Thumbnails AI › Analysis · Published 2026-11-01
We analyzed 555 trending YouTube thumbnails: text, color and faces (Japan, 2026)
The short version
Of 264 regular videos on YouTube's trending list in Japan, 94% had text on the thumbnail and 89% of those used an outline or shadow. White was the main text color in 75%, and white × red, white × black and white × yellow together covered about 60%. A person appeared in 88% and a large face in 57%; in gaming videos an exaggerated expression appeared in 64% versus 30% elsewhere.
How we did it
- Data: 555 videos on YouTube's trending list in Japan on 1 November 2026, collected with the YouTube Data API.
- Breakdown: 264 regular videos (over 3 minutes) and 291 Shorts. 167 of the regular videos are in the Gaming category.
- Method: an image-recognition AI judged each thumbnail for text (presence, lines, color), people, face size, expression and overall brightness.
- Caveats: AI judgments can be wrong on small text or game characters. Trending videos are only videos that already grew, so this is not a comparison with videos that did not.
Unless noted, the numbers below are for the 264 regular videos.
Results at a glance
| Item | Share |
|---|---|
| Has text on the thumbnail | 94% |
| Text has an outline, shadow or band (of those with text) | 89% |
| Text uses white (of those with text) | 75% |
| Shows a person or character | 88% |
| Shows a large face (at least a third of the frame height) | 57% |
| Exaggerated expression (surprise, shouting, big smile) | 52% |
| Text avoids the bottom-right corner (of those with text) | 55% |
Only 6% had no text at all. Nearly every trending video puts text on the thumbnail and makes it readable with an outline.
Text: more than you would expect, but the biggest line is short
Counting every character in the image, the median thumbnail had 31 characters, and 68% had three or more lines — well above the usual "about 10 characters, two lines" advice. That count includes small text such as channel names, counts and dates.
The longest single line had a median of 17 characters. Most thumbnails use a short, large headline and add the details in small text.
- Keep the big headline short (aim for 10–15 characters in Japanese)
- Put series names, counts and dates on a separate line in small text
Color: white plus one more
White was the text color in 75%, followed by red (38%), yellow (26%) and black (24%). By combination:
| Text color combination | Share (of thumbnails with text) |
|---|---|
| White × red | 25% |
| White × black | 18% |
| White × yellow | 15% |
| White only | 6% |
| Red × yellow | 6% |
The top three "white plus one" pairs alone cover about 60%. White text with a black or red outline stays readable on any background, which is probably why so many channels choose it. Across the whole image, the most used colors were black (47%), red (45%) and white (43%). Overall brightness was "bright" for 45% and "dark" for 20%.
People and expressions: it depends on the genre
A person appeared in a similar share of Gaming (87%) and non-gaming (89%) videos. The big difference was expression.
| Gaming (167) | Other (97) | |
|---|---|---|
| Large face | 62% | 48% |
| Exaggerated expression | 64% | 30% |
| Text with an outline | 95% | 79% |
| Characters of text (median) | 34 | 24 |
| Most common text colors | White × red (32%) | White × yellow (20%) |
Gaming thumbnails typically show a large, shocked or shouting face with red text for energy. In other categories, natural expressions were more common and text was lighter. Matching your genre's conventions makes the content clear at a glance.
Shorts: medium brightness and smaller faces
The covers of the 291 Shorts looked different from regular videos:
- Text on 97% and an outline on 93% (about the same as regular videos)
- Large face: 31% (regular videos: 57%)
- Exaggerated expression: 21% (regular videos: 52%)
- "Medium" overall brightness: 55% (regular videos: 36%)
- Text colors: white (72%) and yellow (44%) are the most common
Shorts covers are often a frame from the video with large captions on top, rather than a designed close-up of a face.
Three things you can use today
- Add text, and always add an outline. 94% of trending videos have text, and about 9 in 10 of those have an outline.
- Make white your base text color, plus one more. If in doubt, pick white × red, white × black or white × yellow.
- Match faces and expressions to your genre. Big and dramatic for gaming; natural can work elsewhere.
To check that your own thumbnail still reads at phone size, use the free thumbnail preview checker, and resize or compress images with the resizer.
FAQ
Should I copy this data to get on the trending page?
No. This describes how videos that were already trending are made. It is not a comparison with videos that did not grow, and a thumbnail alone does not make a video trend. Treat it as a set of common patterns and measure the effect with your own click-through rate.
Is more text better?
Not necessarily. The character count includes small supporting text such as series names and dates. The longest single line had a median of 17 characters, so most thumbnails use a short headline plus small supporting text. Prioritize what is readable on a phone.
How were the thumbnails analyzed?
With an image-recognition AI, one image at a time, for text, colors, people, face size, expression and brightness. AI labels can be wrong, especially on small text or game characters.
This is an English summary of our original Japanese analysis. Data collected 1 November 2026.