New anime ViT tagger OppaiOracle X1.1 adds 110k photo dataset
grio43 · reddit · 2026-10-12
Reddit user grio43 released OppaiOracle-experimental-X1.1, an updated anime image tagging model open-sourced on Hugging Face.
- Expanded base anime dataset plus 110k photography photos (nature to museums), all human-screened; museum photos focused on weapons and armor to strengthen part recognition.
- Danbooru tags synced to August 2026; decapitated tags removed; general-tag min frequency threshold lowered from 500 to 400.
- Tuned ASL settings for clearer tag boundaries and applied cleaned tags from the author's data-cleaning project.
The author notes root issues of ViT taggers remain and will pause major tag-cleaning updates while returning to school.
More from Multimodal
- Photo-real Chinook monster truck: locking first frames with Magnific One — techhalla · 2026-10-12
- Nostalgia for the Deforum era: one AI video took 4 months and 100+ versions — ART-ficial-Ignorance · 2026-10-12
- AI-generated Korean mech blockbuster poster looks straight out of a theater lobby — Foreign-Original124 · 2026-10-12
- LocalLightChat, a lightweight Open WebUI alternative, adds ComfyUI image editing — PromptInjection_ · 2026-10-12
- Anomalous Model Browser: free ComfyUI plugin matches models by hash and relinks missing files — demongatanjieu · 2026-10-12
- Seedance 2.5 demo: 'the final boss was the sky itself' video made in CapCut — umesh_ai · 2026-10-12