README.md

metadata

title: Anima Base v1.0 CPU
emoji: 🎨
colorFrom: indigo
colorTo: red
sdk: docker
pinned: false
license: other
short_description: Anime image generation with Anima Base v1.0 on CPU
tags:
  - text-to-image
  - anime
  - cpu
  - mcp-server

Anima Base v1.0 Image Generation (CPU)

Generate anime images with Anima Base v1.0 + Turbo LoRA, packaged as a single Q4_K GGUF for sd.cpp.

• Engine: stable-diffusion.cpp (compiled from source)
• Model: WeReCooking/Anima-v1.0-Turbo-AIO-Q4_K (1.67 GB)
  • Concatenates VAE + LLM (Qwen3-0.6B text encoder) + DiT + Turbo LoRA (from civitai.com/models/2560840)
  • Quantized to Q4_K via sd-cli -M convert
• Settings: CFG 2.0, sampler euler, scheduler simple, 12 steps (range 8-16)
• Hardware: CPU Basic, ~15 min per 512x512 generation at 12 steps

---

API

Python Client

from gradio_client import Client

client = Client("Luminia/Anima-2B-CPU")

result = client.predict(
    prompt="masterpiece, best quality, score_9, score_8_up, 1girl, solo, cyberpunk",
    negative_prompt="worst quality, low quality, score_1, score_2, score_3",
    resolution="512x512",
    steps=12,
    seed=-1,
    api_name="/generate"
)
print(result)  # (image_path, status_message)

MCP (Model Context Protocol)

This Space supports MCP for AI assistants (Claude Desktop, Cursor, VS Code).

MCP Config:

{
  "mcpServers": {
    "anima": {"url": "https://luminia-anima-2b-cpu.hf.space/gradio_api/mcp/"}
  }
}

---

CLI Usage

# Basic generation
python app.py infer -p "anime girl with silver hair, fantasy armor"

# With all options
python app.py infer -p "masterpiece, best quality, 1girl, cyberpunk" \
    -n "worst quality, low quality" \
    -r 768x512 -s 12 --seed 42 -o output.png

# No subcommand = launch Gradio UI
python app.py

---

Credits

• circlestone-labs/Anima base model
• Turbo LoRA v0.1 acceleration
• stable-diffusion.cpp inference engine