The room and the stream want different things.
Not slightly different. Opposite, in several places at once, and the conflicts are settled at design rather than discovered at rehearsal.
A person in row ten has stereo hearing, a mind that filters, and the ability to look wherever they like. A viewer at home has one frame somebody else chose, one audio mix, and a device with a small speaker. Everything that the room supplies for free has to be built for them.
That is why the stream gets its own audio path rather than a microphone pointed at the room, its own lighting requirement, and its own operator. A single unattended camera at the back of a ballroom is a recording, and it is the right answer for some jobs — but it should be chosen deliberately, not arrived at by treating the stream as something the AV package throws in.
The room mix is the wrong mix to send.
This is the single most common failure in a streamed event, and it is audible within about four seconds of a viewer arriving.
The mix in the room is built for the room: it accounts for what the audience hears acoustically from the stage, for the reverberation, and for a level set against several hundred people. Send that same feed down a wire and the viewer gets the parts of it that were never meant to stand alone — a lectern microphone thin because the room was carrying the body, a panel of four voices at wildly different levels because in the room the difference was small, applause that swamps everything because it was flattering live.
A broadcast feed is mixed separately. It is usually derived from the same sources, then balanced for headphones and a phone speaker rather than for a ballroom, with question microphones brought up deliberately and the room brought in at a level that reads as atmosphere rather than noise.
Two more things the viewer notices and the room does not. Audience questions asked from a floor microphone are audible on the stream; questions shouted from a seat are not, so a floor microphone is a production decision rather than a courtesy. And the moment there is playback video with its own soundtrack, somebody has to be responsible for it arriving in the stream at the right level — it is the most common single point where a stream falls apart mid-show.
How many, and what each one is for.
Camera count is not a quality setting. Each position answers a question the others cannot, and an unanswered question is what makes a stream dull.
- The wide
Locked off, covering the whole stage. It is the shot nobody compliments and the one that saves the stream every time something unexpected happens, because it is always right. On a one-camera job this is the camera.
- The tight
The presenter's face, operated. This is what carries a keynote — people watch people — and it is the difference between a stream somebody stays on and a stream somebody leaves.
- The content feed
Slides, video and graphics taken directly rather than by pointing a camera at a screen. Anything shot off a screen carries the screen's moiré, its brightness and its color shift, and text becomes unreadable at the exact moment it matters.
- The angle
A cross shot, a side, or a camera on the audience. It exists so a panel is watchable, so a cut has somewhere to go, and so a two-hour program has visual rhythm rather than one unchanging frame.
A camera position is also a floor plan decision. A tight camera needs a clear line to the lectern that no audience head crosses, which means it stands somewhere seats were going to go, or on a riser somebody has to build. Deciding that on the day means choosing between a compromised shot and moving furniture in front of arriving guests.
The venue's internet is the part nobody checks.
A stream is only as good as its upload, and the number that matters is not the one on the venue's brochure.
Venue internet is sold on download speed, because that is what guests complain about. A stream does the opposite: it pushes constantly, upward, for the entire length of the program, and it cannot pause to catch up.
Encoding a stream at a given bitrate means sustaining that rate for the whole show, plus headroom, plus a second copy if the stream is redundant. For a 1080p program encoded at 6 Mbps, run twice for redundancy, with headroom:
6 Mbps × 2 paths = 12 Mbps sustained
12 Mbps + 50% headroom ≈ 18 Mbps upload, dedicated
and dedicated means not shared with the guest wifi
Which is usually more than a ballroom's house connection will give you on the day, once several hundred phones are on it. So the practical answers are a hard-wired connection the venue reserves, a bonded cellular path, or both — the second one existing precisely because the first one will occasionally fail, and a stream with no fallback has a single point of failure running through somebody else's building.
Redundancy is the job here. Two encoders, two paths out, a recording made locally regardless of what the stream does, and somebody watching the outbound feed rather than the room. Live means there is no second take, and every part of this page exists because of that one fact.
A room that looks good can stream badly.
The camera has one exposure and no ability to adapt the way an eye does, so the lighting brief changes the moment a stream is added.
A dramatic room — low ambient, light from the sides and above, deep shadow, a bright wall behind the stage — is a difficult picture. The sensor exposes for the brightest thing in frame, which is usually the screen, and the person standing in front of it goes dark. What fixes it is not more light everywhere but even front light on faces, set against the wall's output rather than independently of it.
The same goes for color. Your eye adapts between a warm front light and daylight through a window without telling you; a camera is balanced to one value and everything at the other shifts. A face lit warm against a cool window reads orange on the stream no matter how it looked in the room.
None of which is expensive to get right, and all of it is expensive to fix afterward — which is the argument for the lighting package and the stream being scoped as one thing rather than two.
Six answers, and the rest follows.
None of these is a question about equipment. Each one changes the production, which is why they come first.
Who is watching
A public audience, a private one, or a record for later
What happens on stage
One presenter, a panel, or a performance — it sets the camera count
Content on screen
Slides and playback are their own feed, not a shot of a screen
The upload
What the venue can sustain, dedicated, for the whole program
Interaction
Questions from the floor, from the stream, or neither
What happens after
A recording that gets used is edited, and that starts at capture
If the show needs it, we have it. On a streamed event that matters more than usual, because the stream and the room are produced by the same crew at the same moment: an encoder, a camera operator and a broadcast mix are not a second vendor standing in the back of the room with their own idea of the run of show.
Where the figures come from
- 6 Mbps, doubled, plus 50% headroom — a working figure for a 1080p program rather than a fixed requirement. The arithmetic in §04 is computed from it on the page; the point that survives any change in platform guidance is the relationship — sustained upload, multiplied by redundancy, plus headroom, and dedicated rather than shared.