How One Person Posts 4 YouTube Videos a Week

Sam runs a channel about supply chains. Container shipping, port congestion, why a factory in one country stopping means a shelf in another country is empty six weeks later. Eight to twelve minutes a video, no face, no on-camera anything. Four uploads a week.

They've been doing it for two years. What follows is roughly a normal week, including the parts that don't reflect well on the process.

Monday

Ideas. Sam opens ChatGPT and dumps in the week's shipping news, then asks for twenty angles that aren't "here's what happened."

Most of them are unusable. Three or four are worth thinking about, and one is usually good — this week it's the question of what happens to the crew when a vessel gets arrested in port, which Sam hadn't considered as a story and which turns out to have a lot in it.

This is the part of the week that a pipeline can't touch and shouldn't try to. Sam isn't asking a tool to decide, they're asking it to widen the field enough that they can decide. The back-and-forth is the point, and it takes about forty minutes.

Tuesday and Wednesday

Scripts. Four of them, eight to twelve minutes each.

Sam drafts in Claude, working from research they've already gathered, because Claude holds tone across a full script better than anything else they've tried — they paste in the transcript of an older video that performed well and ask it to match the pacing. First drafts come back needing real work, but they come back structured, which is the part that used to take longest.

Then Sam rewrites. Every script gets a manual pass, sometimes two. The AI draft is scaffolding, and Sam has been clear with themselves that the day they stop rewriting is the day the channel stops being theirs.

Wednesday evening: voice through ElevenLabs, using a cloned version of their own voice recorded eighteen months ago.

Thursday

Edit. Descript for the assembly, footage from a stock library and some archival material, and this takes most of the day because four videos is four videos. The stock library is a paid one; Pexels covers a lot of the same ground free, and did for Sam's first year.

By about seven in the evening there are four finished exports sitting in a folder.

This is the moment the week used to fall apart. Four videos, done, and then between two and three hours of metadata work standing between Sam and being finished — transcripts, titles, descriptions in the channel's format, chapters at real timestamps, four thumbnails found and cropped and compressed. On a bad Thursday it just didn't happen, and a video went out with a first-draft title and whatever frame YouTube picked.

Now it's a batch. Sam uploads the four files, selects the full automation template with best-frame thumbnails, and goes to make dinner.

Thursday, later

The review panel has four sets of results in it.

Sam reads the first two properly. Reads the titles, checks the chapters start at zero — the rule that silently kills the whole list if it is broken — looks at the frames the pipeline picked at the size they will actually appear, and changes one title because the generated version buried the interesting part in the second half.

The third and fourth they skim. It's quarter past ten, the output from the first two was good, and the pattern-matching part of Sam's brain has decided the other two are probably fine. They approve all four.

They are probably fine. Sam does this most weeks and it has not yet gone wrong in a way they noticed.

Friday

Scheduling, comments, and the analytics pass. Sam looks at the previous week's CTR by video in YouTube Studio, mostly to see whether the thumbnails the pipeline chose are doing better or worse than the ones they'd have picked by hand. It is a narrow read of a wider set of numbers, and deliberately so.

The answer, over about a year, is that the pipeline is slightly better on average and occasionally much worse. It picks good frames of people gesturing, which Sam wouldn't have thought to use. It also once chose a frame of an empty dock that was technically well-composed and completely dead as a thumbnail, on a video that underperformed for the whole week.

Friday afternoon is free. It didn't used to be.

What Sam does with the free afternoon

Makes a fifth video, mostly.

This was not the plan. The plan, when Sam started automating the post-production end, was to get back to a four-day week and stop working Fridays. What actually happened is that the capacity got absorbed into more output, because more output has been working, and the channel is growing faster than it was.

Sam is aware this is a trade they didn't consciously make. They've mentioned it twice in their community posts, both times slightly defensively. Whether it was worth it is not something they've settled, and they'd probably tell you they'll take Fridays off properly after the next milestone, which is what they said about the previous one.