The mix that works in the room is the wrong mix for the stream, and the reason is not taste. The room supplies half of what the audience hears, the stream supplies none of it, and the person watching at home is in a different acoustic situation with different equipment and different attention.
The room was doing work that does not travel
A reinforcement mix is set against the acoustic sound present in the building. Unamplified voices, the orchestra in the pit, the natural blend of a chorus: all of that reaches the audience directly, and the console is adding only what the room could not deliver by itself.
Take the console output on its own and the acoustic half is missing. What remains is the correction rather than the performance: reinforced elements dominate, anything not amplified is faint, and the balance someone spent six weeks setting is not present in the file at all. This is why a capture path needs microphones in the house alongside the console feed, as described in the article on archival capture.
The listener has a fraction of the range
A quiet hall lets a pianissimo exist. It sits above the room's noise floor and below the point at which a fortissimo becomes uncomfortable, and the distance between those two is generous.
A phone in a kitchen has a noise floor made of appliances and traffic, and a ceiling set by what the household will tolerate. The usable distance between them is a fraction of the hall's. Material that sits below that floor is not quiet on the stream, it is absent, and the listener's response is to turn it up until the next loud passage makes them turn it down again.
Compression on a stream mix is not a stylistic choice, it is what makes the quiet material exist for the listener. The question is how much, and the honest answer depends on where it will be heard rather than on any principle.
Loudness is normalised anyway
Distribution platforms measure programme loudness by a standardised method published by international broadcasting bodies and adjust playback so that one programme does not arrive louder than the next. Mixing hot to seem impressive therefore gains nothing: the level is turned back down at the far end, and what remains is the reduced dynamic range that was spent to get it.
Most people are listening in mono
A single phone speaker is a mono system, and a laptop's pair is effectively one. Anything achieved by width in the mix collapses, and content that relied on phase differences between channels can partly cancel when the two are summed.
This matters more than it used to precisely because the stereo pair was doing separation work. In a hall a listener has two ears, a head and the ability to attend selectively to one voice among several. A phone gives none of that back, so the separation has to be built into the mix by level, by frequency and by arrangement rather than by position.
Small speakers lose the bottom, and find the top
Reproduction on earbuds and laptops is weak in the low midrange, which is where the body of a voice lives, and often forward in the upper frequencies, which is where sibilance lives. The same voice that sounded warm in the room can arrive thin and hissy.
The discipline that follows is unglamorous and reliable: check the stream mix on a phone speaker and on cheap earbuds, not only on good monitors. Good monitors reveal problems that do not matter to this audience and conceal the ones that do.
Reinforcement bleed makes the problem worse
House microphones placed to capture the acoustic half also capture the loudspeakers, because the loudspeakers are in the same room pointing at roughly the same place. What arrives at the pair is therefore the performance plus a delayed, room-coloured copy of the console mix, and adding the console feed on top of that sums a signal with a smeared version of itself.
The audible result is a hollow quality that no amount of balancing removes, because it is interference rather than balance. Aligning the console feed in time with the house pair, exactly as spot microphones are aligned to a main pair, is what makes the combination work. It is the same mechanism appearing for the third time in this publication, and it is worth recognising on sight.
It needs its own operator
The engineer mixing the room is fully occupied mixing the room, and the two mixes want different things at the same moments. Applause is the clearest case: in the hall it is the event, and on the stream it is two minutes of noise at a level nobody chose.
A production that assigns the stream to a spare output of the show console and nobody at all has decided that the remote audience receives whatever falls out. That is a legitimate decision if it is made deliberately, and it usually is not.
What we cannot verify
Loudness targets differ between platforms and territories and are subject to change; the applicable figure is the one in the current specification for the destination, and we reproduce none. Claims about the frequency response of consumer playback devices come from their manufacturers and vary between units. Nothing here substitutes for listening to the actual stream on the actual devices an audience will use.
The short version
- A reinforcement mix is the correction, not the performance; the room supplied the rest.
- A phone in a kitchen has a fraction of a hall's usable dynamic range.
- Compression on a stream mix is what makes the quiet material exist at all.
- Platforms normalise loudness, so mixing hot spends range and gains nothing.
- Most listeners are effectively in mono, so separation must be built into the mix.
- Check on a phone speaker and cheap earbuds; good monitors hide the real problems.