---
name: primitive-folk-cgi-restyle
description: Use for a supplied image reconstructed as intentionally awkward, primitive low-budget 3D while preserving subject count, broad layout, action, camera, scene type, major props, and dominant colors. Unlike preview-paint (relationship 0.61), choose this for deliberate geometry, rigging, texture, lighting, and rendering failure; choose preview-paint for a painted treatment. Unlike restyle-video (0.55), choose this for a source-image reconstruction with optional short continuation. Do not use for flattering low-poly art, VHS, pixelation, horror, or injury.
---
# Primitive Folk CGI Restyle

## Catalog

- Display name: Primitive Folk CGI Restyle
- Categories: Video
- Status: Ready
- Route label: Broad-anchor-preserving primitive-3D reconstruction
- Tone: green
- Short description: Restyle a source image into readable but deliberately crude early-3D imagery, with optional awkward motion continuation.
- Cover image: /skill-covers/primitive-folk-cgi-restyle.png

## What this skill does

- Preserves broad scene anchors while releasing exact anatomy, clean mesh separation, and polished rendering.
- Directs sparse geometry, broken proportions, stiff movement, coarse textures, asset reuse, and naive lighting as observable production traits.
- Protects people and animals from identity-category changes, horror, injury, or unsafe distortion.

## How to use

Upload a source image and state the desired strength, ratio, prompt-only preference, or optional motion intent. Canvas-capable Agents may retain the approved treatment note there; otherwise they provide it in chat.

#### Example

```text
把这张客厅合照改成笨拙的早期民间 3D 画面：保留三个人的位置、沙发和主色，但使用粗糙贴图、僵硬姿势和单一顶灯；不要恐怖或受伤效果。
```

## Workflow

1. Extract only subject count/type, broad arrangement, semantic action, camera, scene category, major props, and dominant colors.
2. Select a conservative or strong primitive-3D treatment. Specify low geometry, imperfect proportions, limited local contact faults, crude materials, sparse repeated background, and naive lighting without adding unsupported subjects or props.
3. Submit one image edit via the current verified image route with a durable `operationKey`; poll the accepted receipt and reconcile unknown outcomes.
4. Inspect broad-anchor retention and whether the result is genuinely crude rather than merely filtered, blurred, dark, or grotesque. Offer optional video only after the still is accepted.

## User-facing output

- Media type: One restyled image and optional MP4 continuation
- Default quantity: 1
- Content per artifact: Broad-anchor-preserving primitive-3D treatment
- Default layout: Source ratio unless overridden
- Model policy: Current verified image-edit route; catalog-selected video route only after still approval
- Downstream use: Stylized stills and short motion tests

## Route

- Upstream assets: One source image
- Downstream handoff: `$generate-video` for confirmed awkward-motion continuation