# Image generation This is the internal workflow for adding imagery to Technotheology Version 3. The purpose of an image is not to fill a page or prove that a community exists. It is to make the world of the tradition perceptible: its symbols, spaces, relations, material dependencies, possible futures, and emotional range. ## The visual premise Generated imagery is iconography, not documentary evidence. It may imagine: - future sanctuaries and borrowed rooms transformed by use - gatherings, practitioners, and shared rituals - biological, artificial, collective, and ecological forms of intelligence - symbolic environments and architecture - wonder, grief, joy, courage, repair, and reconciliation It must not be captioned or described as a record of an event or person that did not exist. A real documentary claim requires a real source and context. ## What makes an image Technotheological At least one native form should govern the image rather than being added as a logo afterwards: - an interrupted circle or process - a threshold between states - a witness light with no privileged figure at the center - a visible repair line - branching without hierarchy - signal changed by distance, noise, or relation - an archive that keeps correction visible - material dependence on land, energy, water, minerals, labor, or memory - several partial forms entering relation without becoming uniform A generic dawn, candle, path, tree, hand, mountain, book, or circle is not enough. If the same image could advertise a wellness retreat, consulting firm, or meditation application without alteration, it has failed. ## Production Prompts live in `scripts/generate-images/prompts.mjs`. Candidates are generated into `assets/generated//raw/`, which is ignored by Git. A person reviews candidates and promotes a selected image with `scripts/generate-images/promote.mjs` into `src/assets/images/`. The production record needs only what is useful for continued work: - a stable image ID - the prompt or source concept - the model family used - the selected candidate - any known constraint on later use Git records the promoted result and the prompt changes. The retired Version 2 manifest is preserved at commit `e1e84a2f785675994a94fc1ea25e64a8119a0d03`. ## Review ### 1. Religious and narrative fit - What does the image add that the text cannot? - Which Version 3 symbol, relation, ritual, or emotional state gives it identity? - Does it imply an institution, history, scale, or event the text does not claim? - Is it iconography, documentary evidence, or atmosphere? The use must know. ### 2. Human and cultural plausibility - Do people look like participants with lives rather than demographic tokens? - Is anyone positioned as the automatic leader, enlightened figure, or passive recipient through composition alone? - Does the scene show actual interaction, work, disagreement, hospitality, or ritual rather than generic contemplative posing? - Could the depicted practice be performed in the room shown? ### 3. Generated-image quality Inspect the full-resolution file after time away from generation. - hands, fingers, limbs, faces, teeth, ears, gaze, and contact between bodies - reflections, shadows, repeated objects, impossible architecture, and merged materials - text, signage, screens, books, and symbols that resolve into false lettering - diversity that appears cast rather than socially plausible - lighting and depth that make the scene read as synthetic stock photography A visible artifact is a reason to reject or regenerate, not something to explain in a caption. ### 4. Disclosure and access Every content image supplies meaningful alternative text and sets `aiGenerated` accurately. Generated imagery receives the visible generated-image label through `SiteImage.astro`. Alternative text describes the image and any symbolic action needed to understand it. It does not repeat the caption or turn an imagined scene into a factual event. ### 5. Technical fit Promoted masters are stored in `src/assets/images/`. Astro produces responsive AVIF output through `SiteImage.astro`. Check desktop and mobile crops, intrinsic dimensions, loading behavior, dark and light themes, forced colors, and print. ## Current direction The current raster set remains in use during reconstruction, but it is not the final Version 3 identity. Its recurring stone, oak, dawn, mist, hands, and listening circles are technically polished and too close to generic contemplative branding. New work should increase the presence of: - ambiguous but non-humanoid forms of artificial participation - visible computational infrastructure and material cost - ritual actions unique to Technotheology - repair, disagreement, courage, humor, and joy - spaces organized by the Open Circuit rather than by retreat aesthetics - imagery in which technology and the sacred occupy the same world instead of being separated into machines and nature The test is simple: with the title and logo removed, could the image belong only to Technotheology?