Fooocus 2.5 introduced automatic Enhance workflows, improved masking, and a Pony preset. The latest official tagged release is v2.5.5, published August 12, 2024; the project is in limited long-term support with bug fixes only.
Update through the official launcher, or fetch the code and install its requirements manually if automatic updating fails. Sources: releases, project status, update instructions.
Fooocus latest version and maintenance status
Fooocus 2.5.0 was published on July 17, 2024; v2.5.5 remains the latest tagged release. Source: releases.
Fooocus is an SDXL project under limited support, with no current plans for newer model architectures. A model's availability elsewhere does not establish Fooocus support. Source: README.
The 2.5 release notes credit substantial work by mashb1t. For everyday operation, our Fooocus practical guide covers the interface. Source: v2.5.0.
To compare other tools, see the AI image generator overview.
What is new in 2.5
The initial v2.5.0 release added:
- Enhance: automatic image refinement using detection prompts and replacement prompts.
- Masking improvements: GroundingDINO/SAM changes and an editor for uploaded masks.
-
Model and preset changes: AnimaPencilXL v500,
DreamShaperXL_Turbo_v2_1, Film Photography Style V1, plus the Pony preset and style. - Configuration improvements: preset VAE support and a configurable default inpaint engine.
- Sampling and metadata changes: a restart sampler and persistent model metadata hash caching.
These changes are listed in the v2.5.0 release notes. Later 2.5.x releases added smaller improvements and fixes:
| Release | Notable change |
|---|---|
| 2.5.1 | Restored JPEG/JPG/WebP metadata reading; added saving only the final enhanced image and improved Enhance image ordering |
| 2.5.2 | Fixed positive prompts with certain styles; expanded input-image configuration |
| 2.5.3 | Changed loading to use weights_only; added control over style changes when describing images |
| 2.5.4 | Fixed configuration validation, directory searching, and Enhance debug-mask previews |
| 2.5.5 | Fixed an unsupported image-type problem in Colab |
Enhance: Correct and Improve
Enhance can upscale an image and refine detected regions automatically. Its detection prompt identifies the area to work on; separate positive and negative prompts describe the intended replacement. Source: Enhance documentation.
For example, try detecting a dress and describing the desired garment color. Treat the output as a candidate to inspect: check the mask and surrounding details before accepting the edit.
For a generated image, keep the seed fixed, open Enhance, and enable the refinement steps you need. For an existing image, open Input Image → Enhance and upload it. Source: Enhance documentation.
The default masking combination uses GroundingDINO to locate objects and SAM to segment them. Additional masking options are available. This separates finding the target from describing what should replace it. Source: Enhance documentation.
Upscaling can run before or after enhancement. The maintainer recommends the earlier position for small details, while enhancing first can suit larger regions such as clothing or backgrounds. Source: Enhance documentation.
New Models and Presets
The Anime preset uses AnimaPencilXL v500. Its current JSON selects animaPencilXL_v500.safetensors; an older model already present can be used through the preset's fallback list. Source: anime configuration.
The Realistic preset uses realisticStockPhoto_v20.safetensors with SDXL Film Photography Style V1 at weight 0.25. Source: realistic configuration.
The added pony_v6 preset selects Pony Diffusion V6 XL, its VAE, and the Fooocus Pony style. The preset sets the inpaint engine to None. Source: Pony configuration.
Fooocus Pony adds score_9, score_8_up, and score_7_up to the positive prompt. Its negative style contains score_6, score_5, and score_4; those tags are supplied by the style itself. Source: style definitions.
Playground v2.5 support and its preset arrived in v2.4.2, with further preset adjustments in v2.5.0. This distinction matters when checking which feature requires which version. Sources: v2.4.2, v2.5.0.
Other Notable Updates
The inpainting tab contains Enable Advanced Masking Features, the renamed mask-upload control. The mask inversion control also moved out of Developer Debug Mode, and uploaded masks gained editing functionality. Source: v2.5.0.
Presets can specify a VAE and default inpaint engine. Inspect the supplied configuration when changing models: checkpoints may require different inpainting settings. Sources: preset keys, Pony.
How to update
Update Fooocus through its launcher
Close Fooocus and back up custom presets, configuration, and any edited project files. For the official Windows package, launch run.bat; its generated command invokes Fooocus\entry_with_update.py. Source: launcher implementation.
For a Python installation, activate the environment you use for Fooocus, open its directory, and run:
python entry_with_update.py
The updater fetches the current branch from origin and attempts a fast-forward update. Startup then checks dependencies against requirements_versions.txt. Sources: updater, startup dependency handling.
Update manually if automatic updating fails
From the Fooocus repository directory, with its Python environment active:
git pull
python -m pip install -r requirements_versions.txt
python entry_with_update.py
git pull fetches and integrates code changes; the pip command installs the required packages. Installing requirements alone does not update Fooocus's source code. Sources: release update procedure, Git documentation.
For the Windows archive installation, open Command Prompt in the inner Fooocus folder containing config.txt. The release's dependency command uses the bundled interpreter:
..\python_embeded\python.exe -m pip install -r .\requirements_versions.txt
Run git pull there first if Git is installed, then restart through run.bat. If you cannot repair that installation, the release notes also offer downloading the official package again. Source: v2.5.0 update instructions.
Windows package names and run_update
The README links the Windows archive Fooocus_win64_2-5-0.7z. That archive name differs from the latest release tag because the launcher handles code updates. Sources: Windows download, updater.
The official launcher generator creates run.bat, run_anime.bat, and run_realistic.bat; it does not create run_update. Use the documented launcher or entry_with_update.py for the official package. Source: launcher generator.
Known issues
- Update failure: local changes or a diverged branch can prevent a fast-forward update. Inspect the terminal error and preserve your edits before resolving the checkout. Source: updater.
- Misleading success text: the updater prints “Update succeeded.” even after its exception handler. Read preceding messages when confirming that an update completed. Source: updater.
- Enhance compatibility: the maintainer recommends Speed or Quality for workflows using the inpaint engine; performance LoRAs can interact poorly with it. Source: Enhance documentation.
- Control inputs: ImagePrompt, PyraCanny, CPDS, and FaceSwap are documented as unsupported within Enhance steps, though they can help generate the starting image. Source: Enhance documentation.
For unexpected edits, inspect the detection mask first. Model-load errors MetadataIncompleteBuffer and PytorchStreamReader indicate corrupted model downloads. Source: README.
FAQ
What is the Fooocus latest version?
The latest official tagged release is v2.5.5, published August 12, 2024. Consult the official release list when checking a different installation or package. Source: releases.
Is Fooocus still maintained?
Its stated status is limited long-term support focused on bug fixes. The README does not announce a migration to newer model architectures. Source: project status.
Does Fooocus 2.5 add FLUX support?
No. The official project remains based on SDXL, and its README directs people interested in FLUX to other interfaces. Source: README.
Do I need a separate extension for Enhance?
No. Enhance is part of Fooocus 2.5.0 and later. Its built-in workflow accepts generated or uploaded images and uses configurable refinement steps. Sources: release, Enhance documentation.
Sources
- Official release history
- Fooocus README and maintenance status
- v2.5.0 features and update procedure
- v2.5.1 release
- v2.5.2 release
- v2.5.3 release
- v2.5.4 release
- v2.5.5 release
- v2.4.2 release
- Maintainer's Enhance documentation
- Anime preset
- Realistic preset
- Pony preset
- Fooocus style definitions
- Preset configuration keys
- Official Windows launcher generator
- Automatic updater implementation
- Startup dependency handling
- Git pull documentation
- Contributor profile
- AnimaPencil model page
- Film Photography Style model page
- Pony Diffusion model page
Last reviewed and updated: September 2026.
Top comments (2)
Great update! The new features in Fooocus 2.5 sound really useful. I appreciate the effort put into making the tool even better. Can't wait to try it out and see how it improves my workflow. Keep up the awesome work!
Just in case someone have the same issues, so you won't spend an entire day trying to fix this XD