Visual & data

Visual taste from data

Use public engagement numbers to find the samples the audience actually rewards.

Paid $15 list, $12 until 2026-12-31 · one skill, yours for good. The outline, the steps and the file open after purchase.

Background

One line: aesthetics are screened by data, not guessed. High-interaction data is the screen, taste is the ruler; only the crossing of the two gives the reference for "what the public really buys *and* what really looks good".

Why it is built this way

The pipeline screens before it looks. Collection pulls covers plus exact interaction counts through the section id, scoring turns those into a spread score so a mid-size video with strong engagement can outrank a huge passive one, and --min-view 100000 drops the small samples that inflate their own like-to-view rate. Only after the ranked list exists does anything get downloaded. Reading a cover before the ranking is built turns the library into a mood board.

Two tracks cover two different losses. Pillow measures palette, light and dark, whitespace and contrast, and those numbers go straight into the visual-quality-gate thresholds. The vision pass reads the cover and writes the five genes plus one copyable rule per sample, which is the part a designer can act on. Step 5 sends the chosen rule into S3 of visual-quality-gate, so the reference becomes a named thing, and the finished asset is accepted by the gate that supplied the threshold.

Where it bites

• High data read as good taste A lurid thumbnail with a fat like count enters the library as a style rule, and the next batch copies the lurid part. The red line forces every candidate through the five-gene breakdown, and the machine track has to agree before the pattern gets written down.
• One library for every section Knowledge-section covers and food-section covers obey opposite rules on type and colour temperature, so a merged summary produces guidance that fits neither. The hard rule stores one file per platform and section, and the section id is passed on every collection run.
• The reference that was never named A designer says the result should feel premium, anchors nothing specific, and the revision loop never converges. S3 of visual-quality-gate demands one named thing such as whitespace or restraint, and style borrowed from two libraries at once fails outright.
• Covers with no provenance An image file saved with no platform, section, view count or collection date cannot be re-checked, and the library turns into a folder of pretty pictures. Samples without data provenance do not enter the library, and downloads stay in references/library/ rather than the workspace root.

When to use it

Use the interaction data on public social-platform pages (views / likes / comments / shares) as a screen to pick the visual samples the public actually votes for, break out the aesthetic rules per section × platform, and bank them into a queryable library that serves as the reference anchor when making cover images, inline images and video thumbnails. Use when the user wants to improve aesthetics with data, learn from viral covers, do data-driven aesthetics, get past weak aesthetic judgement, or asks how to make a cover people actually click, or needs a data-backed reference before producing visual assets.