Here's the link
Transcript :
Key takeaways
- 58 compilations of TikTok clips reached 1,390,902 views and 2.98k subscribers. Once the pipeline ran, filing a clip took one Stream Deck key press.
- Views kept coming after the last upload: about 212,000 from December 2025 to August 2026, according to Social Blade.
- YouTube lists short videos compiled from other social media sites as not monetizable, so AdSense was never an option.
- Its external-links policy counts a URL written in the video as a link and bans links to pornography. My domain led to dating offers through a prelander.
In August 2025 I started a faceless YouTube channel made only of TikTok clips. Once it was automated, filing a clip took one key press on a Stream Deck, and n8n and FFmpeg did the rest. Fifty-eight compilations later, I stopped uploading. By September 2026 the channel had 1,390,902 views and 2.98k subscribers, and about 212,000 of those views came in after the uploads stopped.
The views were the easy part. Monetizing them was not: this is borderline-adult content on YouTube, and very few offers fit it. Below: the pipeline step by step, where the views came from, and why the money side was so hard.
| Measure | Figure |
|---|
| Uploads | 58 between 6 August and 2 November 2025, at most one a day |
| Views | 1,390,902 |
| Views after the last upload | About 212,000, December 2025 to August 2026 |
| Subscribers | 2.98k |
| Videos over 20k views | 17 of 58 |
| Median views per video | 11.6k |
| Best video | 220k, a Blonde compilation from 24 August 2025 |
| Clips per compilation | 18 to 22, on the 33 videos with chapters |
| Video length | 3:06 to 6:02, 4.1 hours in total |
The channel's About panel on 24 September 2026. The lander link is cropped out.
The idea
The niche was TikTok compilations: short vertical clips grouped by theme into videos of about four minutes. Clips are fast to find, cost nothing to make and have decent viral potential.
AdSense wasn't an option, so I registered a short one-word .pm domain and burned it into the videos as the call to action.
I did everything by hand at first. Once the concept worked, I automated it.
Step 1: one key per category
Collecting was the only step that needed a human. I scrolled TikTok on desktop, and when a clip fit, I pressed a key on a Stream Deck. The key sent the page I was on to a webhook, which started an n8n workflow.
Eight keys were lit: Blonde, Brune, Redhead, Sexy, Bikini, Gym, Asian and Next. Brune is French for brunette. Filing a clip took one press, with no copying, downloading or renaming.
One press sends the TikTok URL and its category to n8n.
Step 2: n8n downloads, stores and tags
The workflow called an API that downloaded the video. n8n stored the file and added a row to a table: the clip and its category tag. Then it asked one question: did this category have enough clips for a compilation?
Download, store, tag, then count per category.
Categories filled at their own speed. Redhead never got there: the channel has no Redhead compilation.
Step 3: FFmpeg builds the video
When a category had enough clips, n8n triggered FFmpeg on my server. FFmpeg stitched the clips into one video and burned the domain into the picture at random positions.
Clips joined end to end, the domain burned in at random spots. The command is illustrative.
The 33 videos with chapters hold 18 to 22 clips each, and all 58 run from 3:06 to 6:02. From 21 August to 3 October, descriptions carried those chapters: one timestamp per clip, labelled with the creator's TikTok name.
Step 4: upload and schedule, one a day
The finished file went to YouTube automatically and was scheduled at one video a day. The public upload times show it: 58 videos, never two on the same day. From 6 September onward, all but one went live at 12:00 UTC, almost to the minute. Apart from a few gaps, one went out every day until 2 November.
Every upload from 6 August to 2 November 2025, then views per video. Public data read on 24 September 2026.
Early on I'd hit two goals: a fully automated YouTube-to-affiliate pipeline, and a video scheduled every day, set and forget. The next was a two-week buffer, topped up every 15 days.
Get NoIndex in your inbox
Things you haven't heard before: unusual ideas, real experiments and detailed methods rarely shared elsewhere. Get new NoIndex guides and tutorials, with the steps, results and mistakes from my own projects.
Email address
Subscribe
I agree to receive NoIndex experiments, guides and tutorials from Clement at Theory18 by email.
Privacy policy
Leave this field empty
Where the views came from
| Category | Videos | Views | Median per video |
|---|
| Blonde | 15 | 586,189 | 13.7k |
| Brune (brunette) | 21 | 364,645 | 11.6k |
| Sexy (mixed) | 9 | 275,352 | 10.9k |
| Bikini | 9 | 142,770 | 13.4k |
| Gym | 2 | 16,990 | 8.5k |
| Asian | 2 | 5,033 | 2.5k |
| Redhead | 0 | none | none |
Per-video counts, read on 24 September 2026. They add up to 1,390,979, a few more than the About panel showed minutes earlier.
The first weeks were the best. The nine August uploads took 571,301 views, 41% of the channel's total, with a median of 33k per video. September's median fell to 11k and October's to 8.9k. When I first wrote about the channel, after about 25 uploads, early videos were doing 20k+ each and new ones about 5k. I put that down to the automation or API limits.
Blonde carried the channel: 15 videos, 42% of all views, and the best video at 220k. Asian and Gym stayed small. Hits could still land late in the run: 114k for a mixed compilation on 19 October and 68k for a brunette one on 28 October.
The most-viewed uploads, sorted by YouTube. Thumbnails blurred.
The views didn't stop when the uploads did. Social Blade's monthly figures show 302,497 views gained in September 2025, 311,212 in October and 250,509 in November. After the last upload, the catalogue still pulled about 212,000 more views from December to August, tapering from 84,678 in December to between 5,000 and 7,000 a month over the summer. In September 2026 it still gets around 146 views a day.
Views gained per month, from Social Blade. August 2025, the launch month, is not on its chart.
Subscribers followed the same curve: +790 in September 2025, +510 in October, +280 in November and +130 in December, then almost nothing from January on. That's about one subscriber for every 467 views.
The hard part: an offer that fits
AdSense was out from the start. YouTube's monetization policy lists "short videos you compiled from other social media websites" among the examples of content that can't be monetized, and it applies its reused-content rule to the channel as a whole. Its advertiser-friendly guidelines add that highly sexualized titles, thumbnails or themes are not suitable for advertising. That line depends on intent and focus, not on swimwear as such, but titles built around "sexy girls" sit close to it.
The offers closest to what this audience watches for are adult ones: cam sites and paid subscription platforms. YouTube closes that door too. Its external-links policy counts a URL shown as text in a video as a link and bans links to pornography, and its list of what isn't allowed includes telling viewers to copy an unclickable link from the video to a pornographic site. The advertiser-friendly guidelines put "links to paid subscription adult content platforms" among content that gets no ad earnings. The spam policy adds "off-platform diversion": content made solely to drive viewers off YouTube, especially to sites with prohibited material such as adult content.
That leaves a narrow middle: offers adult enough to interest these viewers without being porn. My domain led to dating offers. YouTube's policy pages don't say either way where dating stands.
The domain opened a dating prelander. It shows a "Pick Your Favorite" grid, filled at random with girls' profile pictures I had pulled. After a choice, a "Searching videos across platforms" loading screen runs, then a button sends the visitor through a tracker to the offers.
Three affiliate providers sat behind it. The domain was burned into every video, and the channel page and ten of the first eleven video descriptions linked it too.
The burned-in domain leads to a dating prelander and three providers. The avatars are placeholders.
Even there, the fit was loose. A dating offer is a step away from what people came to watch. It converted poorly: about $50 over the whole run.
Why I stopped
The last upload went out on 2 November 2025. I stopped to focus on this blog and on other projects that bring in more.
I'm sure this niche has other opportunities. The views are there; the open question is the offer.
For another automation project, read how
I automated 2,700 Twitter accounts.
