Character Lab Reset
Larger isolated characters, rebuilt from scratch.
These are fresh transparent-background renders for the recurring cast. The selected character is shown at a much larger scale so we can judge silhouette, costume details, and face consistency before generating more scene art.
Large isolated transparent renders.
Totally different Nutty style options ready to compare.
Style lanes still available for the next round.
Nutty Nutwobble
Fast-brained squirrel kid who treats every inconvenience like a machine-shaped opportunity.
Focus on silhouette, facial consistency, costume shape, and whether this design will still read when inserted into busy story scenes.
Nutty Nutwobble
Use this as the baseline prompt for the next improvement pass. For stronger consistency, the next round should use the selected render and one anchor scene image as visual inputs.
Use case: illustration-story Asset type: reusable transparent-background character render Primary request: large full-body render of Nutty Nutwobble for recurring story illustration reuse Scene/backdrop: transparent background only Subject: Young squirrel, inventor lead, fast-brained squirrel kid who treats every inconvenience like a machine-shaped opportunity. Style/medium: premium modern children's book illustration, crisp painterly rendering, clean silhouette Composition/framing: character fills the frame, full body visible, centered, no cropping, readable at large preview size Lighting/mood: warm, inviting, highly readable Constraints: keep blue overalls, aviator goggles on forehead, eager grin and bright eyes, compact builder pouch or notebook when useful, transparent background, no text, no watermark Avoid: photorealism, blur, extra limbs, duplicated tail or wings, tiny figure, clutter
Source scenes still available for refinement.
All isolated characters at a more useful size.
Lock one anchor image
Pick the best existing scene image per character, then use it as the visual reference input for every new render.
Render transparent character sheets
Generate full-body character art on transparent background in 3-4 style lanes before choosing a final lane.
Freeze reusable prompts
Once a lane wins, keep the locked traits, pose language, and anchor inputs stable for future scene art.