A Day at the Office

Fourteen interviews with parents, teachers and staff. Two cameras, B-roll, photographs, the whole bloody lot.

I’ve done enough of these shoots now that I’ve developed a fairly decent workflow. But every time I come back to one of these projects, I find another way of doing something.

Yesterday, I worked out how to use Claude to automate the naming of folders on the NAS so they matched the bins in DaVinci Resolve exactly.

Then, for the first time, I used ChatGPT to bring the footage directly into Resolve through prompting.

I’d also jammed timecode on the cameras during the shoot, which meant syncing the interviews took seconds rather than the rather tedious waveform process I’d been using previously.

By the end of the day, I’d built dump timelines, generated transcripts, created selects, developed paper edits and put together a draft of the two-minute hero video.

I estimated that about 75 percent of the work had been done dynamically using AI.

Not bad.

Of course, I’ve no idea how much time I spent working out how to get AI to do the work in the first place.

I have a tendency to tinker.

I’ll start with a perfectly reasonable task, notice something that could be done differently, and follow it. That leads to something else, which leads to another possibility, and before long I’m investigating something that wasn’t remotely on the agenda when I got out of bed.

Sometimes that’s genuinely productive.

Other times, I suspect I’m just finding a particularly interesting way of avoiding the job I’m supposed to be doing.

The trouble is, I don’t always know which one it is until I come out the other end.

But there’s something enormously satisfying about discovering that I can do something today that I didn’t know was possible yesterday.

Even if it takes me six hours to work out how to save four.

Yesterday, I even came up with an idea for a digital clapboard using an ordinary iPad, a spreadsheet and a CSV file. Capture the interviewee’s name, title and approximate timecode, then use the information to automatically name the timelines in Resolve.

Nobody asked me to invent a clapboard.

I was supposed to be editing videos.

But that’s how my mind works. I tend to explore a few different directions, sometimes quite tangentially, until something clicks. Then the pieces fall into place and I lock into a flow.

And I rather enjoy that process.

The funny thing is, everything I’ve worked out over the last few days will probably look terribly primitive in six months.

Some of it might not even work anymore.

I can imagine looking back at this in fifteen years and finding it rather amusing. A bit like listening to 80s music. The synthesizers, the drum machines, the production techniques that once sounded like the future.

Some of it has aged terribly.

Some of it sounds wonderful.

And some of it is wonderful precisely because it sounds so unmistakably of its time.

I suspect these early AI workflows will be much the same.

There’ll probably come a day when someone looks at what I was doing yesterday and wonders why on earth I had to spend so much time telling the computers what to do.

Fair question.

But I don’t imagine I’ll be sitting around mourning the loss of my beautifully constructed workflow.

In six months, this will probably be child’s play.

And there’ll be another hand to play with different rules.

I’ll be quite happy to pick up the cards.

Assuming, of course, I’ve finished editing the bloody videos.