Stencil: Subject-Driven Generation with Context Guidance

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Chen, Gordon, Huang, Ziqi, Tan, Cheston, Liu, Ziwei
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916959769067520
author Chen, Gordon
Huang, Ziqi
Tan, Cheston
Liu, Ziwei
author_facet Chen, Gordon
Huang, Ziqi
Tan, Cheston
Liu, Ziwei
contents Recent text-to-image diffusion models can generate striking visuals from text prompts, but they often fail to maintain subject consistency across generations and contexts. One major limitation of current fine-tuning approaches is the inherent trade-off between quality and efficiency. Fine-tuning large models improves fidelity but is computationally expensive, while fine-tuning lightweight models improves efficiency but compromises image fidelity. Moreover, fine-tuning pre-trained models on a small set of images of the subject can damage the existing priors, resulting in suboptimal results. To this end, we present Stencil, a novel framework that jointly employs two diffusion models during inference. Stencil efficiently fine-tunes a lightweight model on images of the subject, while a large frozen pre-trained model provides contextual guidance during inference, injecting rich priors to enhance generation with minimal overhead. Stencil excels at generating high-fidelity, novel renditions of the subject in less than a minute, delivering state-of-the-art performance and setting a new benchmark in subject-driven generation.
format Preprint
id arxiv_https___arxiv_org_abs_2509_17120
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Stencil: Subject-Driven Generation with Context Guidance
Chen, Gordon
Huang, Ziqi
Tan, Cheston
Liu, Ziwei
Computer Vision and Pattern Recognition
Recent text-to-image diffusion models can generate striking visuals from text prompts, but they often fail to maintain subject consistency across generations and contexts. One major limitation of current fine-tuning approaches is the inherent trade-off between quality and efficiency. Fine-tuning large models improves fidelity but is computationally expensive, while fine-tuning lightweight models improves efficiency but compromises image fidelity. Moreover, fine-tuning pre-trained models on a small set of images of the subject can damage the existing priors, resulting in suboptimal results. To this end, we present Stencil, a novel framework that jointly employs two diffusion models during inference. Stencil efficiently fine-tunes a lightweight model on images of the subject, while a large frozen pre-trained model provides contextual guidance during inference, injecting rich priors to enhance generation with minimal overhead. Stencil excels at generating high-fidelity, novel renditions of the subject in less than a minute, delivering state-of-the-art performance and setting a new benchmark in subject-driven generation.
title Stencil: Subject-Driven Generation with Context Guidance
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2509.17120