Browse public AI image workflows shared by the Prompt Builder community.
A controlled prompt rescue using the same model, LoRA, and settings. Controls: seed 559655067, 9 steps, Euler, CFG 1, 1024x1024, and an empty negative prompt. The rewrite separates both hands from two continuous kimono sleeves while retaining the rainy cyan-magenta alley. Trade-offs: the hands became raised, the second
An AI-generated neon street study that separates the exact recorded baseline into scene, camera, wet reflections, motion-blur, bokeh, film-grain, and color-accent clauses. Camera wording describes the requested visual style. The current example and recorded Draw Things checkpoint, seed, steps, CFG, and dimensions are s
By strxet.nyc
Synthetic metadata fixture for software acceptance testing; not a generated-art showcase.
By Paulo Mendes
An AI-generated architecture study with modular clauses for a coastal concrete pavilion, vertical lines, geometric shadows, diffuse skylight, wide framing, concrete texture, and negative space. The camera reference is a prompt style cue, not a record of physical capture. The existing sample was generated in Draw Things
An AI-generated macro study with an installable library of the exact recorded prompt clauses: brass gears, surface wear, soft window light, a macro camera cue, film grain, and shadow depth. Camera and film names are prompt style cues, not claims about physical equipment. The existing sample was generated in Draw Things
By ember.lens
A compact official workflow for writing FLUX prompts as a photographic brief instead of a tag pile. Replace each clause with your own intent, generate a baseline, and add detail only where the result is ambiguous.
An official Illustrious/NoobAI template that keeps the quality stack short and the composition explicit. Use it to test your checkpoint before adding artist or character-specific tags.
An official safe-content Pony starting point. Establish rating, source, subject, pose, framing, and background before adding a visual style. Checkpoint behavior varies, so treat the settings as a reproducible baseline rather than universal best settings.
An official model-agnostic diagnostic workflow. Save the original recipe, reduce it to subject and composition, restore light and camera, then add style last. Publish only real before/after results and record every setting needed to reproduce them.
An official, intentionally plain SDXL starting point for controlled tests. Keep the seed and settings fixed, generate the baseline, then change only subject, light, camera, or style. This is a template, not a claim that one setting wins on every checkpoint.