Clockwork owl identity consistency
How consistently does one model preserve Herman’s identity across repeated unseeded trials?
clockwork-owlidentity-lockrepeatability
Research registry
Experiments
13 controlled tracks spanning identity, typography, materials, anatomy, composition, and failure analysis.
Beautiful failures
Which failed constraints still produce useful or aesthetically strong artifacts?
failure-museumcuration
Dense multi-constraint briefs
How many simultaneous visual constraints survive in one generated composition?
dense-promptablation
Hands, tools, and mechanical detail
Can anatomy and tool contact remain plausible around small mechanical parts?
handstoolsmechanics
Low-light cinematic composition
Can the subject remain readable under low-key lighting without crushed detail?
cinematiclow-light
Material realism
Can brass, enamel, glass, paper, clay, and textiles remain visually distinct?
materialsrealism
Negative-prompt behavior
Which exclusions measurably alter outputs, and which are silently ignored?
negative-promptablation
Round 1: clockwork owl across models
Which hosted model best preserves Herman's clockwork-owl identity, materials, mechanical coherence, and art-deco composition from one identical prompt?
clockwork-owlcross-modelround-one
Counting and spatial relations
Can the model obey exact object counts and left-right-inside-behind relationships?
countingspatial
Reference and style adherence
How strongly does a provided reference control identity, palette, and composition?
referencestyle
Transparent-background assets
Can the model isolate a useful production asset with a truly clean background?
transparentasset
Typography and poster legibility
Can the model render exact short text while preserving a premium poster composition?
typographyposter
UI and brand illustration
Can one visual identity survive icons, empty states, badges, and hero illustrations?
brandui