Before You Click: Dataset for Visual Framing in YouTube News Thumbnails and Video Content

This dataset accompanies the master's thesis "Before You Click: Visual Framing in YouTube News Thumbnails and Video Content" (University of Amsterdam, 2026). It contains the cleaned primary corpus of 153 YouTube news videos covering the Venezuela–United States conflict, collected via YouTube Data Tools in April 2026, alongside computed visual and affective features extracted through a mixed-methods computational pipeline. Files include color analysis, text overlay detection, CLIP-based affect scoring, and visual divergence measurements for both the primary corpus and the longitudinal upload histories of three representative channels: CBS News, Times of India, and Meganoticias. Thumbnail images and video frames are not included due to copyright considerations. The analysis code is available at https://github.com/tessacalor/before-you-click-thesis.