A strong design can’t rescue a weak frame, so choosing the moment to freeze comes before any styling.
YouTube Channel · gaming clips
A gaming clips channel that drew 47.4K views in its first 28 days.
A system scans gaming streams for short, shareable moments and drafts videos from the best ones. I make the final call on every video.
By my rough estimate, review takes me about 30 minutes a day, against about 3 hours by hand.
The result
47.4K views in the first 28 days live; 150 of 87,764 clips got my review.
Views are self-reported from YouTube Studio; the clip counts cover 84 days.
Why I built it
Find more. Watch less.
Running a gaming clips channel by hand is a grind: hours of streams to scrub through and no quick way to tell which five minutes are worth sharing. I wanted the system to do the digging and leave the judgment, is this moment actually good?, to me.
It searches broadly for channels I’d never think to check and keeps a watchlist of creators I already like. The system prepares; I decide. Every clip that goes out is one I approve.
How it works
Discover → Collect → Narrow → Review → Approve.
- 01 · Discover37,771channels found across three platforms
- 02 · Collect87,764clips gathered in the production window
- 03 · Narrow1,394clips made it into my review queue
- 04 · I review150clips got a human look for taste and quality
- 05 · Approve53clips I approved
What viewers see first
Every video gets a thumbnail drafted from its own best moment.
The system finds the action in each clip, applies the channel’s design, and drafts a set of thumbnail options. I pick the one that goes live.
Three more from the channel. The gameplay comes from the clips each video features; no creator names or faces are shown. Select one to see it larger.
The audience result
What happened once the channel went live.
In its first 28 days live, the channel’s videos drew real viewers. These are self-reported YouTube Studio numbers, kept apart from the production counts above. They show a real audience, not a promise that growth continues.
47.4K
views, first 28 days
304.7
hours watched, first 28 days
1,693
views a day on average
+20
subscribers, first month
What I learned about packaging
The clip matters. The thumbnail decides if anyone sees it.
Every clip gets a full set of thumbnail options, lettered A to Z. I choose one by letter; the system never picks for me.
Rust and Valorant each have their own approved thumbnail design, so a viewer can tell the series apart at a glance.
Every compilation lists each streamer with a timestamp and a link to their channel.
What it took to build
Seven parts work together behind those five steps.
See all seven parts
01
Finding. Three platforms, 37,771 channels searched, 117 creators on a watchlist.
02
Ranking. Weighs view growth, clip frequency, creator, language, and title keywords.
03
Watching. Picks the moment of action and keeps the streamer’s reaction in frame.
04
Assembling. Stitches clips, adds opening and closing cards, and balances the sound.
05
Naming. Drafts titles, descriptions, tags, and a ranked set of thumbnail options.
06
My review. I approve, reject, and pick a thumbnail in one spreadsheet.
07
Posting. Uploads to YouTube and TikTok on a schedule once I approve.
It does the digging. I make the calls.
The system handles finding, assembling, and routine checks. I still approve clips and choose the thumbnail.
One bad source can’t stop the queue.
Errors are logged and skipped, and the day carries on: 3,818 so far.
From an afternoon to half an hour.
About 30 minutes of review a day, against about 3 hours by hand, using the estimate above.
Track review and posting in one place.
Review decisions and published videos live in separate records, so the numbers are hard to compare.
Evidence record
Every number on this page, with its source.
Open the full evidence record
This record shows how big the clip pool was, what reached my review, and what was approved or used. Audience numbers come from YouTube Studio and are dated separately. The checked summary stays public; the larger working record remains private.
These counts describe different groups. Of the 1,394 clips in my review queue, 150 were reviewed (53 approved, 97 rejected by hand) and 1,244 are still pending. The 1,482 clips marked used and the 100 published videos come from separate production records, so they do not line up one-to-one. The 86,282 remainder is every collected clip not marked used (87,764 minus 1,482), including clips still waiting for review.
The claim
How big the clip pool was and how much of it reached a person.
87,764
clips collected
1,394
queued for review
150
reviewed by me
1,482
clips used
86,282
not marked used (includes pending)
The checks
What I approved or rejected by hand, and what is still waiting.
53
approved by hand
97
rejected by hand
64.7%
hand rejection rate
1,244
pending review
The scale
Channels monitored, production days, languages, and the separate publishing record.
37,771
channels monitored
117
creators on my watchlist
84
days in the production window
3
review languages
100
published videos (separate record)
Talk shop
Run a channel too?
The same idea works well beyond video: narrow with rules, and keep a person on the final call. If you’re building something like it, I’d like to compare notes.