README.md

metadata

title: 'SD Dataset Automaker: Scrap image → Face similarity → +Tag'
emoji: 🎴💆📦
colorFrom: indigo
colorTo: pink
sdk: gradio
sdk_version: 6.0.1
app_file: app.py
pinned: false
license: mit

🎴 SD Dataset Automaker

Anime character dataset generator for LoRA/fine-tuning.

⚡ Quick Start

1. Find show on fancaps.net → copy URL
2. Upload 3-5 reference faces of target character
3. Run → download ZIP with images + captions

🎛️ Parameters

Param	Default	Description
URL	—	fancaps.net/.../showimages.php?ID-Name
Refs	—	3-5 face images of target character
Character	auto	Output name & tag (optional)
Max Downloads	200	Frames to scrape
Similarity	32	Lower = stricter (25-30 strict, 35+ loose)
Trigger Word	—	Prepends to all captions

🔌 MCP Server

SSE: https://nekochu-sd-dataset-automaker.hf.space/gradio_api/mcp/sse

Claude Desktop config

{
  "mcpServers": {
    "sd-dataset-automaker": {
      "command": "npx",
      "args": ["mcp-remote", "https://nekochu-sd-dataset-automaker.hf.space/gradio_api/mcp/sse"]
    }
  }
}

Python API (Gallery)

from gradio_client import Client, handle_file

client = Client("Nekochu/SD-Dataset-Automaker")
zip_path, log = client.predict(
    q="https://fancaps.net/anime/showimages.php?3092-Cowboy_Bebop",
    c="spike_spiegel",
    imgs=[
        {"image": handle_file("https://huggingface.co/spaces/Nekochu/SD-Dataset-Automaker/resolve/main/from_url_spike_spiegel1.jpg"), "caption": None},
        {"image": handle_file("https://huggingface.co/spaces/Nekochu/SD-Dataset-Automaker/resolve/main/from_url_spike_spiegel2.jpg"), "caption": None},
    ],
    files=None, mi=200, th=32, at="", an=True, mo=False, tv=False,
    api_name="/process"
)
print(f"ZIP: {zip_path}")

Python API (File - MCP compatible)

from gradio_client import Client, handle_file

client = Client("Nekochu/SD-Dataset-Automaker")
zip_path, log = client.predict(
    q="https://fancaps.net/anime/showimages.php?3092-Cowboy_Bebop",
    c="spike_spiegel",
    imgs=None,  # Skip Gallery
    files=[  # Use File input (simpler schema, MCP-friendly)
        handle_file("spike_ref1.jpg"),
        handle_file("spike_ref2.jpg"),
    ],
    mi=200, th=32, at="", an=True, mo=False, tv=False,
    api_name="/process"
)
print(f"ZIP: {zip_path}")

---

CPU runtime ~5-10 min | Based on Maximax67/LoRA-Dataset-Automaker