Upload your YouTube thumbnail and receive a 0–100 rule-based quality score. The tool analyses contrast, colour vibrancy, brightness balance, and focal area complexity — all in your browser, using zero server requests. No AI model, no account, no upload.
The four heuristics in this tool target the most common low-level issues that hurt thumbnails at preview size — not the content or the concept. A thumbnail can score 80 on these metrics and still perform poorly if the subject matter does not match the audience's interests. Conversely, a thumbnail can score 50 and be a top performer if the concept is perfectly tuned for a specific niche.
Contrast — the luminance spread across the image. YouTube's home grid card is 320×180 pixels on a 1080p screen. At that size, low-contrast thumbnails merge into the background and become invisible. The heuristic measures the interquartile range of luminance values: a thumbnail with both very dark and very bright regions scores high; a flat, near-greyscale image scores low.
Vibrancy — average colour saturation. YouTube's recommended thumbnails tend to use bold, saturated colours because they stand out against the platform's white or dark background. Muted, desaturated thumbnails are harder to spot in a busy grid. This heuristic measures average HSV saturation across sampled pixels.
Brightness balance — how close the average luminance is to a comfortable mid-range (roughly 0.35–0.65 on a 0–1 scale). Very dark thumbnails lose detail in the shadows; very bright thumbnails appear washed out and low-contrast. Both extremes score lower.
Focal area complexity — an estimate of how much of the image contains high-frequency detail (edges, texture) versus flat, empty space. Good thumbnails tend to have one or two clearly defined subject regions surrounded by some breathing room, not a uniformly busy composition that overwhelms the eye at small sizes. This metric rewards that balance.
Is my image uploaded to a server or an AI model?
No. The analysis runs entirely in your browser using the HTML Canvas API and pure arithmetic. No pixels leave your device. There is no AI model call.
How is the score calculated?
The score is a weighted average of four rule-based heuristics: contrast (35%), colour vibrancy (25%), brightness balance (20%), and focal area complexity (20%). Each is computed from pixel-level statistics — luminance spread, HSV saturation, and edge-block density.
Why is face presence not evaluated?
Detecting whether a face is present requires either an on-device ML model (which would significantly increase the page weight) or a server-side AI call. Neither is compatible with the fully local, no-signup design of this tool in v1. The other four heuristics provide a useful proxy — a thumbnail with a clear face tends to score higher on contrast and focal area naturally.
Does a high score guarantee a high CTR?
No. CTR is influenced by many factors this tool cannot assess: the relevance of the thumbnail to the search query, the quality of the title, channel authority, and the browsing context. The score is a rule-based sanity check — it flags low contrast, flat colours, or muddy brightness that are known to hurt performance at small sizes. It is not a CTR prediction.
FatThumb generates high-contrast, vibrant 1280×720 thumbnails with your face. Start free — no credit card required.
Start free