Compressing Image Style Training into a Single Model Forward
Quick Answer
This paper shows that The i2L framework enables efficient style transfer by predicting LoRA weights in a single forward pass, improving style fidelity and perceptual quality on datasets like Z-Image and FLUX.2.
Quick Take
The i2L framework enables efficient style transfer by predicting LoRA weights in a single forward pass, improving style fidelity and perceptual quality on datasets like Z-Image and FLUX.2. This approach eliminates the need for separate training for each style, supporting advanced features like multi-reference style fusion and controllable generation.
Key Points
- i2L predicts LoRA weights for immediate style instantiation without per-style optimization.
- Architecture includes an image encoder, learnable LoRA queries, and compressed decoding heads.
- Training on diverse style pairs enhances appearance cue preservation and reduces content copying.
- Experiments show i2L outperforms existing methods in style fidelity and prompt alignment.
- Supports asymmetric classifier-free guidance and composition with controllable-generation modules.
Paper Resources
Article Content
From source RSS / original summaryarXiv:2606. 13809v1 Announce Type: new Abstract: Diffusion-based style transfer must balance inference efficiency with stylization fidelity. Adapter-based methods are efficient, but they inject style as an external condition and can either weaken reference-specific appearance or copy reference semantics into the generated image. Optimization-based personalization methods such as LoRA internalize style more effectively, but require a separate training process for every new style.
We introduce i2L (image-to-LoRA), a framework that amortizes style LoRA training into a single forward pass. Given one or more reference images, i2L predicts LoRA weights for a text-to-image model, enabling immediate style instantiation without per-style optimization. The architecture combines an image encoder, learnable LoRA queries, and compressed decoding heads that generate adapted matrices.
Training on semantically diverse style pairs encourages the predictor to preserve appearance cues while suppressing reference-content copying. Experiments on Z-Image, FLUX. 2, and Hidream-O1 show that i2L improves style fidelity, prompt alignment, and perceptual quality over existing baselines. Because i2L produces explicit LoRA weights, it also supports asymmetric classifier-free guidance, multi-reference style fusion, and composition with controllable-generation modules.
Reader Mode unavailable (could not extract clean content).
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CV
See more →LLM-Guided ANN Index Optimization for Human-Object Interaction Retrieval
A phase-aware LLM agent optimizes human-object interaction retrieval, outperforming Optuna TPE by 33.3% and VDTuner by 34.2% on the HICO-DET benchmark. This method enhances throughput by 15.3x over UniIR and demonstrates strong transferability across vector database management systems.