OppaiOracle V1.1 (CPU) — anime tagger

Multi-label ViT trained from scratch at 448×448 on a cleaned ~5.9M-image corpus (19,294 tags). Drop in an image to see ranked tag predictions.

This Space runs on CPU. Each prediction takes ~10–30 s on the HF CPU tier (the 448² ViT is ~250M params). For faster turnaround, see the GPU Space at Grio43/OppaiOracle.

Read first: the model card on the model repo documents known noise patterns (color tags, hair-length boundaries, neckwear, missing-tag bias). Predictions are best treated as a fast first pass that a human reviews — not as ground truth.

0.05 0.95
5 200

Model: Grio43/OppaiOracle · Resolution: 448×448 · Tags: 19,294 · Activation: sigmoid (already applied inside the ONNX graph).

This Space runs the V1.1 (448×448) checkpoint on CPU. Match input resolution to the checkpoint you load — feeding 320 to V1.1 (or 448 to V1) hurts accuracy because the ViT position-embedding grid is fixed at load time.