Img2Prompt, developed by Methexis Inc., is a service that provides approximate text prompts based on input images. It is optimized for use with stable diffusion, specifically with the clip ViT-L/14 model. The generated prompts can be used to recreate similar-looking versions of the input image or painting [1].
Img2Prompt is hosted on Replicate, a platform that enables users to explore and run machine learning models. Users can access the service through its web interface by uploading an image, and the Img2Prompt algorithm will generate a text prompt based on the image’s properties [2].
The service is based on a slightly adapted version of the CLIP Interrogator notebook by @pharmapsychotic. Img2Prompt has been used in various applications, such as identifying which author would best write a poem about an image [3].
