Kwang Moo Yi (@kwangmoo_yi)
2025-12-05 | ❤️ 45 | 🔁 8
Zeng et al., “NeuralRemaster: Phase-Preserving Diffusion for Structure-Aligned Generation”
You can LoRA fine-tune existing models to only destroy the magnitude of the image while noising, which then leads to a structure-preserving image/video editor. https://x.com/kwangmoo_yi/status/1997078462117400932/video/1
🔗 원본 링크
미디어
