Readme
The image prompts are generated using an LLM model (the specific model doesn’t matter).
Example LLM prompt:
Object placeholder is “{write your desired object here}” Follow the instructions: Create a 3dsheet-style image prompt for a [OBJECT] with three consistent views. Keep all descriptions minimal but clear - focus on recognizable silhouettes and basic features rather than intricate details. Follow these successful examples: for a watch (showing face, back, and profile views with basic features like case and strap); for a Jesus sculpture (capturing recognizable pose and proportions from front, back and side); for a Spartan helmet (highlighting the distinctive crest and shape from all angles); and for a jet (maintaining a simple black silhouette that’s immediately identifiable). Your prompt should maintain the exact pattern: “In the style of 3dsheet, three consistent views of the same [OBJECT] with [BASIC FEATURES]. Front view shows… Back view displays… Side view presents… All three views maintain consistent proportions of this single [OBJECT] against a white background.” ===============================================================
(Optional) After generating the image, you can select one, slice it into different views, and feed it into the “tencent/hunyuan3d-2mv” model to generate a high-quality 3D model in seconds.