against color bars when the source has them.

Shots meant to belong together have to match in lighting and color, because continuity covers lighting as well as props. A camera's white balance holds only for the light it was set under, so reset it when the color temperature changes. The film-era fix was a timed print, with light changes scene by scene to correct color and density.

Keep important text inside the safe title area, since the edges of the full frame may be lost on a monitor. Make text large, because a small screen shows less detail, and keep artwork flat and matte so reflections do not hide letters. A title shot from artwork also needs the camera centered and level with it.

A lower third is text keyed over the picture to name a speaker. It competes for space with captions, which FCC rules for television say must not block faces or featured text such as graphics, so plan their positions together. A name shown only on screen never reaches a blind viewer, so have speakers identify themselves aloud as well.

A matte blocks part of the frame and lets the rest through, so one image can be combined with another. Film made them from cut-outs in front of the lens or high-contrast strips run through the printer, and used them to build superimposed titles, elaborate wipes and effects; the same device is also called a mask.

A master is the highest-quality version, and each viewing copy is made from it at equal or lower quality, never the reverse. Archives keep the master and serve copies at a bit rate low enough to stream easily. Export each platform's file from the master, not from another export, so quality is lost once rather than again and again.

Keep the source's frame rate when you transcode, even an unusual one; archival guidance is to retain it rather than normalize to a standard rate, and converting risks unwanted artifacts. Keep the original chroma subsampling too. A frame rate above the source's makes the file much larger without the source having more to show.

Bit rate is how much data each second of video gets. Higher rates capture more information and make larger files, so the right rate depends on the use: a master for later work warrants a high one, while a copy meant to stream needs one low enough to play smoothly over an ordinary connection.

Delivery audio is defined by codec, sample rate, bit depth and channel layout, and a master and its viewing copy can differ in every one. One archive's video master carries four channels of 24-bit uncompressed audio at 48 kHz, while its online copy carries compressed stereo at the same sample rate. With more than two channels, pick a format that labels them.

Burned-in, or open, captions are part of the picture: always visible, impossible to switch off, and fixed in the style you chose. Sidecar files such as SRT or WebVTT are not encoded into the video; the player reads their timing and draws the text over it, so viewers can turn them on or off and you can correct the text without exporting again.

Lossy compression throws detail away, and compressing again compounds the loss, much as each tape dub moved one generation further from the original. Go from source to target in a single step where you can, and when a service will encode your file again, give it a version well above its final target.

Newer codecs trade compatibility and encoding effort for efficiency. HEVC improved on H.264's compression, yet its patent situation slowed adoption; AV1 is royalty-free and compresses well but can need longer encoding times. A file that plays on nearly everything may need more bits than one that plays only on newer devices.

Captions serve viewers who cannot hear the audio: they carry the dialogue plus who is speaking and the sounds that matter. Subtitles translate dialogue for viewers who hear it but do not know the language, and they rarely identify speakers or sounds, so under the Section 508 standards a subtitle track does not satisfy the caption requirement.

Under FCC caption rules for television, a caption appears as its speech or sound begins, ends about when it ends, and moves at a speed viewers can read. Lines must not overlap one another or run off the screen. Automatic captions often fall out of sync and drop punctuation, so check timing against the audio rather than trusting a machine draft.

Captions name who is speaking, on camera or off, because a viewer who cannot hear has no voice to tell people apart. FCC caption rules list speaker identity among the information captions should supply. Machine captions often miss speaker changes, so add them by hand and keep the naming style the same throughout.

Caption the sounds that carry meaning, such as a telephone ringing, footsteps, a fire alarm or music, not only the words. A viewer who cannot hear needs them to follow what happens. Automatic captioning tends to leave them out, so they are among the first things to check when you review a caption file.

Audio description narrates a program's key visual elements for viewers who are blind or have low vision, placed in natural pauses and never over dialogue unless it cannot be avoided. Not everything can be described, so choose what matters most, including on-screen titles and speaker changes. Leaving room between lines during production beats retrofitting it.

A transcript is a program's speech and audio as plain text, with no timing attached. Captions are built by dividing that text into time-coded blocks synchronized with the audio, so a script or transcript is a head start on captioning. Plan the transcript alongside captions and description rather than as an afterthought.

Photosensitive epilepsy is a reflex epilepsy: flashing, bold patterns or certain color changes can trigger seizures, and a flash that looks larger, or sits in the center of view, is more provocative. Accessibility rules cap how often content may flash in any one second unless the flashing area stays under set thresholds, so test the full screen and looping versions as well.

Fuses and breakers open a circuit when too much current flows, to keep wires from overheating, so a trip means the circuit is carrying more than it is built for. Flexible cords carry their own ratings, marked along their length, and being exposed they are damaged more easily than fixed wiring. Spread heavy lights across circuits instead of resetting the breaker.

Grounding gives fault current a low-resistance path to earth, so a tool whose metal case becomes live sends current there instead of through you, but only if that path is intact. A GFCI cuts power when the current going out and the current coming back differ by a few milliamperes, and it is the protection used in wet locations.

Cords are damaged by door and window edges, staples and abrasion, and an exposed cord is more vulnerable than fixed wiring, so route runs away from doors. Walkways must be kept orderly and free of hazards, and a cable crossing a path is a trip hazard to cover or reroute.

Tungsten-halogen, or quartz, lamps run very hot; even in copy work their heat can curl an original placed under them, so keep heat-sensitive material well clear of a fixture. Never handle the bulb with bare fingers: oil from skin leaves a concentrated hot spot that can make the quartz bubble and the lamp burn out. Electronic flash does not give off that heat.

OSHA requires fall protection from four feet up in general industry and from six feet in construction, so even a low lighting platform can need it. Set ladders only on stable, level ground unless they are secured or stabilized, and never move, shift or extend a ladder while someone is standing on it.

To fly a small drone for work or business, the person at the controls needs an FAA remote pilot certificate carrying the small UAS rating, or must be directly supervised by someone who holds one. It is earned by passing an FAA knowledge test, and it attaches to the pilot, not the aircraft or the client: owning a drone or commissioning the shot does not cover the flight.

Class B, C and D airspace, and the surface-area Class E airspace around an airport, are controlled: a small drone needs authorization from air traffic control before it enters. LAANC is one FAA route to that access at low altitude near participating airports. Arrange it before the shoot day, because the authorization has to come first, not once you are airborne.

Part 107 caps how high a small drone may fly above ground level. The pilot, any visual observer and whoever works the controls need to be able to see the drone with their own eyes, glasses allowed, to know its location, altitude and direction, and throughout the flight either the observer or the pilot must actually keep it in sight.

Flying over people falls into four categories, set by the aircraft's weight, its performance requirements or an airworthiness certificate, and each has its own limits; moving vehicles have conditions of their own. Over an outdoor crowd the line is sustained flight: hovering, circling or going back and forth above it counts, while a brief, one-time pass on the way elsewhere does not.

Before each flight the remote pilot must assess the area, including weather, airspace, people and property below, and other hazards, and must brief everyone involved on conditions and emergency procedures. Check that the control link works, there is enough power for the planned flight, and anything attached is secure, then confirm the aircraft is in safe condition.

FAA rules are not the only ones that apply. National park superintendents prohibit launching, landing or operating drones on park land under park regulations, and that ban sits alongside the FAA's rules rather than replacing them. Check the land manager's rules for each location, and ask whether a permit or designated area exists.

A recorded song involves two copyrights: one in the musical composition, covering the music and any lyrics, and one in the sound recording of a performance. They are separate works, so permission for one does not cover the other. A new recording you make yourself still uses the composition. This is how the rights are structured, not legal advice.

A stock shot is footage reused across productions, often bought from a stock footage company, and the purchase gives you the uses its license terms allow. Under copyright law a nonexclusive license is not a transfer of ownership, and even an exclusive one can be limited in time or place, so read the terms for where the shot may appear. Educational only, not legal advice.

Get consent from everyone who takes part on camera, in a signed letter or release, before the piece is distributed; agencies that publish video ask for it in writing for exactly that purpose. Explain the project so participants know what they are agreeing to. This describes common practice, not legal advice for a particular shoot.

Permission to film somewhere comes from whoever controls the place, and a public authority may add its own permit. National parks usually need no permit for a crew of eight or fewer using hand-carried gear in public areas, but exclusive use of a site or larger groups may need one, with location fees. Educational only, not legal advice.

Fair use is judged case by case on four factors: the purpose of the use, including whether it transforms the work; the nature of the work; how much is taken, and whether it is the heart of it; and whether the use harms the market for the original. Commentary and criticism are listed examples, yet no number of seconds is automatically safe. Educational only.

Chain of title is the record of who owns each element. A work made for hire belongs to the hiring party, and a commissioned contribution to a film qualifies only with a signed written agreement saying so. Transfers of ownership can be recorded with the Copyright Office, putting them on the public record. Educational only, not legal advice.

Keep the highest-quality version of each video with good descriptive information, and decide whether unedited footage stays as well as the final cut. If you transcode for access, keep the original file as it came. Keep what travels with the video too, such as captions, scripts and location notes, and record how each relates to it.

A format becomes unreadable when the hardware and software it needs disappear, and widely adopted formats are less likely to go quickly. Archives watch the formats they hold and plan transformations to sustainable ones while the old format can still be read. Removable media add a second risk: the carrier itself failing.

No single medium is safe, so archives spread copies across different storage environments, including one in a distant region, and verify each copy when it lands on long-term tape. Mixing media guards against hazards that hit one kind. Whatever the medium, check copies regularly and make fresh media copies on a schedule.

A checksum is a value computed from a file's contents; compute it again later and compare, and a mismatch means the file has changed. Fixity checks catch accidental errors from transfer or storage. NARA's plan audits a sample of its fixity records each year and repairs or replaces failing files, which depends on having a good copy to restore from.

Preservation metadata keeps a file's context, administrative, descriptive and technical information with it, so it can still be understood and managed later. Record technical details such as codecs, frame rate and timecode, prefer formats that can embed rich metadata, and start collecting it at the shoot rather than at the archive door.

An edit decision list records which source shot goes where, so a cut can be rebuilt from the originals. Interchange formats such as AAF carry the creative decisions about combining media, and the media they point to can sit outside the file. A project moves to another system only if its media travel with it.