QSilk: Micrograin Stabilization and Adaptive Quantile Clipping for Detail-Friendly Latent Diffusion
This work addresses detail preservation and artifact suppression in image generation for users of latent diffusion models, but it is incremental as it builds on existing stabilization techniques.
The paper tackled the problem of high-frequency fidelity and rare activation spikes in latent diffusion models by introducing QSilk, a lightweight stabilization layer that combines micro clamping and adaptive quantile clipping, resulting in cleaner, sharper outputs at low step counts and ultra-high resolutions with negligible overhead.
We present QSilk, a lightweight, always-on stabilization layer for latent diffusion that improves high-frequency fidelity while suppressing rare activation spikes. QSilk combines (i) a per-sample micro clamp that gently limits extreme values without washing out texture, and (ii) Adaptive Quantile Clip (AQClip), which adapts the allowed value corridor per region. AQClip can operate in a proxy mode using local structure statistics or in an attention entropy guided mode (model confidence). Integrated into the CADE 2.5 rendering pipeline, QSilk yields cleaner, sharper results at low step counts and ultra-high resolutions with negligible overhead. It requires no training or fine-tuning and exposes minimal user controls. We report consistent qualitative improvements across SD/SDXL backbones and show synergy with CFG/Rescale, enabling slightly higher guidance without artifacts.