Sandwiched Slides, An AI Re-Edit

AI-Modified: Exploring a generative AI editing feature on scanned images of sandwiched slides, a creative technique popular among photo enthusiasts in the 1980s.

<< Click on image for enlarged Lightbox display >>

To better understand AI editing, I am moving further down the road from my past couple of posts to look at regenerative AI edits of scanned images from sandwiched slides, a creative technique popular among photo enthusiasts in the 1980s. Sandwiched slides are made by layering two or more slide images into a single mount, then re-shooting or scanning the combined composition to produce a blended, double- or multiple-exposure-like image.

Stacking the slides together, however, increases the density of the combined image. Dark areas on one slide mask and block light from the other, making the sandwiched result significantly darker than either individual slide. The blend will inadvertently create a double-exposure effect as a natural feature of the new image.

While doing analog re-takes of these layered images (which would have been done way back in the 80s as the sandwiches were mounted) calls for experience and measured exposure controls, re-editing the images with current AI technology is an avenue where both nuances of technical restoration and creative blending can be equally applied.

My main interest here, aside from doing the nominal AI edits to remove dust, scratches, fungus, Newton rings, and adding color corrections and other adjustment bits to the sandwiched slide image, is in the regenerative ability of AI to remove the double-exposure look of the image and make the image look like a single-shot photo.

I did this by trial and error, and after a few prompt revisions and anxious moments waiting for the image to be generated, the double-exposure (sandwiched slide) image, as seen here (click to enlarge), is now an almost single-shot (which is still not quite near enough, though) masthead image.

And the prompt:

'Integrate image as a single shot photograph, eliminate the double exposure effect by properly delineating the trees and rocks on the shore as foreground elements and the sea and island as background elements, sharpen image, smoothen colors, adjust white balance, enhance and brighten foreground and background colors, maintain size and proportion'.

That's the experience. While the result is still short of what I expected, the exercise does show the way forward, and with added knowledge and insight, for better things to come.

As further examples, the images here were also AI-generated with similar prompt initiatives, except for the delineation of the front and rear elements, as they are different in each image.

To get the best of the exercise, technical enhancements of the image should include points on layer separation and depth (identifying foreground and background slides to separate overlapping elements), physical artifact removal (removal of dust, scratches, fungus, and Newton rings), color correction and realism (magenta or blue aging tint of old slides), and exposure balancing (optimizing contrast).

Sandwiched Slides, A Generative Re-Edit

Creative enhancement options, on the other hand, should consider or incorporate prompts for the exposure style (intentional 'double exposure look', or 'single-shot style'), surrealism and image interpretation, and style (e.g., 'grainy film effect', 'warm and vibrant colors of Kodak Kodachrome' or 'the moody green of a C41 cross-processed image').