Background & Origin Story
Not all heroes begin their journey in glory. Our subject, affectionately known as Donnie, began life as a humble Santa Claus lawn decoration, spreading good tidings until his magic faded. Destined for the landfill, Donnie was rescued by my wife who saw potential in his metal frame. With orange and white ripstop, she transformed him into DJT wearing an orange prison jumpsuit. Thus reborn, Donnie began adventuring in local parks and political demonstrations, his new persona both a commentary and a spectacle. Over time, Donnie began to dream, and his dreams—like those of his namesake—are grandiose, unruly, and a little bit nasty.
Model Creation
To capture the surreal psyche of Donnie, we curated 25 images depicting his new form in various settings: posing in front of a Thin Blue Line flag, riding on top of a makeshift bomb, and wearing a crown during the No Kings protest. Using standard Kohya_ss training parameters—optimized for creative, poetic prompts and cinematic scenarios—we distilled the essence of Donnie’s subconscious into a model that generates vivid, dreamlike scenes.
Training Details
Dataset: 25 diverse, high-resolution images of Donnie in his Trump-prison-jumpsuit guise, spanning close-ups and wide shots.
Base Model: Flux dev, chosen for its flexibility in surreal and narrative-driven outputs.
Training Steps: 45 steps per image, totaling 1,125 steps for optimal fidelity and creative flexibility.
Network Dimensions: 32–32, balancing detail and imaginative output.
Workflow: Images captioned and tagged for scenario diversity, with iterative testing in Invoke AI to ensure dreamlike consistency.
Capabilities
The resulting LoRA model excels at rendering Donnie’s dreams—scenes of power, chaos, and spectacle:
Leading armies through storm-lit fields.
Seizing the Supreme Court in surreal legal dramas.
Declaring himself King of the World atop monumental ruins.
Enacting wild, cinematic fantasies inspired by both satire and subconscious longing.
Best Practices
Use the trigger phrase: d0nn13 inflatable figure.
Compose long, poetic prompts for best results.
LoRA strength: 1 for a balance between fidelity and creative freedom.
Steps: 30 - 50 depending on the level of detail in the scene.
Guidance: 4 to ensure you see all his glorious details.