LocalForge AILocalForge AI
BlogFAQ

Fooocus vs RuinedFooocus

RuinedFooocus is Fooocus with the safety filters stripped out. That's the entire comparison. Same UI, same models, same hardware, same output quality. If you came here expecting a feature showdown, there isn't one — this is a single toggle's worth of difference packaged as a separate project.

Feature Comparison

Feature Fooocus RuinedFooocus
Runs Locally Yes Yes
Open Source Yes Yes
NSFW Allowed Yes Yes
Type Local / Offline Local / Offline

The Scenario

You've found Fooocus, liked how simple it was, generated a few images — and then hit a content filter. Maybe it blocked a prompt you didn't expect it to block. You searched for a workaround and found RuinedFooocus. Now you're wondering if it's better, safer to install, or hiding some catch. Here's what's actually going on.

The Quick Verdict

If you don't need unrestricted NSFW content, there's no reason to use RuinedFooocus. Fooocus is the original, actively maintained project. RuinedFooocus is a community fork that removes content moderation defaults. That's the whole story. Anyone telling you one produces "better" images than the other is either confused or selling something.

What People Think the Difference Is vs What It Actually Is

The fork's name implies Fooocus is somehow broken or limited and this version "fixes" it. In practice, here's the complete list of meaningful differences:

  • Content filters removed: RuinedFooocus ships with safety filters disabled or stripped out entirely. Fooocus has some NSFW filtering on by default, though it can be toggled off in settings.
  • NSFW-friendly defaults: RuinedFooocus adjusts default settings and presets to be permissive out of the box. Fooocus's defaults are conservative.
  • That's it. Same Gradio interface. Same model loader. Same style presets (with some additions in Ruined). Same generation pipeline under the hood.

If you were expecting architectural differences, different model support, or performance improvements — there are none. Both projects run the same Stable Diffusion checkpoints through the same inference pipeline.

The Part Nobody Asks: Can You Just Disable Filters in Regular Fooocus?

This is the question that makes the fork feel unnecessary. Fooocus's safety features aren't hardcoded in a way that prevents override. Community guides exist for toggling them off in the config. It takes about 2 minutes. So what RuinedFooocus actually provides is convenience — a pre-configured version where you don't have to find that setting yourself.

Whether that convenience justifies running a separate fork with slower updates is a fair question. For most people, the answer is no.

Setup and Installation

Both install the same way. Clone the repo, run the launcher, wait for models to auto-download.

  • Fooocus: Official repo, well-documented, one-click installers available. Works on Windows, Linux, Mac (MPS). Auto-downloads SD models on first run.
  • RuinedFooocus: Community repo, installation mirrors Fooocus. Documentation is thinner. You're trusting community maintainers rather than the original developer.

Hardware requirements are identical because the underlying engine is identical:

  • Minimum: NVIDIA GPU with 4 GB VRAM (expect slow generations and resolution limits)
  • Recommended: 6–8 GB VRAM (RTX 2060 or better)
  • Reality check: The "4 GB" minimum in both READMEs is technically true but practically painful. 6 GB is where things stop being frustrating.

If you want Fooocus with zero setup and no config hunting, LocalForge AI ships it pre-configured alongside other frontends.

Performance and Output Quality

Identical. There's nothing to compare here, and that matters.

Both run the same checkpoints (SD 1.5, SDXL, community models). Both use the same samplers. Both produce the same images from the same seed + prompt + settings. If someone shows you a "RuinedFooocus produces sharper images" comparison, they changed a setting — not a tool.

The only scenario where output differs: RuinedFooocus won't block or modify a prompt that Fooocus's safety filter would have caught. That's not a quality difference — it's a content restriction difference.

The Real Risk: Fork Lag

Here's where the skeptic in this comparison earns its keep. RuinedFooocus is a fork, not an independent project. Every time Fooocus updates — bug fixes, new model support, performance patches — RuinedFooocus has to manually merge those changes. In practice:

  • Updates lag by days to weeks depending on how active the community maintainers are at that moment.
  • Breaking changes in Fooocus can break RuinedFooocus until someone patches the fork. You're at the mercy of volunteer maintainers.
  • Fooocus development itself has slowed. lllyasviel's commit frequency has dropped, which means both projects get fewer updates. But when they do come, the original gets them first.

If Fooocus ships a critical fix, you might wait. If the fork's maintainers lose interest, you're stuck on a stale version with no migration path except going back to Fooocus anyway.

Who Actually Needs This Fork?

Be honest with yourself about this. The audience for RuinedFooocus is narrow:

  • You specifically need NSFW generation and don't want to modify config files yourself.
  • You want a one-click NSFW-ready setup without reading documentation.

That's a valid use case. But it's a small one. If you're generating landscapes, portraits, concept art, product mockups, or anything that doesn't trip content filters — RuinedFooocus gives you nothing Fooocus doesn't already provide.

And if you're a power user who needs unrestricted generation as part of a larger workflow, you're probably better off with ComfyUI or Forge — neither has content filters by default, both have larger ecosystems, and neither depends on a single fork maintainer.

The Bottom Line

Fooocus RuinedFooocus
Maintained by lllyasviel (original author) Community volunteers
Content filters On by default (configurable) Removed/disabled
UI Identical Identical
Model support Same Same (may lag)
Performance Same Same
Update speed First to receive Follows after delay
VRAM requirement 4–8 GB 4–8 GB
Price Free, open source Free, open source

One column in this table has a real difference. One. The rest is identical by definition — it's a fork.

About Fooocus

Minimal Stable Diffusion frontend for beginners. Type a prompt, get an image — nearly zero configuration required.

Visit Fooocus →

Full Fooocus profile →

About RuinedFooocus

Uncensored fork of Fooocus removing all content filters

Visit RuinedFooocus →

Full RuinedFooocus profile →

Frequently Asked Questions

Does RuinedFooocus produce better image quality than Fooocus? +
No. Both run the same models through the same pipeline. Image quality is determined by your checkpoint, prompt, and settings — not which fork you're using. Same seed, same prompt, same settings will produce the same image in both.
Can I just disable safety filters in regular Fooocus instead of switching? +
Yes. Fooocus's content filters can be toggled in the configuration. It takes a couple of minutes. RuinedFooocus essentially pre-applies this change for you — that's the main convenience it offers.
Is RuinedFooocus safe to install? +
It's open-source and community-maintained, so you can inspect the code yourself. The risk isn't malware — it's maintenance. Fork projects depend on volunteer contributors. If they move on, you're stuck on an outdated version with potential security and compatibility gaps.
What happens when Fooocus updates but RuinedFooocus hasn't merged the changes yet? +
You wait. Bug fixes, new model support, and performance patches arrive in Fooocus first. RuinedFooocus gets them when a maintainer manually merges the upstream changes — which can take days to weeks, or longer if the maintainers are inactive.
Are there better alternatives if I want unrestricted image generation? +
ComfyUI and Stable Diffusion WebUI Forge both have no content filters by default. They're more actively maintained, have larger communities, and offer more features. The only thing you trade is Fooocus's simplicity — both have steeper learning curves.