Recording the talks turns a one-day event into something that keeps being read for years, and it is the commitment most often made lightly. The cameras are the easy part. What decides whether recordings appear is the work that happens in the fortnight afterwards, when the event is over, everybody is tired, and the files are sitting on somebody’s laptop.
Audio decides everything
A recording with excellent video and poor audio is unwatchable; a recording with a still frame and clean audio is perfectly usable. That asymmetry is the single most useful thing to know before spending on equipment.
Clean audio means a microphone on the speaker, not a microphone in the room. A lapel or headset feed taken from the venue’s own desk is almost always better than anything a camera picks up from ten rows back, and it costs nothing extra when the room already has amplification.
The check that saves an event is listening to the first ten minutes of the first talk while the second talk is still running. A fault found at 10am is a fault fixed for the other six talks; the same fault found a week later is six lost recordings.
Capturing the slides
Two feeds make a recording useful: the speaker and what the speaker is showing. Capturing the projector feed directly rather than filming the screen removes the keystone, the flicker and the person who walks in front of it.
Where that is not possible, a fixed camera framing both the speaker and the screen is the fallback, and it is worth setting the framing before the room fills rather than adjusting it during the talk.
The editing that nobody budgeted
Editing a conference talk is not creative work. It is trimming the start, trimming the end, checking that the slide feed is in sync, and exporting. It takes something like twenty minutes per talk when it goes well, and the reason recordings go unpublished is rarely that the work is hard: it is that nobody agreed in advance to do it.
Naming the person before the event, and making it somebody who is not also running the day, is what separates events that publish from events that intend to.
Speaker approval
Speakers are told at acceptance that talks will be recorded and published, and are given a way to say no. Some will, for reasons ranging from employer policy to a talk that named a former client. Asking afterwards, when the recording already exists, puts the speaker in a position they did not agree to.
A short window for review before publication is common and cheap. It also catches the case where a speaker said something they immediately corrected, and would rather the correction was the version that survives.
Multi-track events
Two rooms double every part of the problem and more than double one of them. Capture and editing scale linearly, but the operator does not: one person cannot check the audio in two rooms at once, and the fault that goes unnoticed is always in the room nobody was standing in.
Events that record two tracks well either staff both rooms or accept that the second room is recorded at lower confidence and say so. What does not work is assuming that equipment left running in an unattended room has produced anything usable, which is discovered weeks later.
The scheduling consequence is worth stating. If only one room can be recorded reliably, the programme should put the talks most likely to be watched afterwards in that room, and that is a decision for the people choosing the programme rather than for whoever is holding the camera on the day.
Where recordings live
Wherever they are put, the link belongs on the talk’s own page on the conference site. A recording that exists only on a video platform is discoverable by people searching that platform and nobody else, and it loses the context the programme page carries: the abstract, the speaker, the year.
The archive is also the part of an event that keeps returning value. A talk page that has been linked to for five years is worth more to the event than the day it was recorded on.
Common questions
Is it worth recording a single-track event?
Usually yes, and it is easier than a multi-track event because there is one room, one feed and no scheduling conflict for the operator. The decision still rests on whether somebody has agreed to do the editing.
What if a speaker declines afterwards?
The recording is not published. That is what the review window is for, and honouring it costs the event one talk while breaking it costs the event speakers.
Do recordings reduce attendance?
The evidence from events that publish is that they do not. People attend for the room and the corridor, and watch recordings for the talks they missed in the other track or could not travel to.
How quickly should they go up?
Within a month is a reasonable target and is achievable when the editing is assigned in advance. Past about three months the audience that was waiting has stopped waiting.
Should the questions at the end be recorded?
Only where the room has a microphone for questions. Recording an answer to a question nobody can hear produces a confusing artefact, and it is better to trim it than to publish it.