# cafe still life with diffusion models recently, i moved to a new area and discovered a cafe there. today i randomly took a picture while sitting there and decided to try to generate similar images with diffusion models here's the original: ![[_media/IMG_20250309_153048.jpg]] and below, you can find a few images made by different models it's still quite clear that they are ai-generated, but the line gets more and more blurry --- ## imagen3 ![[_media/Pasted image 20250310004649.jpg]] ```instruction Subject: (Detailed glass sugar-dispenser:1.2) filled with (natural brown-sugar:1.1), (shiny silver metal lid and spout:1.1) positioned center-left, next to (frosted yellow-green glass vase:1.3) containing (dark burgundy leafy branches:1.2), (white menu card:1.0) with bold "COFFEE" text, leaning right side. Setting: (Cafe table-top:1.2), light-grey surface, (emerald-green wall:1.3) background, (soft diffused lighting), (daytime atmosphere), hints of (natural wood paneling) above. Style: (Photorealistic still-life:1.4), (naturalistic rendering), (high detail), (intimate cafe scene), (inviting atmosphere), (soft focus background). Camera: (Medium shot), (eye-level viewpoint), (50mm focal length), (shallow depth of field), (static camera). Negative prompt: [out of focus], [blurry], [grainy], [low quality], [artifacts], [distorted], [unrealistic colors], [overly saturated], [artificial lighting], [harsh shadows], [text illegible], [cluttered scene], [people], [food]. ``` i use poe: https://poe.com/Imagen3 ** ![[_media/Pasted image 20250310004735.jpg]] ** ![[_media/Pasted image 20250310004823.jpg]] ```instruction Create an extremely detailed, high-angle photograph of a white cafe table setting. The central focus is on three objects: a clear glass sugar dispenser with a shiny stainless steel lid and spout, filled with white and light brown granular sugar; a yellow-tinted glass vase with small air bubbles throughout, containing a small amount of clear liquid at the bottom and several dried reddish-brown plant stems with delicate, feathery leaves; and a white, matte paper menu card with sharp corners, printed with black ink text listing 'COFFEE' and other menu items in columns. The sugar dispenser is positioned to the right of the vase and slightly behind the menu card. The vase is to the left and slightly in front of the sugar dispenser. The menu card is angled towards the right edge of the frame and is positioned in front of both the vase and the sugar dispenser, partially overlapping them. The table surface is smooth, matte, and light grey to white. The background is blurred, showing a green painted wall and light brown wooden wall panels, suggesting a cafe interior. The lighting is natural and soft, emphasizing the transparency and reflections of the glass objects, the textures of the dried plant stems and sugar granules, and the matte surface of the menu card. Ensure extreme detail in the textures of glass, metal, paper, dried plants, and sugar. Capture the subtle light refraction and reflections on the glass surfaces and the matte finish of the paper and dried plants." ``` ** ![[_media/Pasted image 20250310004941.jpg]] ```instruction Create a photorealistic still life photograph, high-resolution (8K, 16:9 aspect ratio), capturing a tabletop scene in a cafe. The primary focus is a clear ribbed glass sugar dispenser filled with light brown sugar and a light yellow-green bulbous glass vase containing dried reddish-brown feathery grass, positioned side-by-side on a clean white laminate table. To the right, include a slightly angled white menu card with legible black text "COFFEE" at the top. Background features a vibrant bright green wall and a light brown wooden frame, suggesting a cafe interior. Lighting should be soft, diffused indoor light, highlighting the reflective glass surfaces and creating subtle shadows. Photorealistic rendering, emphasize texture details of glass, sugar, dried grass, and paper. Camera: medium focal length, slightly high-angle perspective for a clear tabletop view. Style: Photorealistic, clean, modern cafe aesthetic. ``` ## flux ![[_media/Pasted image 20250310004853.jpg]] ```instruction Create a photorealistic still life image from a slightly elevated eye-level perspective. The scene is set on a clean white table surface. In the foreground, arrange a clear glass sugar dispenser filled with white sugar and topped with a silver metal lid and spout, positioned slightly to the right. To its left and slightly in front, place a translucent yellowish-green glass vase containing a bundle of dried reddish-brown feathery branches and some clear liquid. To the right of the sugar dispenser, lean a white rectangular menu card with black printed coffee menu text against the dispenser for support, angled slightly towards the viewer. The background is softly blurred, depicting a green wall or window frame with a light brown wooden trim element visible at the top. The lighting is soft and natural indoor light, casting gentle shadows and reflections on the glass and metal surfaces. Maintain a shallow depth of field to keep the objects in the foreground sharp and the background out of focus ``` https://poe.com/FLUX-pro-1.1-ultra --- ## how to do it for this task, i wanted ai to reproduce an existing picture, so i do it in three steps: 1. extract image description from an existing photo or screenshot 2. generate a prompt for an image generation model 3. generate the image that is reminiscent of the original and consistent in total, that's what i call a [[prompt sequence]] to run it, i use a poe bot based on claude37r and a metaprompt: - bot: https://poe.com/imaginator-x365 - metaprompt: [[prompt - imaginator-x365]] - image generator: https://poe.com/Imagen3 feel free to use, adapt, and complain :) --- ## p.s. since recently, poe has a fun feature to `/repeat` the same command multiple times: ![[Pasted image 20250310010651.png]] use sparingly as it eats up your quota real fast :( ![[Pasted image 20250310010731.png]] --- // 10 mar 2025 #image