PROMPT · Image Generation
Split Canvas Editorial Art Poster Prompt
A free GPT image 2 image-to-image prompt that turns any photo into a 3:4 editorial poster: the top half stays photorealistic, the bottom half becomes a tiny handmade mixed-media artwork on warm ivory.
Third prompt from Sania (@saniaspeaks_), made with GPT image 2 on ChatGPT, and by a wide margin her best-performing one.
The idea is a split canvas. Every uploaded photograph gets turned into a premium editorial art poster in a strict 3:4 vertical format, and the canvas is divided into two exactly equal horizontal halves.
The upper half is disciplined: a faithful, photorealistic rendering of the original. Same identity, same facial features, same proportions, pose, clothing, objects, composition, lighting, shadows, mood and natural colour. The only things allowed are sophisticated editorial colour grading and seamless environmental extension where it's needed.
The lower half is where it becomes something else. The same visual story, reinterpreted as a tiny handmade mixed-media artwork sitting in the middle of expansive warm ivory negative space, taking up no more than 10 to 20 percent of the lower section. Expressive ink sketching, layered gouache-like colour fields, collage textures, torn-paper edges, imperfect brushwork, visible fibers, soft pigment variation, charming human imperfections — while still carrying the recognizable silhouette, gesture, objects and emotional narrative of the original.
Two more instructions worth noting. Pull up to four dominant harmonious colours out of each photograph and rework them into a muted, sophisticated palette. And add typography only when it genuinely helps — a poetic title, a place, a date, a single word.
The stated target: a collectible contemporary art publication cover. Minimal, poetic, tactile, elegant, emotionally quiet, and still unmistakably connected to the photograph it came from.
Create a premium editorial art poster for every uploaded photograph, treating each image as its own independent composition and never merging multiple photos together. Use a strict 3:4 vertical format with the canvas split into two perfectly equal horizontal halves: the upper half should remain a faithful, photorealistic presentation of the original image, preserving the subject's exact identity, facial features, proportions, pose, clothing, objects, composition, lighting, shadows, mood, and natural colors, enhanced only with sophisticated editorial color grading and seamless environmental extension where necessary; the lower half should transform the visual story into an entirely different artistic interpretation—a tiny, carefully composed handmade mixed-media artwork centered within expansive warm ivory negative space, occupying no more than 10–20% of the lower section, using expressive ink sketching, layered gouache-like color fields, subtle collage textures, torn-paper edges, imperfect brushwork, visible fibers, soft pigment variations, and charming human imperfections while retaining the most recognizable silhouette, gesture, objects, and emotional narrative from the original photo. Extract up to four dominant harmonious colors from each photograph and reinterpret them in a muted, sophisticated palette. Add only occasional understated editorial typography when it genuinely enhances the composition, such as a poetic title, place, date, or single word. The overall result should feel like a collectible contemporary art publication cover—minimal, poetic, tactile, elegant, emotionally quiet, visually distinctive, and unmistakably connected to its original photograph. 

A small note on the original: this is an image-to-image prompt, and it's written to be run on a batch — "for every uploaded photograph, treating each image as its own independent composition and never merging multiple photos together" is there specifically to stop the model from blending several uploads into one poster. The 10-20% rule for the lower artwork is the part most people would get wrong without being told; the point of it is negative space. Two images came with the tweet, both 1086 x 1448. At the time of archiving it had 263 likes, 31 retweets, 202 bookmarks, 10,367 views and 38 replies. Original post: Sania on X