Img2prompt is an AI tool that generates approximate text prompts to help users create images with similar attributes in a fast and efficient manner. The tool is particularly optimized for stable-diffusion (clip ViT-L/14) and uses machine learning algorithms to predict an image’s attributes and provide a range of approximate text prompts to help users generate more images that match their specifications. The tool is based on the open-source CLIP Interrogator notebook created by @pharmapsychotic and utilizes the OpenAI CLIP models to match an image to a variety of artists, mediums, and styles. The results of the comparison are then combined with BLIP captions to generate a text prompt that can be used to create additional images similar to the original. The tool can be run via an API or the GitHub repository, and predictions typically complete within 24 seconds and run on Nvidia T4 GPU hardward. Img2prompt is a user-friendly tool that requires copy-pasting text prompts and generates predictions within 27 seconde.
