A channel that had not posted in five years, 149 leftover subscribers from an abandoned paranormal channel, no audience worth the name. Its seventh new video took 1,269 views in a day.
That is the number worth starting with, because it is the one that matters if you have nothing yet: the Shorts feed does not need you to have subscribers. It needs one video it wants to show.
This is not a case study from someone with a million views selling you a course. It is four days of our own data, published in full, including the batch of six videos that died on publication and the one that took zero views. If you are at the beginning, small numbers with the mechanics visible are more useful than large ones with the mechanics hidden.
The setup: twenty vertical Shorts over four days, all the same format. A first-person petty-drama story, synthetic voiceover, word-by-word captions, and a real reaction meme on each punchline. Nobody on camera, no gameplay footage running underneath.
The numbers
Twenty videos took just over 3,000 views in four days. The total is the least interesting part; the split is where the lessons are.
| Batch | Videos | Published | Views |
|---|---|---|---|
| First eight | 8 | Days 1–2 | 2,426 |
| Second six | 6 | Day 2, ten minutes apart | 247 |
| Third six | 6 | Days 3–4, three per day | 333 and counting |
The single best video took 1,269 views. The worst took zero. Not "a few", zero, with an empty traffic-sources panel three days later.
Two videos took most of the channel
Of roughly 3,000 views, 1,843 came from two videos. Everything else split the remainder.
That is normal for Shorts and worth knowing before you plan content volume: the feed does not spread reach evenly across what you publish. It picks. Fourteen of the twenty videos sat between 11 and 221 views, and no amount of publishing more of them changed that.
The practical version: your channel's result is decided by whether the feed picks anything at all, not by your average.
Publishing six videos in ten minutes killed all six
This is the finding we did not expect and would not have believed from someone else.
The second batch went out as a burst — six Shorts within about ten minutes. Total across all six, two days later: 247 views. Two of them were never distributed at all.
The first batch had also gone out in a burst, but a slower one, five videos across 35 minutes, and it produced the channel's two best videos. The third batch went out three per day, spaced by hours, and three videos beat the entire six-video batch inside 24 hours.
We cannot prove the mechanism from one channel. But the ordering is consistent: 35 minutes was survivable, ten minutes was not, and spacing by hours worked. If you batch-produce content, publish it slowly.
Fourteen videos, zero comments
For the first fourteen videos, across 2,597 views, the channel got zero comments.
We changed one thing on the third batch: every script now ends with a direct question to the viewer. "Was that overkill, or the only thing that would ever work?" The first genuine viewer comment arrived on the first video of that batch.
One comment is not a result, and the comparison is not clean — we had also started leaving a comment of our own under each video, which invites replies by itself. But going from zero to non-zero after changing exactly one thing is worth the two seconds it costs to write a closing question.
What we still cannot explain
The watermark. Two of the twenty went out without the "Made with memecut.fun" mark in the corner, by a coin flip rather than by choice. Those two are the 1,269 and the 574. Across all twenty, the videos without a mark lead on median views by a wide margin.
That is suggestive and nothing more. Two videos is not a sample, the arms are unbalanced, and one whole batch was suppressed for unrelated reasons. We are still running the split.
Retention. YouTube Studio never produced audience-retention data for any of these videos. Three days after publishing, it still says processing. At this channel size the retention curve simply does not appear, which means the most useful diagnostic for short-form video was unavailable to us the entire time. If you are small, plan to work without it.
Where the videos came from
Every video here was made by pasting a script into Memecut. The voiceover, the word-by-word captions and the meme on each line are generated from the text; nothing was assembled by hand in an editor, and no meme was picked manually.
That last part is the whole reason the channel exists. Placing a reaction meme somewhere in the right paragraph is easy. Placing it on the punchline of a specific line, so the joke lands twice, once in the words and once in the image, is the part that takes a human 30 to 60 minutes per video in CapCut. Memecut does that matching by searching a library of real memes for the one that fits the line, which is also why the memes are recognizable images rather than generated art.
It is not perfect. On these twenty videos the matcher sometimes latched onto a word rather than the joke, and one viewer told us so in that first comment: "Ok i think that one not a meme." Fair.
What to take from this
Publish slowly, even if you produce in batches. Expect one or two videos to carry the channel and plan volume accordingly rather than chasing an average. End on a question if you want anyone to say anything. And do not count on retention data to tell you what worked until the channel is bigger than this one.
If you want to try the format, the script-to-video flow is the same one that produced every number above, and pasting any text into the reaction tool shows you the meme matching on its own, without making a video.