unfk-imagelisted
Install: claude install-skill vonspurter/unfk-image
# unfk-image
Rewrite an image's metadata so it looks like it was taken by a real device instead of AI-generated. Three modes: **strip** (remove all EXIF/GPS), **fake** (replace with a realistic iPhone/Pixel/Canon/Sony template), and **random** (pick a random *real* device per file, mixed brands). Pixel data is never touched — only the metadata block.
## When to use
- User wants to "un-AI" an image, hide provenance, or make a photo look camera-shot.
- Privacy: scrub GPS/device info before sharing.
- Forensics/testing: inject known metadata to check detector heuristics.
## The tool: `unfk.py`
Located at `C:/Users/Gary/unfk.py`. Uses `piexif` (metadata read/write) and `pillow` (image I/O). Run via `uv run --with piexif --with pillow python unfk.py ...`.
### Commands
```bash
# Strip all metadata from a single file or directory (in place)
uv run --with piexif python C:/Users/Gary/unfk.py /path/to/image.jpg --mode strip
# Fake iPhone metadata on a whole directory (recursive), output to DIR
uv run --with piexif python C:/Users/Gary/unfk.py /path/to/dir --mode fake --device iphone --out /path/out
# Random real device per file, mixed brands (iPhone/Pixel/Samsung/Canon/Sony)
uv run --with piexif python C:/Users/Gary/unfk.py /path/to/dir --mode fake --random
# Put the photos somewhere else entirely (keeps originals intact)
uv run --with piexif python C:/Users/Gary/unfk.py /path/to/dir --mode fake --random --out /path/out
```
- `--mode`: `strip` | `fake` (required)
- `--device`: