Text Guided Image Colorization

by fffilonico.huggingface.fffiloni/text-guided-image-colorization

  • Remote endpoint
  • 1 tool

Tools

  • text_guided_image_colorization_process_image Colorize a grayscale or low-color image using automatic captioning and text-guided diffusion. This function performs image-to-image generation using a ControlNet model and Stable Diffusion XL, guided by a text caption extracted from the image itself using a BLIP captioning model. Optional prompts (positive and negative) can further influence the output style or content. Process Overview: 1. The input image is loaded and resized to 512x512 for inference. 2. A BLIP model generates a caption describing the image content. 3. The caption is cleaned using a filtering function to remove misleading or unwanted terms. 4. A prompt is constructed by combining the user-provided positive prompt with the caption. 5. A ControlNet-guided image is generated using the SDXL pipeline. 6. The output image's color channels (A and B in LAB space) are applied to the original luminance (L) of the control image to preserve structure while transferring color. 7. The image is resized back to the original resoluti

As the server's own listing gives them. We never call a tool.