In the three-hour intensive workshop ‘Image-based Generation of AI Images’, I explain how to generate images using existing images or a combination of text and image prompts. The focus is clearly on Midjourney, where we experiment with Blend, Describe, Image Prompts and Faceswap. We test the possibilities of Inpainting and Outpainting (aka Generative Fill / Uncrop / Zoom) on Midjourney, Photoshop and Clipdrop. We use Style Tuner, Upscaler and discuss the Midjourney parameters. I show the possibilities of Midjourney, Firefly and Magnific to work with image references to achieve consistent styles, personas and lighting conditions. We mix style codes to create new aesthetics. I also show how Midjourney differs from the open source models of Flux and Recraft.