Measuring a YouTube course's real length

The question nobody can answer from the page

“How long is this course?” is the first thing anyone evaluating training content wants to know, and YouTube does not tell you. It shows a video count. Forty videos could be three hours or thirty.

The playlist scraper does tell you, because duration is typed as a number in seconds and marked always present. No parsing, no ISO 8601 duration string, no abbreviation. Sum the column.

run = client.actor("apidojo/youtube-playlist-scraper").call(run_input={
    "startUrls": [f"https://www.youtube.com/playlist?list={playlist_id}"],
    "maxItems": 100,
})
videos = list(client.dataset(run.default_dataset_id).iterate_items())

hours = sum(v["duration"] for v in videos) / 3600
print(f"{len(videos)} videos, {hours:.1f} hours")

Two filters before the sum means anything

A raw sum is usually wrong, for two reasons that are both visible in the output.

Livestreams. isLive is in the record. A single eight-hour stream parked in a course playlist adds more running time than the rest of the curriculum combined, and it is not something anybody is going to watch as a lesson.

Private and deleted entries. isPrivate appears in the sample. Entries get removed after a playlist is built and still come back as rows, billed like any other. They contribute nothing to the course and, worse, their presence means the video count overstates what you can actually watch.

watchable = [v for v in videos if not v["isLive"] and not v.get("isPrivate")]

print(f"{len(watchable)} of {len(videos)} usable, "
      f"{sum(v['duration'] for v in watchable)/3600:.1f} hours")

The gap between those two counts is itself a finding. A course with nine dead entries out of forty is a course nobody has maintained.

Why a nickel covers it

The pricing makes this job almost free. A playlist query is $0.05 and carries the first hundred videos, and most real courses are well under a hundred videos. So the per-video rate never applies at all: an album, a conference track, a module, a series, all come back complete for five cents.

That changes what is worth doing. Auditing one playlist is cheap enough to be worth doing casually. Auditing a shelf of twenty, to decide which training library to buy, costs a dollar and answers a procurement question that would otherwise be guesswork.

The allowance is per playlist rather than per run, so twenty playlists carry twenty allowances. Breadth is the cheap direction here, exactly as it is on the location and hashtag actors.

The number worth reporting

Not the video count, which YouTube already shows and which overstates. Report watchable hours, the count of dead entries, and the longest single item. Those three together describe a curriculum in a way the platform’s own page does not, and they cost five cents to produce.