OmniPainter performs training-free and inversion-free stylized text-to-image generation by extracting style key/value statistics from a single noisy LCM forward pass and mixing them into self-attention during six-step denoising.
In International Conference on Pattern Recognition
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Training-free Stylized Text-to-Image Generation with Fast Inference
OmniPainter performs training-free and inversion-free stylized text-to-image generation by extracting style key/value statistics from a single noisy LCM forward pass and mixing them into self-attention during six-step denoising.