Style Transfer / LoRA

Image & 2D Art

Open Source

• AL

  Abstract-LoRA: Unlocking Single-Image Style Transfer through Targeted U-Net Block Training Discovered by AI

  2026

  A single-image style transfer method built on SDXL and LoRA that improves style-content disentanglement by training LoRA weights on targeted U-Net blocks. It introduces background-aware disentanglement and abstract color representation extraction to handle complex backgrounds and preserve content structure.

  lora sdxl style-transfer

  GitHub · arXiv

• AA

  AnyStyle: A Single LoRA is Sufficient for Image-Guided Style Transfer Discovered by AI

  2026

  A streamlined framework for image-guided style transfer that uses a single style LoRA adapter combined with training-free structural guidance derived from the content image's internal attention layers to achieve a balanced content-style composition.

  diffusion lora style-transfer

  GitHub · arXiv

• CA

  CLeaR: A Unified Framework for Resolving the Leakage-Degradation Dilemma in Style Transfer Discovered by AI

  2026

  A training-free framework designed to resolve the leakage-degradation dilemma in style transfer by suppressing content leakage while maintaining high style fidelity. It operates across feature separation, feature-space grounding, and diffusion generation stages.

  content-leakage diffusion style-transfer

  GitHub · arXiv

• CK

  ComfyUI Krea2 Style Transfer Discovered by AI

  2026

  A training-free local style reference node for ComfyUI that replicates Krea2-style reference behavior, enabling single-image style transfer and two-reference blending with low content leakage.

  comfyui style-reference style-transfer

  GitHub

• CI

  Compressing Image Style Training into a Single Model Forward Discovered by AI

  2026

  Introduces i2L (image-to-LoRA), a framework that amortizes style LoRA training into a single forward pass of a neural network, enabling instant generation of style LoRA weights from reference images without traditional fine-tuning.

  diffusion lora style-transfer

  GitHub · arXiv · Hugging Face

• FF

  FreeStyle: Free Control of Style-Content Dual-Reference Generation from Community LoRA Mining Discovered by AI

  2026

  A dual-reference generation framework that leverages community LoRA models to construct large-scale style-content triplets, using disentanglement mechanisms to prevent semantic leakage from style references.

  disentanglement lora style-transfer

  GitHub · arXiv

• ST

  StyleGallery: Training-free and Semantic-aware Personalized Style Transfer from Arbitrary Image References Discovered by AI

  2026

  A training-free framework for arbitrary style transfer that enables region-based, fine-grained control over style elements. It performs unsupervised semantic clustering on content images using intermediate diffusion features to achieve semantically-aware and interpretable results.

  diffusion semantic-aware style-transfer training-free

  GitHub · arXiv

• TV

  TeleStyle V2: Beyond Content-Preserving Style Transfer with Self-Distillation and Distribution-Matching-Distillation Discovered by AI

  2026

  An advanced image and video style transfer model that achieves content-preserving style transfer and text-guided image editing in a single run. It uses a LoRA-driven DiT architecture with self-distillation and distribution-matching-distillation (DMD) to ensure high-fidelity style transfer with minimal content leakage.

  dit lora style-transfer video-style-transfer

  GitHub · arXiv · Hugging Face

• TC

  TeleStyle: Content-Preserving Style Transfer in Images and Videos Discovered by AI

  2026

  A lightweight model built on Qwen-Image-Edit designed for content-preserving style transfer in both images and videos using Diffusion Transformers.

  dit style-transfer video-style-transfer

  GitHub · arXiv

• CE

  ConsisLoRA: Enhancing Content and Style Consistency for LoRA-based Style Transfer Discovered by AI

  2025

  A LoRA-based style transfer method that improves content and style consistency by predicting original images and separating content from style learning.

  consistency lora style-transfer

  GitHub · arXiv

• KL

  K-LoRA: Unlocking Training-Free Fusion of Any Subject and Style LoRAs Discovered by AI

  2025

  A training-free LoRA fusion approach that seamlessly merges independently trained style and subject LoRA models. It uses a Top-K selection mechanism in each attention layer to balance the contributions of subject and style, preventing content leakage and style misalignment.

  diffusion lora style-transfer training-free

  Project · GitHub · arXiv

• NL

  NP-LoRA: Null Space Projection for Subject-Style LoRA Fusion Discovered by AI

  2025

  A training-free framework that uses asymmetric null-space projection to fuse subject and style LoRAs. It models how independently trained LoRAs interact in the shared parameter space, preventing content-style interference and leakage.

  lora lora-fusion null-space style-transfer

  arXiv

• OS

  Only-Style: Stylistic Consistency in Image Generation without Content Leakage Discovered by AI

  2025

  A training-free style transfer framework that mitigates content leakage from reference images during diffusion-based generation. It localizes content leakage during inference and adaptively tunes parameters within the reference image patches containing the subject to preserve style consistency.

  content-leakage diffusion style-transfer

  GitHub · arXiv

• SL

  SplitFlux: Learning to Decouple Content and Style from a Single Image Discovered by AI

  2025

  A framework designed to decouple content and style from a single image by fine-tuning the single-stream blocks of FLUX via low-rank adaptation (LoRA), allowing style and content to be recontextualized independently.

  disentanglement flux lora style-transfer

  arXiv

• SC

  Style Composition within Distinct LoRA modules for Traditional Art Discovered by AI

  2025

  Proposes a zero-shot diffusion pipeline that blends multiple artistic styles by performing style composition on predicted noise-clean latents during each timestep of the flow matching denoising process using distinct LoRA modules.

  diffusion lora style-transfer

  GitHub · arXiv

• SP

  StyleKeeper: Prevent Content Leakage using Negative Visual Query Guidance Discovered by AI

  2025

  A training-free style transfer method that prevents content leakage from visual style prompts in text-to-image diffusion models. It introduces Negative Visual Query Guidance and self-attention feature swapping to decouple style from semantic content.

  content-leakage diffusion style-transfer

  GitHub · arXiv

• SS

  StyleSSP: Sampling StartPoint Enhancement for Training-free Diffusion-based Method for Style Transfer Discovered by AI

  2025

  A training-free diffusion-based style transfer method that enhances the sampling startpoint to prevent content leakage and preserve the original layout of the content image.

  diffusion style-transfer training-free

  GitHub · arXiv

• UU

  USO: Unified Style and Subject-Driven Generation via Disentangled and Reward Learning Discovered by AI

  2025

  A unified framework for style-driven and subject-driven image generation that disentangles content and style representations to achieve high-fidelity style transfer and identity preservation.

  diffusion flux style-transfer subject-driven

  GitHub · arXiv

• CS

  CSGO Discovered by AI

  2024

  A framework for content-style composition in text-to-image generation that achieves high-fidelity style transfer and content preservation using a curated dataset of stylized image triplets.

  content-preservation diffusion style-transfer

  GitHub · arXiv

• DD

  DiffStyler: Diffusion-based Localized Image Style Transfer Discovered by AI

  2024

  A localized image style transfer framework that uses a text-to-image Stable Diffusion model-based LoRA to encapsulate style targets, combined with cross-LoRA feature and attention injection for precise style control.

  diffusion lora style-transfer

  GitHub · arXiv

• DS

  DiffSynth-Studio Discovered by AI

  2024

  A high-performance training and inference framework for diffusion models that supports instant 'Image-to-LoRA' (i2L) generation, allowing users to extract and apply custom styles from a single reference image without traditional training.

  diffusion-framework lora style-transfer

  GitHub

• IS

  Implicit Style-Content Separation using B-LoRA Discovered by AI

  2024

  A method that leverages Stable Diffusion XL (SDXL) and Low-Rank Adaptation (LoRA) to implicitly separate the style and content components of a single input image, enabling high-quality style transfer and content manipulation.

  lora sdxl style-transfer

  Project · GitHub · arXiv

• IP

  InstantStyle-Plus: Style Transfer with Content-Preserving in Text-to-Image Generation Discovered by AI

  2024

  An advanced style transfer framework that improves upon InstantStyle by introducing content-preserving mechanisms to balance style strength and structure preservation.

  content-preserving diffusion style-transfer

  GitHub · arXiv

• IF

  InstantStyle: Free Lunch towards Style-Preserving in Text-to-Image Generation Discovered by AI

  2024

  A tuning-free framework that decouples style and content from reference images using IP-Adapter to achieve precise style-preserving text-to-image generation.

  ip-adapter style-preserving tuning-free

  GitHub · arXiv

• LR

  LoRA.rar: Learning to Merge LoRAs via Hypernetworks for Subject-Style Conditioned Image Generation Discovered by AI

  2024

  A hypernetwork-based method for merging subject and style LoRAs. Instead of slow optimization-based merging, it uses a hypernetwork to predict the optimal merge weights, enabling high-quality subject-style conditioned image generation in real-time.

  hypernetwork lora lora-merging style-transfer

  GitHub · arXiv

• SA

  StyleShot: A Snapshot on Any Style Discovered by AI

  2024

  A generalized style transfer framework that enables high-quality stylized image generation from any reference image without test-time tuning. It utilizes a style-aware encoder trained on a curated dataset to extract expressive style representations such as shading, layout, and lighting.

  diffusion encoder style-transfer tuning-free

  GitHub · arXiv · Website

• ST

  StyleStudio: Text-Driven Style Transfer with Selective Control of Style Elements Discovered by AI

  2024

  A text-driven style transfer framework that introduces cross-modal Adaptive Instance Normalization and selective control to prevent overfitting to reference styles.

  diffusion selective-control style-transfer

  GitHub · arXiv

• US

  UnZipLoRA: Separating Content and Style from a Single Image Discovered by AI

  2024

  A method for decomposing a single input image into its constituent subject and style components, represented as two distinct and compatible LoRAs. This allows for independent manipulation, recombination, and high-fidelity style transfer with enhanced disentanglement.

  diffusion disentanglement lora style-transfer

  Project · GitHub · arXiv

• AR

  ArtAdapter Discovered by AI

  2023

  A text-to-image style transfer framework that utilizes a multi-level style encoder and explicit adaptation to capture high-level style elements such as composition and distinctive artistic expression.

  adapter art diffusion style-transfer

  GitHub · arXiv

• DR

  DreamStyler Discovered by AI

  2023

  A framework designed for artistic image synthesis and style transfer that optimizes style inversion with text-to-image diffusion models, capturing unique brushwork and color tones.

  diffusion style-inversion style-transfer

  GitHub · arXiv

• SA

  Style Aligned Image Generation via Shared Attention Discovered by AI

  2023

  A training-free technique that establishes style alignment across a series of generated images by sharing self-attention layers during the diffusion process.

  diffusion shared-attention style-aligned

  GitHub · arXiv

• SI

  Style Injection in Diffusion: A Training-free Approach for Adapting Large-scale Diffusion Models for Style Transfer Discovered by AI

  2023

  A training-free artistic style transfer method that manipulates attention features in pre-trained large-scale diffusion models to preserve content structure while transferring style.

  diffusion style-transfer training-free

  GitHub · arXiv

• SA

  StyleAdapter: A Unified Stylized Image Generation Model Discovered by AI

  2023

  A unified stylized image generation model that produces high-quality stylized images matching both the content of a prompt and the style of reference images without test-time fine-tuning.

  diffusion style-adapter tuning-free

  arXiv

• ST

  StyleDrop: Text-to-Image Generation in Any Style Discovered by AI

  2023

  A method that enables text-to-image generation in any specific style using a text-to-image model. It leverages adapter-based parameter-efficient fine-tuning (PEFT) to capture fine-grained details of a style, such as color schemes, shading, and brushstrokes, from a single or few reference images.

  adapter peft style-transfer

  GitHub · arXiv

• ZA

  ZipLoRA: Any Subject in Any Style by Effectively Merging LoRAs Discovered by AI

  2023

  A method to cheaply and effectively merge independently trained style and subject LoRAs to generate any user-provided subject in any user-provided style. It optimizes the merging weights to preserve both subject fidelity and style alignment.

  diffusion lora lora-merging style-transfer

  arXiv · Website

Proprietary / Industry Tools

• SC

  Scenario Discovered by AI

  2023

  A creative AI infrastructure platform designed for game developers to train custom Flux and SDXL LoRA models on proprietary art styles, ensuring visual consistency across game assets.

  game-assets lora-training style-consistency

  Website