4 Thumbnails I made 👨‍🍳 Which one is your favourite? DMs Open 📩

The best thumbnails start before photoshop 👨‍🍳 Always start by understanding: who is this for, and what should they feel the second they see it? DMs Open 📩

There’s a graph hiding on almost every YouTube video, and most people have never noticed it, even though it’s been there since 2022. Hover over the progress bar on any long video with enough views and you’ll see a wave appear. Most people who do spot it assume it’s a retention graph showing where viewers stayed and where they left. It isn’t. It’s YouTube’s “Most replayed” graph. The higher the line, the more often people rewatched or jumped back to that exact moment. A peak can mean the moment was so good people watched it twice. It can also mean it went by too fast and people rewound to catch it, or that viewers skipped straight to it and ignored everything before. Have you ever noticed this graph?

Recent W on a first video together 🔥

Thumbnail I made for UnspeakableTV 🧪👨‍🍳 DM me for Thumbnails 📩

Prepare to see all the "AI will take our jobs 😢" tweets

Before --> After 👨‍🍳👀

"Ai Is replacing designers" lol.

Recent Thumbnail for Unspeakable Reacts👨‍🍳 Love working with the GOATs DM me for Thumbnails 📩

Thumbnail I made for Kaye 👨‍🍳 Inspiration concept and brief of the idea DM me for Thumbnails 📩

Testing Thumbnails is costing you views Changing a shirt color or moving the text a few pixels isn't really a test. If the message of the thumbnail stays the same, the result almost always stays the same too. So here's the order I'd follow instead: 1) Composition first. Build three completely different layouts and concepts. Each of the concepts should also ideally convey a different emotion and vibe. 2) Elements second. Once you know which layout works, test what's actually in the frame. Not a color swap, but a different subject, a different object, a different thing for the eye to land on. 3) Styling last. Font, size, color, expressions etc.. only matter as something extra you can do once the first two are settled Timing matters just as much as what you test. The first one or two days of data are close to useless, because your most loyal subscribers click no matter what the thumbnail looks like. Start reading the numbers around day three or four, and wait until day seven before you trust them. After that first test, leave the video alone for at least one or two weeks before testing it again. The bigger the difference between your versions, the clearer the answer you get back. Now, there's a point where you should stop using YouTube's built-in test tool altogether. Once the video is somewhere between 1-2 weeks and a month old, switch to hard swaps. That means manually replacing the thumbnail yourself instead of letting the tool split your traffic. The reason is a blind spot in the tool. Someone who already saw thumbnail A and scrolled past it then gets shown thumbnail B, and because B looks new to them, some of them click. The data leans toward B for that reason alone, and YouTube ends up calling it the winner even when it isn't. After a hard swap, give it time and watch for the views. If they dip sharply, swap back. If they stay flat, the change didn't make a difference. If they climb, you've found your winner. And don't get excited about an early spike, because some viewers will think it's a brand new video and click just out of curiosity.

Thumbnail for Slob Clan 👨‍🍳 Composition Sketch --> Final Design 🪄 DMs Open 📩

YouTube announced it would generate thumbnails and titles for creators automatically. At its Made on YouTube event on September 23, 2026, YouTube revealed that Studio will now generate thumbnails and titles based on a video's content, matched to the creator's existing style. It will also automatically monitor thumbnail performance and swap out underperforming ones without the creator doing anything. YouTube CEO Neal Mohan described YouTube Studio as evolving into "an end-to-end creative partner to help across production, packaging, and channel strategy." (Source: TechCrunch, Tubefilter) Earlier this year, YouTube also updated its monetization guidelines to aggressively go after what it calls 'unsatisfying or off-putting content'. @MarioJoos , one of the most respected YouTube strategists (if not the most), broke this down publicly: the policy targets content that "relies heavily on manipulative formulas, mimic formats, and stories where the video feels interchangeable, or content designed purely to shock or surprise viewers for the sole purpose of getting views." He also noted that content using AI tools to create a truly unique piece with a creative narrative is fine, but generic templated content violates the guidelines. (Source: Mario Joos on X) So YouTube is telling creators: be original, be human, don't template. And then building a system that generates your packaging for you automatically. Both of these things are true at the same time. Mario has also been outspoken about AI regulation more broadly, co-founding Team Human alongside creators with over 130 million subscribers, meeting with members of the European Parliament, and calling for opt-in rather than opt-out AI data policies, stronger copyright protections for original creators, and stricter guidelines around photorealistic generative imagery. (Source: Mario Joos on X, Team Human campaign) His point isn't that AI is bad. His point is that the incentives need to be right. And that's exactly what I keep coming back to when I look at yesterday's announcement. A tool that auto generates packaging makes it cheaper and faster to produce content. That's genuinely useful for creators who are overwhelmed. But it also makes it cheaper and faster to mass produce low-effort content at scale, which is exactly what YouTube says it wants to stop. The creators who win with these tools will be the ones who use them as a starting point and then apply actual strategic thinking on top. The ones who use them as a replacement for thinking will produce the kind of content that YouTube's own policy is now targeting. The question isn't whether AI will replace the packaging process. It's whether the incentives that YouTube is building around automation are actually aligned with the quality standards it's simultaneously enforcing. What are your thoughts? Will this make AI slop better or worse?

Recent Retouching for a Thumbnail👨‍🍳

Recent Thumbnail for Slob Clan 👨‍🍳 What do you think? DMs Open 📩

Trust me it's not over 😭 And someone has to come up with the initial "sketch" anyways, AI can't do that yet

Officially over 1,000,000,000 views on YTJobs 🎉 ytjobs.co/@TadlosDesigns