# Hidden portrait in a flower garden Garden light and shadow follow a supplied face structure at different viewing scales. Web page: Other languages: [简体中文](https://promptseen.dev/zh-hans/prompts/hidden-portrait-flower-garden/index.md) ![Garden light and shadow follow a supplied face structure at different viewing scales.](https://promptseen.dev/media/a_284/hidden-portrait-flower-garden-detail-6dbaa15b7fd1.webp) Upload a high-contrast portrait guide and use a tool capable of strong reference or structure conditioning. ## Test record - Tested with GPT Image 2.5 - 1254 × 1254 · 1:1 - Tested 2026-10-29 - Reference image required - Updated 2026-10-29 ## Prompt · English ```text Create an optical-illusion landscape photograph, using the uploaded portrait only as a very faint low-frequency arrangement guide. Photograph an expansive natural English flower garden from a high oblique viewpoint: small irregular beds of pink and white roses, lavender, winding narrow gravel paths, many scattered shrubs, a few cypress trees, patches of dappled sunlight. At full resolution every patch is ordinary separate foliage, flowers, gravel or deep garden shadow. ONLY when viewed as a tiny thumbnail or with eyes squinted, the overall distribution of light and dark loosely suggests the supplied face. Break every facial contour into disconnected irregular plant clusters. No physical human face, skin, anatomical eyes, carved face, face-shaped paving, topiary head, bust, portrait mural or vertical face sculpture. Keep the main garden plane receding into the distance; prioritize a convincing garden over precise portrait likeness. No text. ``` ## How to use Upload a high-contrast portrait guide and use a tool capable of strong reference or structure conditioning. Inspect both full size and thumbnail. Plain text alone is insufficient; our recorded workflow supplies the guide image directly. ## Recorded result The example was generated with the built-in image tool. Its exact model ID was not returned. The recorded run contains the submitted prompt and ordered inputs. ## Result limitation The current result uses the supplied silhouette with the built-in image tool, not ControlNet. Individual flower beds, trees and paths form an oblique garden plane, while a thumbnail reveals the face. The face can still be noticed at full size; subtlety and likeness vary across generations. The earlier, stronger face-shaped result remains in the run history. ## Test prints ### 2026-10-29 · GPT Image 2.5 - Provider: OpenAI What we saw: Improved oblique garden plane: distinct cypress trees, flower beds and gravel paths remain individually readable at full size; thumbnail aligns to supplied face. The face is still noticeable at large size, so describe as a structural illusion rather than invisible-at-full-size. Built-in image conditioning, no ControlNet. #### Reference images 1. Silhouette ![Silhouette](https://promptseen.dev/media/a_231/hidden-portrait-flower-garden-reference-detail-68df57ea47a9.webp) [Full-size image](https://promptseen.dev/media/a_231/hidden-portrait-flower-garden-reference-original-dc285ba557fe.png) #### Output ![Garden light and shadow follow a supplied face structure at different viewing scales.](https://promptseen.dev/media/a_284/hidden-portrait-flower-garden-detail-6dbaa15b7fd1.webp) ### 2026-10-29 · GPT Image 2.5 - Provider: OpenAI What we saw: The supplied facial structure is recognizable at thumbnail size, while local regions consist of stone paths, foliage and flowers. Strong conditioning leaves the face visible even at full size; this is an advanced structural demonstration, not a subtle or text-only illusion. Prompt used for this print: ```text A lush flower garden with stone paths, dense foliage and patches of warm sunlight. Arrange the major light and dark masses to follow the supplied high-contrast silhouette, while making each local region look like natural plants, stones and shadows. At normal viewing distance it should read as a believable garden; when reduced in size, the hidden silhouette should become recognizable. Avoid explicit drawn facial lines and avoid text. ``` #### Reference images 1. Silhouette ![Silhouette](https://promptseen.dev/media/a_231/hidden-portrait-flower-garden-reference-detail-68df57ea47a9.webp) [Full-size image](https://promptseen.dev/media/a_231/hidden-portrait-flower-garden-reference-original-dc285ba557fe.png) #### Output ![Garden light and shadow follow a supplied face structure at different viewing scales.](https://promptseen.dev/media/a_230/hidden-portrait-flower-garden-detail-f31a9685f413.webp) ## Source and licence - Prompt: By Prompt Seen · Licence: CC BY 4.0 - Image: Creator: Prompt Seen · [Licence: CC BY 4.0](https://creativecommons.org/licenses/by/4.0/) · Generated image ## Inspiration & source evidence This prompt is an independent adaptation. The image shown here is our own recorded result. [Reddit · seniorbrinco](https://www.reddit.com/r/StableDiffusion/comments/15wdoqa/been_trying_these_optical_illusions/) · 2023-08-20 · [Whole post or image gallery](https://www.reddit.com/r/StableDiffusion/comments/15wdoqa/been_trying_these_optical_illusions/) · Post score: 1032 · Observed 2026-10-28 · [Indexed primary-source snapshot](https://www.reddit.com/r/StableDiffusion/comments/15wdoqa/been_trying_these_optical_illusions/) · Source prompt or excerpt available; independently adapted These figures belong to the original post or video. They are dated snapshots, not live counters or ratings of our image.