niu-lai-translator

Supply the image and invoke the default crude_bootleg_cgi preset, which rebuilds the scene with coarse continuous character meshes, broken proportions, failed rigging, local clipping, and flat lighting. The low quality comes from production failures, not blur or pixelation filters.

It preserves only minimal semantic anchors: subject count and type, relationships, action verbs, indispensable props, one or two color cues, and broad pose topology. Composition, costume design, makeup, exact joint angles, and attractive silhouettes are deliberately simplified or broken.

Yes, you can lower reconstruction_strength from extreme to high or medium, raise identity_lock for better likeness, or choose gentler presets like clean_bootleg_cgi. You can also request scoped follow-up edits that change only named dimensions such as eyes, trees, fur, or lighting.

No, an input image is required unless you explicitly ask for a new scene. The workflow begins by confirming an image exists and requesting one if absent, since the process is a source-based reconstruction rather than text-to-image generation.

The skill treats overly faithful, attractive, or professionally lit results as failures under its inverted quality gate. It retries up to two times with targeted recovery prompts that restate anchor locks and degrade only the failed dimension.

It never invents new subjects, props, rain, or gestures absent from the source, and it does not change a real person's apparent age, ethnicity, or body category. It also avoids horror, gore, VHS damage, and mosaic effects as degradation shortcuts.