LFIT

SKILL.md

---
name: lfit
description: "⚡ LFIT — Local HD image generation via stable-diffusion.cpp on Vulkan/iGPU. Free, private, on your hardware."
version: 1.0.0
metadata:
  openclaw:
    emoji: "⚡"
    homepage: https://github.com/jchandler187/lfit
---

# ⚡ LFIT — Local Free Image Tool

**Local HD image generation. Free. Private. On your hardware.**

LFIT runs stable-diffusion.cpp on your Vulkan-capable GPU. No cloud API keys, no per-image fees, no data leaving your machine. Three locked presets — pick by name, get consistent results.

## Quick Start

```bash
# Install
npm install -g @lowwattlabs/lfit

# Set up the server (first time only)
lfit-server setup

# Start the server
lfit-server start --daemon

# Generate
lfit --preset standard --prompt "a knight at a castle gate"
lfit --preset background --prompt "misty pine valley at dawn"
lfit --preset hero --prompt "ancient dragon, key art" --yes
```

## Presets

| Preset | Use for | Size | Time |
|--------|---------|------|------|
| `standard` | Characters, items, single subjects | 1024×1024 | ~2m30s |
| `background` | Scenes, environments, wallpapers | 1344×768 | ~2m30s |
| `hero` | Final/max-quality key art (no LoRA) | 1024×1024 | ~13 MINUTES |

## LoRA Self-Check

The #1 failure mode: the server ignores `<lora:>` prompt tags. If LoRA fails silently, you get a blurry base-SDXL image. LFIT detects this two ways:
1. **Small output warning** — image data under 1000 bytes
2. **Fast completion warning** — standard/background finishing under 30 seconds means LoRA didn't apply

If you see these, check: `curl -s http://127.0.0.1:7860/sdapi/v1/loras`

## Requirements

- Python 3.8+
- Vulkan-capable GPU (4GB VRAM min, 8GB for hero)
- sd-server binary with Vulkan support
- SDXL base 1.0 model from [HuggingFace](https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0)
- SDXL Lightning 8-step LoRA from [HuggingFace](https://huggingface.co/ByteDance/SDXL-Lightning)

## Pricing

Free. If it saves you time, [buy us a coffee](https://buymeacoffee.com/lowwattlabs).

## License

MIT-0 — Low Watt Labs ⚡