Curated clusters that have been qualitatively evaluated and labeled.
Loading cluster data...
Surfacing White Thymos in Alt-Right Corecore Edits
This visualization explores recurrent visual symbols in alt-right TikTok "corecore" edits— hyperaestheticized video montages that mobilize affect through carefully assembled imagery. Drawing on Bharath Ganesh's concept of "white thymos" (symbols that beget white supremacist affects), this project uses computational methods to inductively identify and catalog the visual signs that distinguish alt-right corecore from its mainstream counterparts.
Clusters are generated using CLIP embeddings and HDBSCAN, then annotated through a interpretation of a Peircean semiotic framework: Text (image captions as object), Interpretant (audience comments revealing affect/meaning), and Sign(the composite images for each cluster). I also add audio info under Genre (music descriptors). All items are derived using log-odds ratios to surface cluster-distinctive terms.
Browse: Scroll through cluster cards showing representative frames. These composites are the "symbols."
Use the Selected Results tab for curated findings or All Results to explore everything.
Click: Select any cluster to open a detailed view with the composite image,
representative audio, and the full range of frames from most typical to most atypical. These frames are calculated
via Euclidean distance from the order-centroid frame, and occupy the 10th, 25th, and nth percentiles.
Listen and View: Each cluster has representative audio from the video most central to that cluster.
Click images to enlarge them.
Offensive Content: This research analyzes alt-right propaganda.
Images, audio, and text may include racist, sexist, antisemitic, and otherwise hateful or
disturbing material. Viewer discretion is strongly advised.
Work in Progress: This is an ongoing research project. Clusters may contain noise, and interpretations are preliminary.
Limited Sample: Current analysis is based on approximately 900 videos scraped
from alt-right-associated hashtags (hyperborea, agartha, vril, "another X classic").
Findings represent patterns within this specific dataset, and may change with larger samples.
Curated clusters that have been qualitatively evaluated and labeled.
Loading cluster data...