Video Switching Fundamentals

Introduction

There is a moment about ninety seconds into most conference sessions when the wide shot of the stage becomes a tight shot of the speaker's face. Nobody in the room notices it. That is the point. Somebody pressed a button at exactly the right instant, and the fact that nine hundred people did not register the change is the whole measure of whether they did it well.

This is the strange economy of switching. Every other department is judged by what the audience sees. The switcher operator is judged by what the audience fails to see: the cut that landed on the word rather than after it, the graphic that cleared before it outstayed its welcome, the camera that was checked before it went to air rather than discovered live.

What this book actually does

It teaches the craft rather than a particular desk. What a switcher is for and where it sits in the chain. Why sources must agree not only on format but on TIMING, which is the half most beginners never hear about until a cut glitches. How program and preview work, and what the rows above them are. How to place a take on the right moment. What each transition says, and why the cut carries most professional shows. How keys layer a picture, where the lower third really lives, and why a chroma key is mostly a lighting job. How to work to a director. And what to do in the two seconds after something goes wrong on a machine that has no undo.

Every chapter shows you the thing on a real job, walks you through the procedure, warns you about the failure you will actually meet, and gives you something you can take to work.

Who this is for

Somebody heading for their first video call, or already doing it and wanting to be trusted with the show rather than the cable run. You do not need to have touched a switcher. You do need to be willing to prepare, because almost everything that makes switching look easy happens before the doors open.

If you have read AV Fundamentals 101 you already have signal flow and the vocabulary. Video Basics covers formats, connectors and displays, and this book begins where it ends. Neither is required; the ideas are introduced as they arrive.

What this book will not do

It will not teach you one manufacturer's desk. Button layouts and menu names change every job and learning them is an afternoon; learning what the buttons are FOR is what transfers. It will not make you a director either, though it will make you a much better one when your turn comes.

What it will do is let you walk up to an unfamiliar switcher, work out what it is in ten minutes, and cut a clean show on it. That is the skill that gets somebody booked again.

How to Use This Book

Each chapter follows the same shape, and the shape matters more than it might look. You read the ideas, then you see them play out in a real situation, then you get the procedure, then you find out what goes wrong, then you practice, then you get a checklist you can actually take to work.

1. Read the chapter body first. This is the understanding layer, and it is written to be read straight through.
2. Read On the job. The same material as a real situation on a real show, which is where it stops being abstract.
3. Work through How to do it. These are procedures you can follow on a job this week.
4. Study What goes wrong. Experienced operators are not people who never make mistakes; they are people who recognise mistakes early.
5. Do the exercise before reading the answer key. Recognition feels like learning and is not.
6. Take the checklist with you. Photograph it on your phone. Use it until you no longer need it.

Keep a notebook

One habit is worth starting today, before the first chapter: keep a small notebook and write the details in it as you meet them. This job runs on specifics you will need exactly and will not remember — which input the lectern laptop was on at that venue, which house format a particular room runs, the auto-rate that felt right on a keynote, the name of the producer who calls a clean show.

Write it by hand rather than photographing a screen. Copying a number puts it on paper; rewriting it in your own words is what puts it in your head, and deciding how to phrase it is most of the learning.

And sit behind a working operator for a show, then write down what surprised you. You will see immediately that most of the work is not pressing buttons — it is looking at the multiview, listening to the room, and cueing the next thing while the current one is still live. The desk is the easy part; the rhythm is the craft, and the rhythm is learned by watching somebody who has it.

A note on the words

The same object has different names depending on who trained the person speaking. A switcher is a vision mixer in Britain and much of the broadcast world. A dissolve is a mix. A take is a cut when the transition is instant. None of these are wrong, and an operator who bristles at the other word is telling you more about themselves than about the equipment.

Where a second name is in common use, this book gives it once and then picks one. If somebody on a job uses the other, they are talking about the same thing.

A note on safety

A gallery is one of the safer places on a show, but the rest of the chain is not. Cameras go on tripods in walkways, screens go overhead, and the power that runs all of it is the same power that runs everything else. Tape what crosses a walkway, keep clear of anything being flown, and never defeat an earth pin to chase a fault. If you are not qualified and authorised to do rigging or electrical work, do not do it.

What a Switcher Does

A switcher looks like a router with better buttons, and that comparison is where most misunderstandings start. A router moves a signal from one place to another. A switcher does that too, but it does it at an exact instant, between sources that have been made to agree with each other, and it is the agreeing that turns out to be the hard part.

Many pictures come in; one goes out. The switcher is the live decision point of the video system.

The switcher in the chain

A video switcher — also called a vision mixer, the British term you will still hear on many crews — takes many video sources and lets one operator choose which single source is sent to the output at any instant, with a controlled transition between them. The sources are whatever the show needs: several cameras, a laptop running slides, a media player rolling video, a graphics or titling system. The switcher is the one place all of those feeds come together and a decision is made about which the audience sees.

In the source-to-processing-to-destination model you learned in AV Fundamentals, the switcher is the processing stage for live video. Cameras and laptops are the sources; the projector, the LED wall, the stream, and the recorder are the destinations; and the switcher sits in the middle deciding, moment to moment, what flows from one to the other. Everything upstream exists to feed it and everything downstream simply shows whatever it sends.

This is why the switcher is the heart of a live video show. Think of a conference session: three cameras (a wide of the stage, a tight shot of the speaker, and one on the audience for reactions) plus the speaker's slides — four sources, but the audience screen and the livestream can only show one picture at a time. Every time the picture changes from the wide shot to the slides, or from the slides to the speaker's face, the switcher operator made that change happen. The show the audience experiences is, quite literally, a sequence of the operator's decisions.

That makes switching both a technical job and a creative one. Technically you are managing inputs, formats, and a piece of equipment under time pressure. Creatively you are deciding what the audience should be looking at and when — building the rhythm and story of the show in real time. Good operators hold both at once: the discipline of the machine and the eye of a storyteller.

A beginner often assumes the cameras or the screens are where a video show is 'made.' They are not — they are the ends of the chain. The show is made at the switcher. Once you see it that way, a wall of monitors and a desk full of buttons stops being intimidating and becomes what it is: the one seat where the live picture is shaped.

It is worth being precise about what the switcher is doing at the moment of a cut, because everything else in this book follows from it.

A video signal is a stream of frames arriving at a fixed rate. When you cut from camera one to camera two, the output must stop taking frames from one and start taking them from the other -- and it has to do that BETWEEN two frames, not part-way through one. If it lands mid-frame, the output has half of one picture and half of another, and what reaches the screen is a tear, a roll, or a black flash while the display re-syncs.

So a switcher is not choosing a cable. It is choosing a source at a frame boundary, and to do that it needs every source to have its frame boundaries in the same place at the same time.

That single requirement is the reason for most of the equipment in a video rack. It is why reference generators exist, why a frame synchroniser is built into every input on a modern switcher, and why "it plugged in and it looked fine on the multiview" is not the same as "it will cut cleanly". A source can look perfect and still be arriving at the wrong moment.

The rest of this chapter is about making sources agree: on format, which most people know about, and on timing, which most people do not.

Inputs and outputs

Sources connect to the switcher's inputs, almost always over SDI (on BNC, common for cameras and professional gear over long runs) or HDMI (common for laptops and consumer devices over short runs), exactly as you met them in Video Basics. Each physical input becomes a selectable source on the switcher. Part of setup is simply knowing, without hesitating, which input is which: input 1 is the wide camera, input 3 is the lectern laptop, and so on. On a busy show, fumbling for 'which button is the slides again?' is how mistakes reach the screen.

On the output side, the single most important output is the program output — the finished live feed that goes to the audience displays, the projector or LED wall, the stream encoder, and the recorder all at once. Program is the show. Most switchers also provide a preview output (the shot you are lining up next, covered in "Program, Preview & the Take"), a multiview (one screen showing all sources plus program and preview together, the operator's situational awareness), and one or more auxiliary or 'aux' outputs that can send a chosen source to a specific destination such as a stage monitor or an overflow room.

Every source feeding the switcher must agree on a common format — the resolution-and-frame-rate 'contract' from Video Basics, for example 1080p at 59.94. A switcher mixes and cuts cleanly only between sources that share that format; many switchers will scale or frame-synchronise inputs to bring them into line, but you cannot assume it. The classic beginner failure is plugging in a laptop set to the wrong resolution or refresh rate: it may glitch, flicker, or drop to black at the exact moment you take it to air. Confirm every source is in the house format during setup, long before the doors open.

Because program is what the audience sees, professionals treat it as sacred: nothing reaches program that has not been checked first, and program is kept clean and stable at all costs. Preview, multiview, and aux outputs are your workspace for getting things right; program is the published result. Keeping that line clear in your head — workspace versus published — is the foundation of calm, clean switching.

The output side deserves more than a list, because on a real show you will be asked for outputs that do not exist yet and you need to know what is possible.

PROGRAM is the show. Everything the audience sees, with every graphic, exactly as it goes out. It feeds the screens, the stream and the record, and it is the one output nobody may experiment on.

PREVIEW is your workspace, and it is private. Whatever is on it costs nothing and risks nothing, which is precisely why the discipline of using it is the foundation of the job.

MULTIVIEW is your situational awareness -- every source, plus program and preview, on one screen. It is the thing you actually look at for most of a show. Learn its layout before doors, because in a fast sequence you will be finding sources by POSITION rather than by reading labels.

AUX OUTPUTS are the ones people forget to ask for until the day. An aux sends a chosen source, independently of program, to a specific destination: a stage confidence monitor so the speaker can see their own slides, an overflow room, a translator's booth, a camera's return feed. Some are switchable during the show and some are set once and left. Knowing how many you have, and whether they follow program or stand alone, is a question to answer at setup rather than at showtime.

AND THE CLEAN FEED, which is the one that surprises people. It is program WITHOUT the graphics -- no lower thirds, no logo. A broadcaster taking your feed wants it clean so they can add their own branding; a recording made for re-editing wants it clean so the titles are not burned in permanently. On most switchers you get a clean feed by taking the output from BEFORE the downstream keyer, which is covered in the keying chapter. If somebody asks for a clean feed and you hand them program, they will discover the problem in the edit, weeks later.

Genlock, and why sources have to agree in time

Two cameras can both be producing perfect 1080p at 59.94 and still be impossible to cut between, because although they are producing the same NUMBER of frames per second they are not producing them at the same MOMENTS. One starts each frame a fraction of a second after the other. Cut between them and the switcher is asked to change source part-way through a frame.

GENLOCK is the fix, and it is old and simple. A reference generator produces a timing signal -- historically black burst, on modern high-definition systems tri-level sync -- and it is distributed to every device that will accept it. Each camera locks the start of its frames to that reference. Now every source in the building starts its frames at the same instant, the frame boundaries line up, and a cut has somewhere clean to land.

You will meet it as a BNC input on the back of a camera or a switcher labelled REF, GENLOCK or SYNC, fed from a distribution amplifier so one generator can reach a dozen devices. It is usually terminated at the last device in a chain, and a loop- through that has been left unterminated is a classic cause of an intermittent, baffling picture fault.

WHY IT STILL MATTERS when every modern switcher has frame synchronisers on its inputs: because a frame synchroniser costs you something. It has to hold a picture and re-clock it, which adds delay -- typically a frame or so per synchronised input -- and that delay is not the same on every path. Genlocked sources pass through with essentially no added delay and stay in step with each other. A show where the cameras are genlocked and only the laptops are synchronised is a show where the camera cuts are tight and nothing drifts.

It also matters because not everything HAS a frame synchroniser. Older switchers, some routers, and a lot of the small equipment that turns up on modest shows will simply glitch on an unreferenced source, and no amount of menu-diving will fix it.

The practical shape of it: on a show with proper cameras, ask whether there is a reference generator and whether the cameras are locked to it. On a show with laptops and consumer sources, accept that those will be frame-synchronised on the way in, and expect them to sit a frame or two behind the cameras.

Formats, scaling and the frame synchroniser

A format is a contract: a resolution, a frame rate, and a scan type, written the way Video Basics writes it -- 1080p59.94, 2160p30, 1080i50. Every device in the chain has to be told the same contract, and the commonest cause of a source that will not behave is that one of them was not.

WHAT A FRAME SYNCHRONISER ACTUALLY DOES is worth knowing rather than trusting. It captures an incoming frame into memory and reads it back out in step with the switcher's own timing. If the source is running slightly slow, the synchroniser will eventually have to repeat a frame; if slightly fast, it will have to drop one. On a static laptop slide nobody sees that. On motion it is a very small stutter, once in a while, and it is why a genlocked camera looks subtly better than a synchronised one.

A SCALER is a different job that often sits in the same box: it changes resolution, turning a laptop's 1920x1200 into the house 1920x1080. Scaling always costs some sharpness and it costs latency, and a source that has been scaled AND synchronised can sit several frames behind a camera that went straight in.

THE LAPTOP IS THE RECURRING PROBLEM and it is worth its own paragraph. Laptops negotiate their output with whatever they are plugged into, so a machine that was set correctly in rehearsal renegotiates when it is unplugged and replugged, or when the presenter closes the lid, or when a screensaver engages. It may come back at a different resolution, a different refresh rate, or in a mirrored mode that changes the aspect. This is why professional shows put a scaler in front of every laptop: it gives the switcher one unchanging format regardless of what the laptop decides to do.

AND THE CHECK IS ALWAYS THE SAME. During setup, look at every source on the multiview and confirm the switcher reports the format you expect on every input. Not "it looks fine" -- the actual reported format, input by input. A source running at the wrong rate can look completely normal on a still image and fall apart the moment it is taken to air.

Tally, UMD, and telling the room what is live

Tally is the light that tells a camera operator their camera is live. It is a small red lamp on the camera body, on the viewfinder, or both, and it is fed from the switcher, which knows exactly which source is on program.

It matters more than its size suggests. A camera operator who knows they are live holds the shot, keeps it steady, and does not pan away to find the next one. An operator who does not know will do all the things you do when nobody is watching: reframe, rack focus to check something, zoom out to reposition. Every one of those goes out to the audience.

It matters for the people in front of the camera too. A presenter looks at the camera with the red light on. Without tally they are guessing, and a speaker addressing a camera that is not live, while the live camera catches them in profile, is a shot nobody can rescue.

TALLY IS ALSO A SAFETY LINE FOR YOU, in a smaller way. On the multiview, the program source is normally outlined in red and preview in green. That border is how you confirm at a glance what is actually live, which matters most in exactly the moment you are least able to reason it out.

WHAT GOES WRONG is nearly always wiring or configuration. Tally is often carried on its own cable or over the camera control system, so a camera that has been moved or repatched can end up with the tally of a different input. That is worse than no tally at all, because now two people are confidently wrong. Check it at setup by taking each camera to program in turn and having somebody confirm the right light came on.

PARALLEL TALLY is the simple form. The switcher closes a contact -- a GPO, a general-purpose output -- and that contact drives a lamp. One circuit per camera, red for on air and often green for preview. It is robust, it is easy to fault-find with a meter, and it can do exactly one thing: turn a light on.

SERIAL TALLY sends data instead of closing a contact, and the common protocols are TSL's UMD v3.1 and the newer v5.0. Because it is data rather than a switch it can carry more than on and off: which source, what text, what colour.

WHICH BRINGS IN UMD -- the UNDER MONITOR DISPLAY. It is the label strip under each window of a multiviewer, showing what that window is currently carrying. On a show where sources get reassigned it is the difference between a multiview you can read and a grid of pictures you have to memorise, and it is driven by the same serial tally data.

WHY THIS MATTERS WHEN YOU ARE ASKED. "Do you have tally?" has two different answers. If the rig is parallel you can light lamps and nothing else, so a request for named labels on the multiview is a no. If it is serial over TSL, labels are available and it becomes a configuration question rather than a cabling one. Knowing which you have before somebody asks is the whole point of knowing the words.

SDI, and the move to IP

Everything so far has assumed a cable per source: SDI on coax, one picture down one wire. That is still the commonest arrangement and will be for years. But the transport underneath live video is changing, and an operator who does not know the vocabulary will be lost the first time they walk into a facility that has moved.

SDI IS THE BASELINE. One coaxial cable, one uncompressed picture, essentially no latency, and a fault you can find with your hands. Its limit is that it is one cable per source, so a large facility becomes a very large amount of copper and a router to match.

NDI IS THE SOFTWARE-FRIENDLY ANSWER. It is compressed, it runs over ordinary Ethernet, and it moves many sources over one network at modest bandwidth. Its strength is interoperability: software tools speak it to each other easily, which is why it turns up constantly in streaming, in smaller productions, and anywhere a laptop needs to be a source. The trade is compression and some latency.

SMPTE ST 2110 IS THE BROADCAST-GRADE ANSWER, and it is a different animal. Uncompressed, and it carries video, audio and ancillary data as SEPARATE streams rather than one bundled signal -- which is powerful, because a device can subscribe to just the audio, or just the video, without taking the rest.

AND IT REPLACES GENLOCK WITH PTP. Precision Time Protocol distributes time over the same network rather than sending a reference signal down its own coax. The requirement has not changed -- everything still has to agree about when a frame starts, for all the reasons in this chapter -- but the mechanism has, and a facility on PTP has no black burst generator to look for.

SRT IS THE ONE THAT GOES BETWEEN BUILDINGS. Where NDI and 2110 are for inside a facility, SRT is built to carry a stream across the public internet and survive the packet loss that implies.

WHAT ACTUALLY HAPPENS ON SHOWS IS A MIXTURE, and expecting the mixture is the useful part. A facility may run 2110 as its core, NDI for software and portable kit, SRT for a feed arriving from another city, and SDI for the last few metres to a monitor. Switchers increasingly ship with IP connectivity built in rather than as an option card.

WHAT DOES NOT CHANGE is everything else in this book. Program and preview, the take, transitions, keys, checking before you commit -- all of it is the same whether the picture arrived on coax or as packets. The transport changed; the craft did not.

A video switcher (vision mixer) takes many sources and sends exactly one to the output at a time, with a controlled transition — the processing stage of the live video chain, and the heart of the show. Sources arrive on inputs over SDI or HDMI; the key output is program (the live feed to audience, stream, and record), supported by preview, multiview, and aux outputs. Every source must share the house format or it will not switch cleanly. Program is sacred: only checked sources reach it, and it stays clean and stable.

The outputs, and who each one is for

Ask which of these the show needs at setup rather than during it. The one people forget is the clean feed, and it is the one that cannot be recovered afterwards.

Why a source will not cut cleanly

Format and timing are different problems with different fixes, and confusing them is what makes this hard to diagnose under pressure.

On the job

A corporate session, three cameras and a lectern laptop. Setup went quickly, everything looks right on the multiview, and in rehearsal you cut between the cameras and it is clean.

Then you take the laptop, and the output flashes black for half a second before the slide appears.

The instinct is to blame the laptop's resolution, and here it is not that -- the switcher reports 1080p59.94 on that input, the same as everything else. The format contract is satisfied.

What is failing is timing. The cameras are genlocked to the house reference and start their frames together. The laptop is not, and cannot be. Its frames arrive whenever the graphics card feels like sending them, so the switcher has to catch and re-clock them, and this particular input is not doing that -- frame synchronisation is switched off on it, or that input does not have one.

The fix is either to turn the input's frame synchroniser on, or to put a scaler in front of the laptop that outputs a locked, correctly timed signal. Either way the laptop will now sit a frame or two behind the cameras, which is fine for slides and would matter if it carried audio.

What you must not do is spend the rehearsal changing the laptop's resolution, because the resolution was never wrong.

How to do it

Bringing a switcher up from cold. Done in this order, because each step assumes the one before it is settled.

1. Agree the house format before anything is plugged in — resolution, frame rate, scan type — and write it where everyone can see it.
2. Find the reference generator and confirm what is locked to it. On most shows that is the cameras; laptops and consumer sources will not be.
3. Terminate the end of the reference chain. An unterminated loop-through causes intermittent faults that look like anything but a reference problem.
4. Patch sources in, then check the switcher's REPORTED format on every input one at a time. Not how it looks — what the switcher says it is.
5. Put a scaler in front of anything that renegotiates its output, which means every laptop.
6. Take each source to program in rehearsal and watch the moment of the cut, not the picture afterwards. A timing fault only shows at the transition.
7. Check tally: take each camera to program in turn and have somebody confirm the right light lit.
8. Confirm which outputs the show needs, including whether anybody wants a clean feed, and route the aux outputs before doors.

What goes wrong

Practice it

Six questions from a first video call.

Program, Preview & the Take

Two rows of buttons: what is live, and what is next. Almost everything that makes switching safe comes from keeping those two ideas separate in your head, and almost everything that goes wrong comes from letting them blur.

The core of switching is two rows of buttons: what is live, and what is next.

Program and preview

Almost every switcher is built around two main buses — a 'bus' simply being a row of source-selection buttons, one button per input. The program bus shows what is currently live: press a button on the program bus and that source is on air immediately (a hard cut). The preview bus shows what you have lined up to go next: selecting a source on preview changes nothing the audience sees — it just cues that shot on your preview monitor so you can confirm it before committing.

This two-bus design is the whole secret of clean switching. It gives you a private workspace (preview) separate from the published output (program). You can check that camera 2 is actually framed and in focus, that the laptop is on the right slide, that the graphic is correct — all on preview, with the audience none the wiser — and only then send it to air. Switching without preview, by punching sources straight to program, is how unchecked mistakes end up in front of everyone.

Thinking in program-and-preview is the mental model you carry for the rest of your switching life. At every moment you should be able to answer two questions instantly: what is live right now (program), and what am I preparing to show next (preview)? Hold those two thoughts and the desk stays calm. Lose track of them and even a simple show becomes a scramble.

The habit that separates professionals from beginners is being one step ahead. While the wide shot is live on program, the good operator has already cued the speaker's tight shot on preview, checked it, and is simply waiting for the right moment to take it. The beginner waits until they need the next shot and only then starts hunting for it — always a beat behind the show. Anticipation, not reflexes, is what makes switching look effortless.

The word BUS is worth pinning down, because it is used constantly and rarely explained. A bus is a row of source- selection buttons where exactly one is active at a time. The switcher has several, they all list the same sources, and what distinguishes them is what happens to the source you pick.

Pick on the PROGRAM bus and it goes to air instantly. There is no confirmation, no preview, no undo -- the button IS the cut. This is why experienced operators mostly do not touch the program bus during a show. It is a hard cut with no safety on it, useful when you need to be somewhere immediately and dangerous the rest of the time.

Pick on the PREVIEW bus and nothing reaches the audience. The source appears on your preview monitor and gets a green border on the multiview, and it sits there until you take it. Every mistake you make on the preview bus is free.

That asymmetry is the whole design. One bus is a workspace where being wrong costs nothing; the other is a published output where being wrong costs the show. An operator who works through preview has bought themselves a check before every decision, for the price of half a second.

THE HABIT TO BUILD, and it is worth being deliberate about early because it is hard to add later: the moment you take a shot, cue the next one. Not when you need it -- immediately, while the current shot is still live and you have time. An operator who is always one shot ahead looks unhurried even in a fast sequence. An operator who cues on demand is permanently half a beat behind, and it shows in every take.

The take

The take is the moment you send preview to program — making your cued next shot the live shot. The simplest take is a cut: an instant switch, performed with a button usually labelled CUT or TAKE, or by pressing the source directly on the program bus. The take can also be a timed transition — a dissolve or a wipe ("Transitions") — driven by an AUTO button or a T-bar. Whatever the type, the take is the instant the audience sees the change, so it is never random; it is timed to the show.

Timing the take is what turns mechanical switching into good switching. You cut to the speaker as they begin to talk, not three seconds after; you take the wide shot on the musical downbeat; you bring up the slides exactly as the presenter says 'as you can see here.' The same cut, a half-second early or late, is the difference between a show that feels sharp and one that feels loose. The operator is reading the show continuously and placing each take on the right moment.

The discipline underneath all of it is simple and absolute: check preview, then take. Never send a source to program that you have not first confirmed on preview. It feels obvious until the pressure is on and you are tempted to punch a shot you 'know' is fine — and that is exactly when a camera is mis-framed, a laptop has gone to a screensaver, or a feed has dropped. The half-second it costs to glance at preview is the cheapest insurance in live production.

Recovery is part of the craft too. If you do take a bad shot, you do not freeze or panic-mash buttons — you calmly cut back to a known-good source (often the wide shot, your safety shot) and reset. A clean recovery the audience barely notices is a sign of a seasoned operator; visible flailing is the opposite.

There are more ways to take than most people realise, and knowing which is which stops you reaching for the wrong one under pressure.

CUT or TAKE performs an instant change from preview to program and swaps the two buses -- what was live is now cued, what was cued is now live. That swap is useful and worth internalising: after a cut, your old shot is sitting on preview, which means getting back is one more press.

AUTO performs the currently selected transition over the preset duration. Same swap, but timed.

THE T-BAR does the same thing under your hand, and it is the only control that lets you stop half-way, holding both sources on screen at once. That is occasionally what you want and usually what you have done by accident.

PRESSING A SOURCE DIRECTLY ON THE PROGRAM BUS cuts to it immediately without involving preview at all. It is the fastest route to any source and it skips every check, which is exactly the trade.

AND THE ONE THAT CATCHES PEOPLE: on many desks the transition you last selected stays selected. If somebody set a two-second wipe during rehearsal and nobody reset it, your next AUTO is a two-second wipe in the middle of a keynote. Look at what is armed before you use it, and make a habit of returning the desk to a cut with a short auto-rate as its resting state.

M/E rows, and what a mix/effects bank actually is

On a small switcher, program and preview are all there is. On a larger one there are extra pairs of buses stacked above them, labelled M/E 1, M/E 2 and so on, and a beginner looking at that desk usually assumes it is more of the same. It is not, and the difference is the most useful thing you can understand about bigger desks.

AN M/E -- MIX/EFFECTS -- IS A COMPLETE LITTLE SWITCHER. It has its own program and preview buses, its own transition controls and its own keyers, and it produces one output. What makes it powerful is that its output is available as a SOURCE to the buses above it.

So you can build a picture inside M/E 1: a camera with a presenter keyed onto a background, transitions and all. Then, on the main program bus, the whole of M/E 1 is a single selectable source. You cut to "M/E 1" and the audience sees the composite you built, exactly as though it were a camera.

WHAT THAT BUYS YOU is preparation. Anything you assemble inside an M/E is invisible to the audience until you take it, which means you can build a complicated shot -- a picture-in-picture with a graphic over it, framed and positioned -- live, during the show, while something else is on air. Then you take it in one clean move.

WITHOUT AN M/E you would have to build that shot on program, where every adjustment is public. With one, the whole assembly happens off-air.

THE MENTAL MODEL is a stack, and it always runs the same way: lower banks feed higher ones, and the main program bus is at the top of the stack. If you can hold that, an intimidating desk becomes readable in a couple of minutes, whoever made it.

AND THE VOCABULARY MATTERS because a director will use it. "Put the PiP in M/E 1 and give me M/E 1 on preview" is an ordinary instruction and it needs no translation once you know what an M/E is.

The safety shot, and designing a show you can recover

Every experienced operator has a shot they can cut to when they do not know what else to do. It is usually the wide of the stage. It is always framed so that nothing embarrassing can happen in it, it is always live and checked, and its whole purpose is to be the answer to a question you have not got time to think about.

WHY A WIDE SHOT: because it contains the event. If a speaker moves, the wide still has them. If a camera operator is repositioning, the wide is unaffected. If you have entirely lost track of what is happening, the wide is never actively wrong -- it is just not the most interesting choice, and not-interesting is a survivable state where wrong is not.

KEEP IT AVAILABLE. On a busy desk it is worth knowing the safety shot's button position by feel, so you can reach it without reading. Some operators keep it on a fixed input number across every show they do for exactly this reason.

AND DESIGN THE SHOW SO RECOVERY IS POSSIBLE. That is a real thing you can influence at setup: ask for one camera that is never repositioned mid-show, so there is always one shot you can trust. A show where every camera is doing something dynamic is a show with no floor under it.

THE RECOVERY ITSELF is a rhythm rather than a decision. Cut to the safety shot. Breathe. Find your place in the run of show. Cue the next real shot on preview, check it, take it. That whole sequence takes perhaps four seconds and the audience registers a slightly long wide shot, which is nothing.

What they register instead, if you improvise, is a sequence of half-checked shots going past while somebody hunts. Panic is visible in a way that a wide shot never is.

Switchers are built around two buses: program (live now) and preview (cued next). Preview is your private workspace; program is the published show. The operator cues the next source on preview, checks it, then takes it to program — a cut (instant) or a timed transition — placed on the right moment of the show. Stay one step ahead by cueing while the current shot is live. Check preview, then take; if a bad shot reaches air, calmly cut to a safe shot and reset.

The buses and what they do

Same rows of buttons, different consequences. The distinction that matters is whether being wrong is free.

The take controls

Know which is armed before you use it. A wipe somebody selected in rehearsal is still selected at showtime unless somebody reset it.

On the job

A three-camera panel discussion, and the director wants a picture-in-picture: the moderator small in the corner while a panellist speaks full-frame.

A beginner builds it on program. They turn on the PiP, and the audience watches the box appear in the wrong corner, get dragged across the screen, resize twice and settle. Everything the operator did while finding the right position went out live.

What the desk is for is doing that off-air. You select the panellist on M/E 1's program bus and the moderator into M/E 1's keyer, position the box, size it, and check it -- all of which is invisible, because M/E 1 is not on air. The audience is still watching the wide shot from the main program bus.

Then, when the director calls it, you cue M/E 1 on preview, confirm it looks right, and take it. One clean transition into a finished picture.

That is the whole reason mix/effects banks exist. An M/E is a rehearsal room you can use during the performance, and the difference between an operator who knows that and one who does not is the difference between a composite that appears and a composite the audience watches being assembled.

How to do it

The switching cycle. This is the loop you run for the whole show, and it should become automatic enough that you can do it while listening to a director.

1. Know what is live. Look at the red border on the multiview, not at your memory of what you pressed.
2. Select the next shot on the PREVIEW bus. Never the program bus unless you mean to cut instantly.
3. Check it properly — framing, focus, the right slide, the right graphic. This is the step the whole design exists to give you.
4. Wait for the moment. The take is placed on a word, a beat or a director's call, not on the moment you happen to be ready.
5. Take it — CUT for instant, AUTO or the T-bar for a timed transition. Confirm on the multiview that the borders swapped.
6. Immediately cue the shot after that on preview, while the new one is live. This is what keeps you a step ahead.
7. Build anything complicated inside an M/E, off air, and take the finished M/E as a single source.
8. If you lose the thread: cut to the safety shot, breathe, find your place in the run of show, and restart the loop.

What goes wrong

Practice it

Six about the two buses and the banks above them.

Transitions

A transition is a sentence break. A cut is a full stop, a dissolve is a paragraph, and a wipe is an exclamation mark -- which is why you use one of them constantly and the other two on purpose.

How you change from one picture to the next sets the feel of the show.

The three core transitions

There are three transitions you will use constantly, and almost every show is built from them. A cut is an instant change from one source to the next — no in-between, just one frame the wide shot and the next frame the speaker. It is clean, fast, energetic, and by far the most common transition in most live work. A dissolve (also called a mix or crossfade) fades one source into the next over a set duration, the two pictures briefly overlapping. It feels softer and more deliberate, and it signals a change of pace or place. A wipe replaces one source with the next using a moving edge or shape — a line sweeping across, a box growing — which is more stylised and draws attention to itself.

Each transition carries a different emotional message, so choosing one is partly a creative act. A cut says 'keep up, here's the next thing.' A dissolve says 'settle, we're shifting gently.' A wipe says 'notice this change.' Matching that message to the moment is the craft: a fast concert lives almost entirely on cuts timed to the music; a reflective interview segment might dissolve between speakers; a sports show might wipe into a replay to make the shift obvious.

As a beginner, make the cut your foundation. Clean, well-timed cuts carry the overwhelming majority of professional shows, and a show cut well with nothing but cuts looks polished. The dissolve is your reliable second tool for softer moments. Treat wipes and fancier digital effects as seasoning, not a main course — an operator who wipes between every shot looks like someone showing off the switcher's features rather than serving the show. When in doubt, cut.

A useful rule of thumb: let the content choose the transition, not the other way around. If you find yourself reaching for an effect because it is 'cool,' stop and ask what the moment actually needs. Usually the answer is a well-timed cut.

It is worth saying plainly what a beginner is not told: the cut is not the beginner's transition that you graduate from. It is what professionals use for the overwhelming majority of every show they cut, and the instinct to use something more interesting is the single most reliable sign of inexperience.

WHY THE CUT WORKS is that it matches how attention already behaves. Look across a room and your eye jumps -- it does not slide, and it does not fade. A cut is the visual grammar closest to that, so it costs the audience nothing to process. They do not think "the picture changed", they simply look at the new thing.

A DISSOLVE COSTS THEM SOMETHING, and that cost is the point. During the overlap the audience is briefly aware of both pictures, which reads as time passing or place changing. Used where the show has genuinely changed gear, that is exactly right. Used between two cameras pointed at the same speaker, it is a small confusion for no reason.

A WIPE COSTS MORE AGAIN because the moving edge is itself a thing to look at. There are shows where that is correct -- sport going to a replay wants you to know it is a replay -- and a great many where it is somebody demonstrating that the desk has wipes.

A USEFUL TEST before any transition that is not a cut: what am I telling the audience by not cutting? If there is an answer -- time has passed, we have changed location, this is a replay -- use it. If the answer is "it looked nice", cut.

Timing and the T-bar

Any transition that is not an instant cut has a duration, and the operator controls it. Two controls do this. The AUTO button performs the selected transition over a preset time — set it to, say, one second and every auto-transition takes one second, consistent and repeatable. The T-bar is a physical lever the operator pushes by hand: the transition follows the lever, so how fast you move it sets how fast the dissolve or wipe happens, and you can even hold it halfway to park both sources on screen at once. AUTO gives consistency; the T-bar gives feel and manual control.

Learning to time transitions is what makes switching feel musical rather than mechanical. A quick quarter-second dissolve reads almost like a soft cut; a slow three-second dissolve feels dreamy; the same dissolve started a beat too early or too late breaks the spell. Operators talk about transitions in the language of rhythm because that is what good timing is — placing the change where the show wants it, the way a drummer places a hit.

Timing is, in the end, the whole game in live switching. A cut a half-second late on a music accent, a dissolve that drags while the audience waits, a take during the wrong word of a sentence — each one quietly tells the audience that someone is operating the show. The goal is the opposite: changes that land so naturally nobody registers them as changes at all. That is why switching is performed live, in the moment, like playing an instrument, rather than set up once and left alone.

Practically, this means an operator watches and listens to the show constantly — the program monitor, the preview, the speaker, the music, the director's voice on comms — and anticipates the moment a beat before it arrives, so the take lands exactly on time rather than just after it.

Timing is learned rather than known, but there are a few specifics that shorten the learning considerably.

CUT ON THE MOTION, not after it. When somebody stands, sits, turns or gestures, the cut placed during the movement is almost invisible, because the audience's attention is already moving. The same cut a second later, once everything has settled, is conspicuous.

CUT ON THE WORD, not on the thought. If a presenter says "as you can see here", the slide is up on "see", not after "here". You are always slightly ahead of the meaning, because the audience needs a fraction of a second to take in a new picture before the words land on it.

LET A SHOT BREATHE. A common beginner fault is cutting too often, on the theory that variety is engagement. It is not. A shot needs long enough for the audience to read it -- roughly three seconds as a floor for anything with information in it, and much longer for a speaker who is making a point. Cutting every two seconds is exhausting to watch and it makes the show feel nervous.

MATCH THE TRANSITION SPEED TO THE SHOW. A quarter-second dissolve reads almost as a soft cut and is useful when a hard cut feels abrupt. A one-second dissolve is the standard, useful default. Three seconds is dreamy and belongs to a slow moment. Setting your AUTO rate to something sensible at the start of the show, and leaving it, removes one decision from a busy night.

AND THE T-BAR IS FOR FEEL. Its value is that a human hand can accelerate and decelerate in a way a preset cannot, so a dissolve can start slowly and finish quickly to land on a beat. That is a real skill and it is worth practising -- but not during a show you cannot afford to get wrong. AUTO is repeatable and repeatable is what a nervous night needs.

Fade to black, and the transitions that are not between sources

Not every transition goes from one source to another. Two of the most useful go to nothing at all, and beginners often do not know they exist until they badly need one.

FADE TO BLACK -- FTB on most desks -- takes the entire output to black, graphics and all, regardless of what is on program. It is the end-of-segment transition: the applause finishes, you fade to black, and the room resets. It is also the single most useful button on the desk when something has gone properly wrong, because black is the only picture that is never embarrassing.

Learn where it is by feel on every desk you sit at, before the show starts. It is the one control you may need to find without looking.

CUT TO BLACK is the same destination arrived at hard, and it is almost always wrong on a live show. A hard black reads as a fault -- the audience assumes the system has failed. A fade of half a second reads as a decision.

AND THE THING PEOPLE FORGET: fade to black also kills the stream's picture and, on many desks, does nothing to the audio. A stream audience staring at black while a speaker keeps talking has no way to know the show is fine. If you are going to hold black for more than a moment, somebody needs to have thought about what the stream sees -- often a holding slide rather than black.

STINGERS AND DVE TRANSITIONS exist too: an animated clip that plays over the change, or a digital effect that pushes one picture off while another arrives. They belong to shows with a designed look and a rehearsal to get them right. On a corporate session they are almost always the wrong instinct, and they consume a key layer and a media player that something else probably needs.

Cutting to music, and cutting to speech

The two most common kinds of show ask for opposite instincts, and an operator who cuts a conference the way they cut a band will get it wrong in a specific, visible way.

MUSIC HAS A GRID. There is a tempo, there are bars, and there are moments -- a downbeat, a chorus arriving, a drum fill ending. Cuts placed on those moments feel intentional, and cuts placed between them feel accidental, even to an audience with no musical training. So on music you are counting: you find the pulse, you anticipate the beat, and you press slightly BEFORE it so the picture lands on it. Reaction time is real; a cut you start on the beat arrives late.

A concert is also the one place a fast cutting rate is correct. Energy is the content, and a chorus can carry a cut every two bars without feeling nervous.

SPEECH HAS NO GRID, it has meaning. The moments are the start of a sentence, the arrival of a new idea, a gesture, a laugh, a question from the floor. Cutting on those feels like listening; cutting mid-clause feels like interrupting.

And the cutting rate is far slower. A speaker making an argument needs to be on screen long enough for the audience to attach the words to the face. Ten seconds on one shot is entirely normal and often right, where in a concert it would feel dead.

THE SHARED RULE UNDER BOTH is that you are cutting to what the audience is about to want to look at, not to what you have just noticed. That is why anticipation matters more than reflexes, and why an operator who knows the run of show cuts better than one who is merely quick.

Three core transitions: the cut (instant, energetic, most common), the dissolve (a timed fade, soft and deliberate), and the wipe (a moving shape, stylised, used sparingly). Each sends a different feel; let the content choose the transition, and when in doubt, cut. Timed transitions are driven by an AUTO button (consistent preset time) or a T-bar (manual feel). Timing is everything — watch and listen, anticipate the beat, and land each take on the right moment so the switching stays invisible.

What each transition says

Choosing a transition is a small piece of storytelling. The last column is the one worth remembering.

Placing a transition

Timing is learned by doing, but these are the specifics that shorten the learning.

On the job

A corporate awards evening, and it opens with a three-piece band before the host comes on.

For the band you cut on the music. You find the pulse in the first eight bars and then you are counting: wide on the intro, drummer on the fill, guitarist as the solo starts, back wide for the chorus. Roughly a cut every couple of bars, each one pressed a fraction before the beat so the picture lands on it. It feels fast and it looks right, because energy is what the audience is there for.

Then the host walks on and the whole grid disappears.

An operator who keeps that rate is now cutting a speech every two seconds, and it reads as panic. The show has changed instrument and so must you: a wide as the host arrives, then a tight shot once they are settled and speaking, and then you leave it there. Ten seconds. Fifteen. You cut when they gesture, when they turn to the screen, when they ask the room a question — on meaning rather than on tempo.

The mistake is not knowing the cuts. It is not noticing the show changed what it wanted, and the moment it changes is exactly when the audience is most likely to notice you.

How to do it

Choosing and placing a transition. In practice this runs in under a second, but it is worth being able to say it out loud while learning.

1. Default to a cut. It is what carries most of every professional show, and it costs the audience nothing to read.
2. Before choosing anything else, ask what you would be telling them by not cutting. If there is no answer, cut.
3. Set your AUTO rate once at the top of the show — around a second is a useful default — and leave it, so it is one less decision later.
4. Confirm which transition is armed before you use AUTO or the T-bar. Whatever was last selected is still selected.
5. Find the moment. On music, the beat; on speech, the start of a sentence or a movement.
6. Press slightly early. Reaction time means a take started on the beat arrives after it.
7. Let the new shot breathe — three seconds minimum for anything with information, much longer for a speaker.
8. For an end of segment, fade to black over about half a second, and know what the stream is seeing while you hold it.

What goes wrong

Practice it

Six about choosing and placing.

Keying & Layers

Everything so far has treated the switcher as choosing one picture. It also builds pictures, by laying sources on top of each other, and the difference between a bare camera feed and something that looks like television is almost entirely what is layered over it.

Switchers do more than choose one source — they layer sources together.

What a key is

So far we have treated the switcher as choosing one source at a time, but it also composites — it layers sources on top of each other. The tool for this is the key. A key makes part of an upper layer transparent so the layer beneath shows through, exactly like a sticker with a transparent background placed over a photo. Keys are how a switcher puts a name graphic over a live camera, a logo in the corner, or a presenter standing in front of a virtual background. Modern switchers offer several key layers, so you might stack a background, a presenter, and a lower-third name all at once to build a finished picture.

The three keys you will meet most often each solve a common need. A lower third is a name-and-title graphic that sits across the lower portion of the screen to identify whoever is speaking ('Jordan Lee — Production Manager'). A picture-in-picture (PiP) places one source in a box inside another — a presenter beside their slides, or a sign-language interpreter in the corner. A chroma key replaces a single solid colour, almost always green, with another source — the green-screen effect that drops a presenter onto a graphic background or a weather map.

Two technologies sit behind most keys. A luma key keys out based on brightness and a chroma key keys out based on colour; graphics systems usually also provide a separate transparency channel called an alpha or 'fill and key,' which gives the cleanest edges for titles and logos. You do not need the deep theory yet — what matters now is the concept that a key is a transparency-based layer, and that a switcher can hold several of them at once.

Keying is what turns a bare camera feed into something that looks like television. Without it you have raw shots; with it you have named speakers, branded corners, presenters composited onto designed backgrounds — the polished, finished look audiences expect.

A key needs two things and it is worth naming them, because every keying problem you will ever meet is one of the two going wrong.

THE FILL is the picture you want to see -- the graphic, the logo, the presenter. THE KEY (sometimes called the matte or the alpha) is the instruction that says which parts of the fill are solid and which are transparent. The switcher combines them: wherever the key says opaque, the fill is drawn; wherever it says transparent, whatever is underneath shows through.

HOW THE KEY SIGNAL IS DERIVED is what separates the types.

A LUMA KEY derives it from brightness. Anything darker than a threshold becomes transparent. It is crude, it is instant, and it works well for white text on a black background -- which is exactly why it survives on shows where somebody has a graphic and no alpha channel.

A CHROMA KEY derives it from colour. A chosen hue, almost always green, becomes transparent. This is the green-screen effect and it is the one that depends most on things outside the switcher.

A LINEAR OR ALPHA KEY does not derive it at all -- the graphics system sends the key as a SEPARATE SIGNAL alongside the fill. Two cables, "fill and key". Because the transparency was authored rather than guessed, the edges are clean and can be partially transparent, which is what makes properly soft edges and drop shadows possible.

THAT LAST ONE IS THE PROFESSIONAL ANSWER for titles and logos, and it is why a graphics playout machine has two outputs. If somebody hands you a graphics source with two SDI cables and says "fill and key", they are telling you which input pairs go together on the switcher.

Using keys cleanly

Keys are powerful, and precisely because they are powerful they are easy to overuse. The professional approach is purposeful and restrained. A lower third comes on to identify a speaker, stays just long enough to be read, and clears — it does not live on screen the whole show. A picture-in-picture earns its place by showing two genuinely relevant things at once, not by cramming the screen. A chroma key is set up with even, well-lit green-screen lighting so the edges are crisp. A screen cluttered with permanent graphics, or a ragged chroma key with green fringing around the subject, looks worse than a clean single camera.

Most key quality is won before the show, not during it. Chroma key in particular is really a lighting job: light the green screen evenly with no shadows or hot-spots, keep your subject far enough off the screen to avoid green spill on their shoulders, and the key practically sets itself. A badly lit screen cannot be rescued by twiddling the key controls — garbage in, garbage out. The same is true of graphics: build and load the lower thirds and set the PiP positions in advance.

Operationally, then, the pattern mirrors transitions: prepare in advance, execute on time. With graphics built, positions set, and the chroma key tuned during setup, the live job becomes simply bringing each key on and off at the right moment — the speaker's name appears as they start, and clears before it overstays. Calm live execution is the reward for thorough preparation.

A simple discipline keeps it clean: every graphic that comes on must have a planned moment to come off. If you cannot say when a key will clear, it probably should not go on yet.

A few specifics turn "use keys purposefully" into something you can actually check.

KNOW WHEN IT COMES OFF BEFORE IT GOES ON. This is the single discipline that keeps a screen clean. A lower third is up for long enough to read a name twice -- call it five to eight seconds -- and then it clears. A graphic with no planned exit becomes permanent by accident, and a permanent lower third is the clearest sign nobody is minding the show.

BUILD AND POSITION EVERYTHING COLD. The positions of a picture-in-picture, the size of the inset, which corner it sits in, the exact file loaded into each key -- all of that is setup work. During the show your job is on and off at the right moment, and nothing else.

WATCH WHAT IS UNDERNEATH. A lower third designed against a wide shot can be illegible against a tight one, because the background moved from a dark stage to a bright shirt. If a key lives over changing pictures, check it over the worst of them, not the best.

AND KEYS COST SOMETHING. Each key layer is a resource, and a switcher has a finite number. An operator who has spent every upstream keyer on decoration has none left when the show genuinely needs a picture-in-picture. Spend them the way you would spend channels.

The downstream keyer, and where the lower third really lives

There are two places a key can sit, and knowing which is which is the difference between a lower third that behaves and one that does something baffling at every transition.

AN UPSTREAM KEYER lives INSIDE an M/E, before the transition. It is part of the picture that bank produces, which means it transitions along with everything else -- dissolve from that M/E to something else and the key dissolves with it. That is what you want for anything that belongs to a particular shot: a picture-in-picture, a chroma-keyed presenter, a virtual background.

A DOWNSTREAM KEYER -- the DSK -- lives AFTER everything else, on the final program output. It sits on top of whatever program is showing, and it is unaffected by cuts and transitions underneath it. Cut from camera one to camera two and the DSK stays exactly where it is.

THAT IS WHY LOWER THIRDS LIVE IN THE DSK. A name graphic identifying the person speaking should not flicker or dissolve every time you change camera. It should sit there calmly while the pictures change beneath it, and then be faded off when the name has been read. The DSK does exactly that, and putting the same graphic in an upstream keyer produces a lower third that vanishes and reappears at every cut.

IT IS ALSO WHY THE CLEAN FEED COMES FROM BEFORE THE DSK. Since the downstream keyer is the last thing applied, an output tapped before it is program with all the switching and none of the branding -- which is precisely what a broadcaster or a re-editable record wants.

THE PRACTICAL RULE, and it survives every desk: if the graphic belongs to ONE SHOT, key it upstream. If it belongs to THE SHOW, key it downstream.

And know how many DSKs you have. Two is common -- typically one for lower thirds and one for a permanent logo bug -- and running out mid-show is a problem with no elegant answer.

Chroma key is a lighting job

More chroma keys are ruined before the switcher is switched on than by anything anybody does at the desk, and an operator who understands that will fix problems the settings cannot.

WHAT THE SWITCHER IS ACTUALLY DOING is looking for one colour and making it transparent. It can only do that well if the colour is genuinely one colour. A green screen lit unevenly is not one green -- it is a bright green in the middle and a darker green at the edges, and no single threshold covers both. Pull the key hard enough to remove the dark corners and you start eating into the subject; pull it gently enough to protect the subject and the corners stay.

SO THE FIRST FIX IS ALWAYS LIGHT THE SCREEN EVENLY, separately from the subject, with no hot-spots and no shadows. That is a lighting department job and it is worth asking for by name.

THE SECOND IS DISTANCE. A subject standing close to the screen bounces green light back onto their own shoulders and hair -- spill -- and now the thing you are keying out is also on the person you are keying in. A metre and a half of separation removes most of it for free, and it also lets you light the subject without spilling onto the screen.

THIRD, LIGHT THE SUBJECT TO MATCH THE BACKGROUND they will appear on. A presenter lit flat and dropped onto a dramatic evening scene reads as a cut-out, however clean the key is.

WHAT THE DESK CAN STILL DO, once the physical job is right: SPILL SUPPRESSION removes the residual green cast from edges, and a GARBAGE MATTE crops away everything outside the screen so that light stands, the studio floor and the edge of the cyclorama never have to be keyed at all. The garbage matte is the most underused control in this whole area and it solves a great many "the key is picking up the room" problems in one move.

AND WHY GREEN. Human skin has very little green in it, so keying green takes less of the person with it than keying blue would. Blue is still used where a subject must wear green, which is the one situation that decides it.

Safe areas, and a graphic that survives every screen

A graphic that looks right on your monitor can be cropped on the hall screens, cut off on the stream, and unreadable in the overflow room. Safe areas are the old convention that stops that, and they are still worth respecting even though the reason they were invented has largely gone.

TITLE SAFE is roughly the inner eighty percent of the frame. Any text placed inside it will survive cropping, overscan, and the assorted things different displays do to the edges of a picture. ACTION SAFE is a little wider, around ninety percent, and is where important action should stay.

THE ORIGINAL REASON was that cathode-ray televisions overscanned -- they deliberately displayed slightly more than the frame, so the edges fell off the tube. Modern flat panels mostly do not. But the problem did not go away, it changed shape: a stream is re-encoded and sometimes cropped, a projector is often keystoned or slightly misaligned, an LED wall may be a different aspect ratio from the source entirely, and a hall screen frequently has a physical mask over its edges.

SO THE RULE STANDS. Text goes inside title safe. Anything that must be seen goes inside action safe. Most switchers and graphics systems can display safe-area markers on the multiview, and turning them on during setup takes seconds.

THE OTHER HALF IS SIZE. A lower third that is legible on a 27-inch monitor a metre away may be unreadable on a phone watching the stream, and the stream is often the larger audience. When in doubt, bigger and simpler, with real contrast between the text and whatever might end up behind it.

AND CHECK IT ON THE REAL DESTINATION. Look at the actual hall screen, the actual stream, and the overflow feed if there is one. A graphic approved only on the operator's monitor has been approved on the one screen nobody in the audience is watching.

A key layers one source over another by making part of it transparent (by brightness, colour, or a dedicated alpha channel). The three you will use most: the lower third (speaker's name/title), picture-in-picture (a source within another), and chroma key (green-screen replacement). Use keys purposefully, and remember chroma key quality is mostly even green-screen lighting, won before the show. Prepare graphics and positions in advance and bring each key on and off on time — and give every graphic a planned moment to clear.

The keys you will meet

Which type, what it needs, and the thing that goes wrong with it. Fill is the picture; key is the instruction about what is transparent.

Upstream or downstream

The rule that survives every desk: if the graphic belongs to one shot, key it upstream. If it belongs to the show, key it downstream.

On the job

A panel session. The client has sent lower thirds for six panellists and wants a logo in the corner all evening, and there is a green screen in the next room for a remote guest.

The first decision is where each thing lives, and getting it wrong here is the difference between an easy night and a fight.

The logo goes in a downstream keyer, because it belongs to the show and must not flinch when you cut between cameras. The lower thirds go in the other downstream keyer, for the same reason — a name that dissolved away every time you changed camera would look broken. That is both DSKs spent, which is worth noticing now rather than at eight o'clock.

The remote guest is different. They belong to one shot, not to the show, so the chroma key goes in an upstream keyer inside an M/E: guest keyed onto the background, built and checked off air, then taken as a single source when the director calls it.

Then somebody from the stream team asks for a feed without the client's logo. That is the clean feed, and it exists precisely because the logo is in a DSK — you tap program before the downstream keyers and hand it over. Had you keyed the logo upstream, inside the M/E, it would be baked into the picture and there would be no clean feed to give them.

None of that is difficult. All of it is decided in the first ten minutes of setup, and none of it can be changed comfortably once the show is running.

How to do it

Setting up keys before doors. Almost all key quality is won here rather than during the show.

1. Decide upstream or downstream for every graphic first. Belongs to one shot, upstream; belongs to the show, downstream.
2. Count your keyers, especially the DSKs. Two is common, and running out mid-show has no elegant answer.
3. Load every graphic and confirm the fill and key are paired to the right inputs. A graphics machine sends two signals for a reason.
4. Position and size every picture-in-picture now, off air, and note which M/E each one lives in.
5. Turn on safe-area markers and confirm all text sits inside title safe.
6. Check every key over the WORST background it will meet, not the best — a light shirt, a bright slide, a busy wide shot.
7. For a chroma key, fix the room before the desk: light the screen evenly, get the subject well clear of it, then set a garbage matte to crop out everything that is not screen.
8. Only then touch spill suppression and the key threshold.
9. Route the clean feed from before the DSKs and confirm with whoever asked that it is what they wanted.
10. Write down, for every graphic, when it comes on and when it clears.

What goes wrong

Practice it

Six about layering.

Switching a Live Show

Everything so far has been about the desk. This chapter is about the night: the person calling the shots, the document you both work from, the audio that travels with your pictures, and what you do in the two seconds after something breaks on a machine with no undo.

On the night, the switcher operator turns many feeds into one clean show, following the director.

The director and the run of show

On larger shows the switcher operator does not decide the shots alone — they work to a director who calls them. The director watches the multiview and the show and calls the next shot a beat ahead, in a standard rhythm: a 'ready' or 'standby' to warn you, then the command to execute. 'Ready camera two… take two.' The operator cues camera two on preview the instant they hear 'ready,' confirms it, and takes it cleanly on 'take.' On smaller shows one person both directs and switches, calling the shots in their own head — but the discipline is the same.

All of it runs on a run of show — the planned, ordered sequence of the event: segments, who is speaking, when the video roll happens, when graphics appear. Knowing the run of show is what lets you prepare instead of react. If you know a video package comes after the welcome speech, you have the playback cued and the transition ready before the director even calls it. The run of show turns switching from guesswork into a rehearsed performance.

Comms discipline is part of the job. You keep your headset channel clear, listen more than you talk, acknowledge calls briefly ('camera two, ready'), and never clutter the line during a fast sequence. The director needs to trust that when they call a shot it will happen, on time, without discussion. That trust, built over a show, is what lets a gallery move fast and stay calm.

Underneath the teamwork are the cardinal rules that never change: always check preview before you take, never put garbage on program, and keep program clean and stable. The audience should experience a seamless show and never once feel the switching happening behind it.

The call-and-response between a director and an operator is a small, fixed protocol, and it is worth learning as one because it removes almost all the ambiguity from a fast night.

THE READY IS A WARNING, not an instruction. "Ready two" means put camera two on preview and check it. You do it immediately and you say so -- "two ready" -- and now the director knows the shot exists and is good before they commit the show to it.

THE TAKE IS THE INSTRUCTION. "Take two." You take it. You do not evaluate it, you do not improve on it, and you do not wait for a better moment, because the director is watching the show and you are watching the desk.

WHAT YOU SAY BACK is short and it confirms rather than discusses. "Two ready." "Taken." Comms is a shared channel with several people's attention on it, and a running commentary from the switcher blocks the person who needs to speak.

WHEN YOU MUST SPEAK UP, and there are times, keep it to the fact: "two has lost signal", "the laptop is on the wrong slide", "I have no preview". Those are things the director cannot see and needs immediately. Anything that is not urgent waits for a gap.

AND WHEN YOU ARE BOTH: on small shows one person directs and switches. The protocol still helps, because the discipline it enforces -- cue, check, then take -- is exactly the discipline that stops you punching an unchecked shot when you are busy. Plenty of good operators say the calls under their breath.

THE RUN OF SHOW is the document all of this rests on, and the time to read it is not during the show. Mark your own copy: where the video roll is, which segments have a lower third, when the panel comes on and how many chairs, where the awards sequence changes the cutting rate. An operator who has annotated the run of show is preparing; one who has not is reacting.

Cutting cleanly under pressure

Everything in this book is a defence against one fact, and it is worth stating flatly: the moment a source hits program, the audience has seen it. There is no undo, no second take, and no fixing it later.

THAT IS WHY THE HABITS ARE SHAPED THE WAY THEY ARE. You cue on preview because there is no undo. You check before you take because there is no undo. You build composites in an M/E, prepare graphics cold and stay one shot ahead for the same reason. None of it is fussiness; all of it is the craft arranging itself around an irreversible act.

PRESSURE DOES A SPECIFIC THING TO PEOPLE and knowing it helps. It narrows attention. Under stress an operator stops looking at the multiview and starts staring at one monitor, stops hearing comms, and starts pressing faster rather than thinking. The counter is mechanical rather than emotional: go back to the loop. What is live. What is next. Check. Take. Saying it in your head is enough to widen your attention back out.

THE ANSWER TO PRESSURE IS NEVER SPEED. It is calm and a plan. An operator following the run of show at a steady pace will out-perform a fast one with no plan on every show that has ever been cut, because the fast one is making more decisions and checking fewer of them.

AND THE RECOVERY IS ALWAYS THE SAME SHAPE. Cut to the safety shot. Breathe. Find your place. Cue, check, take. Four seconds, and the audience saw a wide shot.

Everything in this book points at one outcome: a show so smooth nobody thinks about the switching. Clean cuts on the beat, graphics that appear and clear at the right moments, every source looking right the instant it goes live, transitions that feel like part of the show rather than interruptions of it. That invisibility is the mark of a good operator — and it is not luck. It comes from setup and format checks, the program-and-preview discipline, good timing, prepared keys, clear comms, and a calm head.

Master those and you can walk up to an unfamiliar switcher, learn its button layout in a few minutes, and cut a clean show — because the equipment changes from job to job but the craft you have learned here does not.

Audio, and why the picture arrives late

A switcher is not only carrying pictures. SDI embeds audio inside the video signal, so a camera feed usually arrives with sound in it, and what the switcher does with that sound is something an operator needs to know even when somebody else mixes it.

AUDIO-FOLLOW-VIDEO means the sound switches with the picture. Cut to camera two and you get camera two's audio. It is convenient for simple work and almost always wrong on a professional show, because it means the sound of the room changes character every time you cut -- which is exactly what an audience notices.

AUDIO BREAKAWAY is the alternative and the normal arrangement: the audio is separated from the video and handled independently, usually by an audio desk mixing microphones properly. Then the switcher cuts pictures and nobody hears anything change.

THE PART THAT CATCHES PEOPLE IS LATENCY. Every stage of video processing costs time. A frame synchroniser costs about a frame. A scaler costs more. A switcher itself adds some. By the time a laptop has been scaled, synchronised and switched, its picture may be several frames behind where it started -- and a frame at 59.94 is about seventeen milliseconds, so five frames is most of a tenth of a second.

Audio, meanwhile, has taken a much shorter path through an audio desk. So the sound arrives first and the picture arrives late. On a speaker's face that is lip sync error, and audiences detect it easily -- sound arriving ahead of picture is far more obvious than the reverse.

THE FIX IS ALWAYS TO DELAY THE AUDIO, never to try to hurry the video. Audio desks have delay for exactly this, measured in milliseconds, and the audio engineer sets it once the video path is settled. Which means the number changes if you change the video path -- adding a scaler mid-show moves lip sync, and it is your job to say so.

WHAT YOU OWE THE AUDIO ENGINEER, then, is a heads-up: how many frames of delay your path adds, and any change to it. "I have put the laptop through a scaler, so it is a couple of frames later than it was" is a complete and useful sentence.

When it goes wrong

Live video fails in a small number of ways, and having thought about each one in advance is most of what separates a calm gallery from a loud one.

A SOURCE DIES. A camera loses power, an SDI cable is kicked out, a laptop sleeps. If it is live, cut away first and diagnose second -- the audience does not care what happened, only that they are looking at black. If it is not live, take it off preview so nobody takes it by reflex, and tell the director in one sentence.

A GRAPHIC IS WRONG. The wrong name is under the wrong panellist. Take the key off immediately -- it is better to have no name than the wrong one -- fix it off air, and put it back when it is right.

YOU TAKE THE WRONG SHOT. Do not snap back instantly, which reads as a flinch and draws attention to the mistake. Let it sit for a beat, then cut on the next natural moment. Half the audience will read a two-second wide shot as a choice.

THE SWITCHER ITSELF FAILS, which is rare and total. This is why serious shows have a plan: a bypass route that sends one camera straight to the screens, a second switcher, or at minimum a known answer to "what do we put up if the desk dies". Ask about it at setup, because inventing it live is not possible.

AND THE ONE NOBODY PLANS FOR: you lose your multiview. A dead monitor takes away the thing you were using to know what is happening. It is survivable if you know your source numbers and can work from program and preview alone, which is a good reason to learn the input list rather than relying on labels.

THE THREAD THROUGH ALL OF IT is that the recovery is decided before the show, not during it. Knowing your safety shot, your bypass, and which failures require a cut versus a fix is preparation, and preparation is what calm actually is.

Setting up a switcher you have never seen

You will be put in front of unfamiliar equipment, probably sooner than you would like. It is far less frightening than it looks, because every switcher is the same handful of ideas in a different arrangement, and you already know the ideas.

FIND THE TWO BUSES FIRST. Program and preview, usually two long rows of identical buttons, usually with program above or below preview and marked. Press a source on preview and watch the multiview to confirm which row is which. That single test tells you more than the manual.

FIND THE TRANSITION CONTROLS. CUT or TAKE, AUTO, the T-bar, and the buttons that select which transition is armed. Check what IS armed and set it to a cut with a sensible auto rate.

FIND FADE TO BLACK and learn where it is by feel. It is the one control you may need without looking.

COUNT THE M/ES AND THE KEYERS. How many banks, how many upstream keyers per bank, how many DSKs. That tells you what the show can and cannot do before anybody asks.

FIND THE OUTPUT ROUTING, which is where desks differ most. Program, preview, multiview and aux outputs, and whether there is a clean feed available. This is usually a menu rather than a button and it is worth finding calmly rather than urgently.

CHECK THE INPUT FORMATS, source by source, as reported by the desk.

THEN REHEARSE THE LOOP a few times with nobody watching: cue on preview, check, take, cue the next. Ten minutes of that on an unfamiliar desk buys you a whole show, because by the time it matters your hands know where things are and your attention is free for the show rather than the equipment.

Practical sign-off

Switching is learned on a switcher. The learner sets up sources, cues on preview and takes to program, performs cuts and dissolves on time, brings a lower third on and off, and cuts a small show to a director. A mentor signs off when the learner cuts a simple show cleanly.

• Set up sources and identified program vs preview
• Cued on preview and took to program with cuts and dissolves
• Brought a lower third / key on and off on time
• Cut a simple multi-source show following a director

The comms protocol

Short, confirming, and off the line the rest of the time. The director needs to trust that a called shot will happen.

What to do when something breaks

Decided before the show, not during it. The recovery is a rhythm, not a judgement call.

On the job

An awards evening, and forty minutes in the director calls "ready three, take three" for a wide of the room as the applause starts.

You take it, and camera three is black.

What happens in the next two seconds is the whole job. You do not investigate, because nine hundred people are looking at black. You cut to the safety shot — the locked wide you asked for at setup, the one nobody repositions — and the room is back.

Then one sentence on comms: "three has lost signal, I am on the wide." The director now knows what happened, what you did, and what they have to work with. They can carry on calling the show around camera three.

Only now do you look. The multiview shows three completely black rather than frozen, which points at power or cable rather than the camera itself, and a runner finds the SDI kicked half out of the back of the camera during the applause.

Three comes back. You put it on preview, check it properly, and tell the director it is available again — you do not take it just because it returned.

Total elapsed: about a minute, and the audience saw a slightly long wide shot during some applause. That is what a recovery looks like when the safety shot was decided at setup instead of invented at forty minutes past.

How to do it

Show day, from arriving to the end of the last segment.

1. Read the run of show and mark your own copy — video rolls, graphics, panel sizes, anything that changes the cutting rate.
2. Set the desk up cold: find the buses, the transition controls, FTB by feel, the keyers and the output routing.
3. Confirm every input's reported format, and take each source to program once to prove it cuts and not just displays.
4. Check tally camera by camera with somebody confirming.
5. Agree the safety shot with the director and ask that camera not be repositioned.
6. Ask what the bypass plan is if the switcher dies. Invent it now, not later.
7. Tell the audio engineer how much delay your video path adds, and tell them again if you change it.
8. Set graphics: positions, safe areas, upstream or downstream, and a written moment for each to clear.
9. Rehearse the loop — cue, check, take, cue — until your hands know the desk.
10. During the show, work the loop and stay one shot ahead. When something breaks, cut away first and explain in one sentence.

What goes wrong

Practice it

Six about the night itself.

Where to Go Next

The switcher seat is unusually well placed for seeing the whole of a show, because everything upstream feeds you and everything downstream depends on you. That view is why switcher operators end up in so many different chairs.

• If the pictures interested you most, go towards camera and directing. The eye you have been training at the desk is the same eye a director uses.
• If the plumbing interested you — formats, reference, routing — go towards video systems and engineering, where the question becomes how the whole facility is timed.
• If the graphics interested you, go towards playback and graphics operation, which is its own craft and increasingly its own department.
• If the LED wall behind the stage interested you, go to LED Video Basics, then towards media servers.
• If the show itself interested you rather than the equipment, go towards production management, where the run of show is written rather than followed.

Choosing without guessing

Those are honest directions and none of them is a decision. What actually decides whether a route suits you is more ordinary than the subject: whether you would rather be responsible for a moment or for a system, whether you like being in the room or in the rack, how much of your week you want to spend travelling, and whether you would rather be judged on taste or on reliability.

The only reliable way to find out is to work beside people doing each of them and notice which conversations you want to join. Nobody chooses well from a description, and the switcher chair gives you a better view of all of them than most jobs do.

Glossary

What the gear actually looks like

Every device named in this book, in one place. Recognition is the whole point of Chapter 4 — use this page to attach a picture to each name before you meet it on a dock at six in the morning.

Audio

Dynamic microphone

Condenser microphone

Wireless handheld

Radio rack

Mixing console

Stage box / snake

Power amplifier

Main loudspeaker

Subwoofer

Stage monitor wedge

In-ear monitors

DI box

Video

Camera

Video switcher

Scaler / processor

Projector

LED wall panel

Confidence monitor

Lighting & control

Conventional fixture

Moving-head fixture

Lighting console

Dimmer

Show controller

Media server

Networking & infrastructure

Network switch

Patch bay

Equipment rack

Power distribution

Cable tester

A properly coiled cable

Connectors, close up

XLR

TRS / TS ¼-inch

SDI on BNC

HDMI

Cat5e / Cat6 (RJ45)

Fiber optic

DMX (5-pin XLR)

AC power — not signal