The slow video is the cheap one

The recording nobody opens

Picture the message. "Missed the call, but I recorded it for you, it's an hour and a half." You thank whoever sent it, put the link on a tab you'll get to later, and that tab sits there for two weeks until you finally close it out of guilt, still unwatched. Everyone I've trained has some version of that folder. Recorded webinars, an all-hands, a course video from the subscription you keep meaning to cancel.

The honest reason nobody opens these is that watching is slow and mostly wasted. Ninety minutes of someone talking usually holds about four minutes you actually need. Skipping around by hand to find them is its own kind of work.

Two ingredients instead of a video file

Here's the workaround, and it isn't complicated once you see it. Instead of feeding Claude the raw video, you give it two things: a still frame every time the picture on screen genuinely changes, not on a fixed timer, and a full transcript of everything said, lined up to the moment it happened. Hand both over, and Claude can answer questions about the recording. What did they decide about the budget? Did anyone commit to a date? Was the pricing slide even shown? You get the answer in a minute, without watching a second of it yourself.

Why length turns out to be the wrong number

The part that took me a while to get used to is what this does to cost. It doesn't track with runtime. It tracks with how often the picture actually changes. A calm, single-camera talk, one person sitting still for forty minutes, produces very few new frames, because very little on screen is different from one moment to the next. A three-minute video cut every couple of seconds, the kind built for short attention spans, produces far more frames than the lecture does, and costs more to process, despite being a fraction of the length.

So the ninety-minute all-hands with one speaker and one slide deck is the cheap file. The ninety-second highlight reel someone edited for social media is the expensive one. That's backwards from how anyone budgets their own attention, which is exactly why it's worth knowing.

Where I actually use this

I use a version of this myself to get through long recordings without watching them in real time. Internal workshop sessions I ran and want to check for one specific moment. Screen recordings of someone walking through a tool, where I need to know if they hit the one step that usually trips people up. An hour of class footage where the only thing I actually need to know is whether a topic got mentioned at all. In every one of those, the recording behaves like the calm lecture, not the fast-cut short, so the economics work in my favor.

What it doesn't fix

It's not magic, and it's worth saying where it falls short. A transcript is only as good as the audio it came from, so a recording with heavy background noise, or three people talking over each other, still needs a closer listen regardless of how few frames it produced. And a fast-cut video stuffed with on-screen text and graphics can generate plenty of frames even when nothing complicated is actually being said, so the trick doesn't rescue every format equally. Treat what comes back as a strong first pass, not a replacement for watching the ten seconds that genuinely matter.

Try it on the next one you're avoiding

Next time a recording lands in your inbox and your honest plan is to never open it, don't add it to the pile. Hand it to Claude and ask the one question you actually care about. "Did anyone commit to a deadline in this call?" "Was pricing mentioned at all?" You'll have an answer before you would have finished skipping through the first ten minutes by hand, and you'll know, for the first time, whether the other eighty minutes were worth going back for.