The clip sounded worse than the show

Mamba Smith is the founder of Queen City Garage, which produces The Grand National Show: a weekly NASCAR panel show on Grand National Today where broadcasters and drivers argue about the last race and pick the winner of the next one.

The first social clips cut from this week's Truck Series episode sounded worse than the episode they were cut from. Same recording, same people, same words. About twelve decibels more noise.

That is not a small difference. Twelve decibels is the gap between a clean voice and a voice with a room behind it, and it was entirely our doing.

Where the noise came from

A remote panel recording gives you two kinds of sound for each person. There is the audio baked into their camera file, which is whatever their laptop microphone heard: the voice, plus the fan, plus the buzz from the power supply, plus the room. And there is the separate microphone track, which is what the show is actually mixed from after it has been cleaned.

The show master uses the cleaned tracks. The first clip cut, in a hurry, pulled its sound from the camera files. Buzz and hiss on every clip. The test copies were pulled before anyone outside the building heard them.

Three rebuilds

The second cut used the cleaned microphone tracks and was still unusable, because the cleanup that works across a thirty-minute show does not automatically hold on a fifteen-second excerpt with a hard start.

The third cut had good sound that arrived 4.4 seconds early.

The fourth is the one that shipped: nine clips, best takes only, with light cleanup on two panelists whose microphones had carried a steady tone. Nine instead of the twelve we had planned, because three of the candidates were not clean enough to send and we would rather send fewer.

The check that exists now

Every clip is now measured against the master before it is packaged, three ways. Noise floor: is the clip's background within range of the master's. Steady tones: is there a hum or whine the master does not have. Sync: does the sound sit within 50 milliseconds of the picture, checked against the master's timing rather than the clip's own.

The packaging step refuses a clip that fails any of the three. It does not warn. It refuses. A person can override it, on purpose, with a reason. The default is no.

What went out

Sixteen clips across the two shows this week, nine from the Truck episode and seven from the NASCAR O'Reilly Auto Parts Series episode, to seven panelists. Each panelist's package runs about 60 to 100 megabytes: their clips in a captioned vertical cut for reels and stories, a clean widescreen cut, and a clean vertical, plus covers and posting copy. The encodes are phone quality on purpose. The master quality stays on the shelf. Nobody needs a two-gigabyte reel on a phone.

Old links were revoked when the clean set went out, so nobody can post the noisy version by accident.

The rule underneath

Every clip comes off the master chain, never off a camera file, and every clip is measured against the master before anyone sees it. We could have written that rule on day one. We wrote it after three rebuilds and a pulled test send, which is when most rules get written.

What a person did: pulled the noisy test copies, chose nine clean clips over twelve, wrote the rule. What the machine did: now measures noise, tone and sync on every clip against the master, and refuses to package one that fails.

This is No. 12 in How We Make the Show, the Queen City Garage build log on producing The Grand National Show. Watch the episode: Jake Garcia Breaks Through at Bristol, Kansas Picks.