cinelet Watch on Cinelet

Data · Short films on YouTube

YouTube doesn't reward the film. It rewards the channel.

We found 229 short films uploaded twice: once by the person who made them, and once by a big channel. Same film, frame for frame. This is where the views, the search results and the money went.

By the maker of Cinelet2026-11-0127,126 films checkedDownload the data (JSON)

"Desire", the same 9-minute animated film, uploaded twice in 2014

Same scale. The maker's bar would be less than a pixel wide; it is drawn at 2. Desire is one of the largest gaps we found, not a typical one; the typical gap is below.

I'm a cinephile, and short films are my favourite part of it. For years I watched YouTube bury them. I assumed it was taste: maybe the films that get millions of views are simply better. Then I found a way to test it.

Short films are often uploaded twice. The filmmaker puts the film on their own channel, and later a big showcase channel (Omeleto, DUST, CGMeetup, Short of the Week, ALTER) uploads the same film. The film doesn't change. Only the uploader does. If the views follow the film, both copies should do about the same.

They don't.

1The same film, 12 times apart

We matched copies by picture, not by title: frames that look alike and a length within 3 seconds. Then we kept only the pairs where one copy sits on a channel whose name matches the film's credited maker. That gave 229 films.

12×median views, big copy vs maker's copy
173of 229 films where the big copy won
124films with a gap of 10× or more
56films with a gap of 100× or more

How many times more views the big channel's copy got (229 films)

maker won1–3×3–10×10–100×100×+
The maker's own copy won in 56 of 229 films. In the other 173, the big channel's copy got more views, and in 124 at least ten times more.
FilmMaker's channelBig channel
DesireTheCGBros, 19 days later4,69127,311,966
ShatteredCGMeetup, 26 days later6,5386,840,244
DownpourNITVShorts, 3 years later102346,894
STOPOmeleto, 2.5 years later38187,036
LOOPShort of the Week, 3 months later21077,144

2"You just measured subscribers"

That's the first objection, and it's fair: a big channel has more subscribers. So we looked them up. The big channels in these pairs have a median of 3.5 million subscribers. The makers have a median of 1,024. That's 870 times more reach.

If views simply followed subscribers, the big copy would win by something like 870×. It wins by 12×. Per subscriber, the maker's own copy does better in 203 of 226 films: about 7 views per subscriber on the maker's channel, against 0.05 on the big one.

Big channel vs maker, median of 226 films (log scale)

The film pulls its audience far harder than the channel does. The channel's size decides how many people the film ever reaches.

The film does the work. The channel collects the views.

3Search for the film by name. You get someone else's copy.

Recommendations are a black box, but search shouldn't be. If you type a film's exact title, the person who made it is the obvious first answer. We searched YouTube for the maker's own title of all 229 films, logged out, and recorded the top 20 results. The film showed up in the top 20 in 100 of those searches; many of the others have titles too common to find, like "Ember".

Searching a film's exact title: which copy comes first? (100 searches)

Maker's copy firstAnother copy above the maker'sMaker's copy not in the top 20

In 71 of 100 searches, the first copy of the film YouTube shows is not the maker's.

Logged-out searches from one location on 2026-11-01, top 20 results. When both the maker's copy and the big channel's copy appeared, the big one ranked higher in 24 of 32. In the other 129 searches the film wasn't in the top 20 at all, often because its title is a common word or phrase.

4Uploading first doesn't save you

The usual advice to filmmakers is to post on your own channel first and let it grow. In 154 of these films the maker did exactly that, with a median head start of 170 days. The later copy still overtook them in 109 of 154 (71%).

Being first helps a little: when the big channel posted first, it won 60 of 69 times (87%), by a median of 51×. But half a year alone with your own film is worth little against one upload on a big channel.

5The money goes the same way

YouTube's Content ID lets a rights holder claim a video, run ads on it and collect the money, or simply track it. 87% of the big channels' copies carry a claim (193 of 222). Only 16% of the makers' own copies do (37 of 229).

And half of the makers (113 of 226) have under 1,000 subscribers, below the bar YouTube sets before it shares any ad money with a channel. On their own channel, most of these filmmakers can't earn from their film at all. On the big channel, someone can.

To be fair to the showcase channels: many of them license films with the maker's permission. Whether that deal is good is between them and the maker. What the numbers show is that on YouTube, a short film gets an audience by handing itself to a bigger channel.

6Not every big channel is the same

The gap depends a lot on who uploads the copy. Some channels multiply a film's views; others barely move them, and one does worse than the makers.

Big channelFilmsMedian gap
Omeleto6314×
ALTER1262×
Crypt TV741×
DUST2134×
CGMeetup4627×
KIS KIS - keep it short99.5×
TheCGBros357.4×
Short of the Week176.5×
Viddsee54.9×
Omeleto Drama111.3×
Film Shortage100.1×

Channels with at least 5 films in our pairs. A gap under 1× means the maker's copy has more views.

7Why I care

I built Cinelet because I wanted a place where short films come first, and where a film is judged by what it is, not by the size of the channel that posted it. It plays films in YouTube's and Vimeo's own players, so the views still count for whoever uploaded them. When a film exists in several copies, we play an official one first (the maker's, their studio's or a licensed showcase) rather than a random re-upload.

That doesn't fix YouTube. But the numbers above should bother anyone who makes things for the internet: the same work gets a fraction of the audience when the person who made it posts it. And the platform's own search, asked for the work by name, often points somewhere else.

How we measured

  • Catalogue: 27,126 short films on YouTube and Vimeo, collected by Cinelet's crawler from channels, festival lists, curated collections and search, with view counts as YouTube reports them.
  • Same film: copies matched by picture (frames that look alike, length within 3 seconds), never by title alone.
  • Maker's channel: the channel's name matches the film's credited maker. Big channel: a channel we classify as a curator or a studio. 229 films qualify.
  • Subscribers: YouTube Data API, 2026-11-01, for the 226 pairs with public counts.
  • Search: the maker's own title, YouTube search, logged out, one location, 2026-11-01, top 20 results.
  • Content ID: YouTube's own "licensed content" flag on each copy.
  • Data: every pair, with both copies' views, subscribers, upload dates, claims and search ranks: same_film_two_uploads.json.

What this doesn't prove

  • We can't see YouTube's recommendation system, only its results. A big channel also has better thumbnails, titles, timing and a habit of posting; all of that is in the 12×.
  • The sample isn't random. Big channels choose which films they take, and our crawler reads them more than small channels. Films no big channel picked up aren't in these pairs.
  • View counts are a snapshot. We can't see how each copy grew over time.
  • Search results vary by country, account and day.