Controllable image editing with FLUX.1: a practical system for masked editing and sketch-to-image
Abstract
Recent advances in diffusion-based models have made it possible to generate and edit high-quality images from text prompts. However, achieving precise and controllable edits in practice is still difficult. In many cases, text alone is not enough to describe exactly what should change in an image or what should remain unchanged. This project investigates controllable image editing using a self-hosted system that supports multiple interaction methods, including masked inpainting and sketch guided image-to-image generation. A local pipeline and lightweight interface were developed to allow users to edit images using brush masks, sketches, and text prompts. To better understand how these editing methods behave, a series of controlled experiments were performed by varying parameters such as guidance scale, infer ence steps, mask dilation, and mask feathering. The outputs were evaluated using both quantitative metrics (CLIP similarity, LPIPS, and SSIM) and qualitative inspection of the generated images. The experiments show that mask dilation is particularly important for successful object replacement, as it gives the model more context when generating new con tent. Prompt phrasing also has a strong influence on whether the model follows the intended edit. For sketch-guided image editing, the strength parameter plays a key role in balancing how much of the original image is preserved versus how much new content is generated. These observations highlight practical factors that…
Read the paper
Related studies
- Learning from First-Generation Qualitative Approaches in the IS Discipline: An Evolutionary View and Some Implications for Authors and Evaluators (PART 1/2)
- Commentary on the clinical and preclinical dosage limits of interstitially administered magnetic fluids for therapeutic hyperthermia based on current practice and efficacy models
- Prosthesis: New Technological Opportunities and Innovative Biomedical Devices
- The history and future of neural modeling for cochlear implants
- Ryanodine receptors: waking up from refractoriness
- Novel Transgenic Rabbit Model Sheds Light on the Puzzling Role of Matrix Metalloproteinase-12 in Atherosclerosis
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.