Upload 2-3 thumbnails. AI-vision-based predicted CTR + pattern recommendations from the YouTube cluster playbook.
Upload 1-3 thumbnails. AI-vision analyzes hierarchy, readability at thumb size, curiosity gap with title, contrast, and legibility. Predicts CTR band and prescribes specific fixes.
Directional. The model evaluates against the 5 craft fundamentals that historically correlate with CTR — visual hierarchy, readability at thumb size, curiosity gap, color contrast, legibility — but actual CTR also depends on your channel's baseline, your niche's norms, and your audience's familiarity with you. Treat the band as "vs the average thumbnail in your niche", not absolute.
A/B/C test math: 3 variants gives the model enough signal to identify a winner with reasoning, without diluting the comparison. Beyond 3, decision fatigue sets in and the actionable advice gets generic. Run multiple sessions if you have a larger variant pool.
For most niches in 2026, 5-8% is solid, 8-12% is elite, 12%+ is once-in-a-channel. Banking on 8-12% as a baseline expectation is unrealistic; treat it as a stretch goal. The bigger lift is moving from <5% to 5-8%, which is where most thumbnail optimization work lands.
Often, yes — especially at thumb scale. Humans evaluate thumbnails at desktop size; the algorithm and most viewers see them at 200×112px on mobile. Common surfaces this tool catches: text below 80px on mobile, low-contrast text on busy backgrounds, fonts that anti-alias into mush at small sizes.
Kompozy automates everything this tool estimates — multi-format generation, brand voice, scheduling across 9 platforms. Free 500-credit trial, no card required.