A mid-range gaming PC with an 8 GB graphics card has enough VRAM to run a full SDXL checkpoint without touching the cloud. That GPU sits idle while you pay a subscription to generate art on a remote server that logs every prompt and filters your output. Off Grid AI Desktop is a free, open-source app that runs Pony Diffusion V6 XL directly on your PC or Mac.
Free, open-source, runs fully offline.
On-device image generation in Off Grid AI Desktop. SDXL, Lightning, Turbo, and more, run from your chat.
Why Pony Diffusion V6 XL
Pony Diffusion V6 XL is a versatile SDXL checkpoint built for character work. It is strong at anime, stylized illustration, and original character art, and it stays consistent across poses and styles in a way base SDXL struggles with. It is one of the most popular community checkpoints for a reason: it does characters well.
It uses tag-based prompting rather than long descriptive sentences. You feed it comma-separated tags, often with quality and rating tags up front, and it responds to that structure. If you have used Danbooru-style tagging before, you already know the dialect.
We publish it in GGUF for stable-diffusion.cpp:
Pony Diffusion V6 XL on Hugging Face →
The GGUF build quantizes the weights to fit consumer hardware. One file, point the app at it, generate. No Python, no diffusers install.
What You Need
Two tiers. Match your machine.
| Tier | Windows | macOS |
|---|---|---|
| Minimum | NVIDIA GPU with 8 GB VRAM (or Vulkan-capable), 16 GB RAM, 12 GB free disk | M1, 16 GB unified memory, macOS 13+, 12 GB free disk |
| Recommended | RTX 3060 12 GB or better, 32 GB RAM | M3 / M4 Pro or Max, 24 GB+ unified memory |
SDXL checkpoints are heavier than SD 1.5, but the GGUF build keeps the download manageable. Leave memory headroom so the system does not swap during a render.
What Off Grid AI Desktop Can Do
The image studio lives inside a full local AI app. For character and illustration work, here is what you get.
- Generate character art and anime from tag-based prompts, fully on-device.
- Watch each step render live, with progress, ETA, and cancel.
- Bring a sketch or base image in with img2img to iterate on a pose or composition.
- Apply style presets (Anime, Sketch, Cinematic, and more) on top of the checkpoint.
- Keep every render in an artifacts gallery with a lightbox for comparing variations.
The same window includes local LLM chat, voice in and out, and a Hugging Face model browser, so you can build a tag list with a local model and render it without leaving the app.
How Hardware Acceleration Works
On Windows, Off Grid AI Desktop runs stable-diffusion.cpp on CUDA for NVIDIA, Vulkan for wider GPU support, or CPU when there is no usable GPU. An 8 GB NVIDIA card handles SDXL comfortably.
On a Mac, the engine uses Metal. Apple Silicon shares one memory pool between CPU and GPU, so weights are not shuffled across a bus, and a 16 GB MacBook can run a checkpoint that needs a discrete 8 GB card on a PC.
Quantization is what makes the file fit. The GGUF build trades a little precision for a much smaller download, and at SDXL resolution the difference is hard to notice.
Prompting Pony: Tags, Not Sentences
This checkpoint rewards structure. A few practical pointers.
- Lead with quality and rating tags the model was trained on, then your subject and style tags, comma separated.
- Keep tags concrete: character traits, pose, setting, art style. The model reads tags better than prose.
- Use the negative prompt to push away artifacts and unwanted styles.
- Generate at 1024x1024 native SDXL resolution, then upscale a keeper instead of rendering everything large.
Privacy: Stronger Than NovelAI
NovelAI runs on their servers behind a subscription, and your generations pass through their infrastructure. Off Grid AI Desktop flips that.
Your prompt never leaves your machine. The image is computed locally and saved to your disk. No account, no telemetry, no API key, no content pipeline you do not control. The app is AGPL-3.0, so the source is open to read. Disconnect from the internet and it keeps generating.
Getting Started
- Download or clone Off Grid AI Desktop from the GitHub repo.
- Install and launch it (Windows and macOS builds available).
- Open the model browser and download Pony Diffusion V6 XL, or grab it from Hugging Face.
- Open the image tab, select Pony Diffusion V6 XL, and write a tag-based prompt.
- Generate. Watch the preview, save the variations you like from the gallery.
# Example tag-based prompt
score_9, score_8_up, 1girl, short hair, hoodie,
city street at night, neon lights, anime style
What's Coming
- More community checkpoints curated in the model browser.
- Cross-device sync so your art moves between machines.
- Unified search across your generated artifacts.
The built-in model browser. Download text, vision, image, and voice models from Hugging Face, sized to your machine.
FAQ
Q: Is it really free?
Yes. The app is free and open-source under AGPL-3.0. The model is a free download. Image generation is not gated.
Q: Does it work offline?
Yes. Once the app and model are downloaded, disconnect from the internet. Everything runs on your device.
Q: Why does it use tags instead of sentences?
Pony Diffusion V6 XL was trained on tag-based captions. Comma-separated tags, with quality tags up front, give you the most control.
Q: How much VRAM do I need on Windows?
8 GB is the floor for SDXL. 12 GB gives comfortable headroom.
Q: Does it run on Mac?
Yes. The same engine runs on Metal on Apple Silicon. 16 GB unified memory is the floor.
Q: Is my data private?
Prompts and images stay on your disk. No account, no telemetry, no server we own ever sees them.
Run character and illustration generation on your own hardware, for free.
