PixelHacker introduces a latent class guidance (LCG) paradigm that injects foreground and background embeddings into a diffusion model, training on 14 million image‑mask pairs and achieving state‑of‑the‑art structural and semantic consistency across Places2, CelebA‑HQ and FFHQ benchmarks.
TurboFill introduces a fast image‑inpainting model that trains a repair adapter on a few‑step text‑to‑image diffusion backbone, achieving state‑of‑the‑art results with only four diffusion steps while dramatically reducing computational cost.
Alibaba’s Mama Intelligent Creation team has open‑sourced a FLUX‑based ControlNet inpainting model that leverages a DiT‑backed Interleave design for superior repair quality, and an 8‑step LoRA‑Turbo model that cuts inference time three‑fold while preserving near‑original image fidelity, both now available on Hugging Face and ModelScope.
This article explains how Kuaishou Y‑Tech combines image inpainting, semantic editing, and advanced GAN techniques to accurately locate and remove facial wrinkles—especially neck wrinkles—while preserving realistic skin texture and achieving high‑quality results suitable for production deployment.
