Best Stable Diffusion Extensions for A1111 & Forge (2026)

Extensions add features to the Stable Diffusion web UI: automatic face fixing, pose control, per-region prompts, bigger images. This page lists the ones worth installing in 2026, what each one does, whether it still gets updates, and which ones Forge already includes.

First: A1111 or Forge?

Before picking extensions, know where your web UI stands (dates from each project's GitHub, checked 15 October 2026):

All of them use the same extension system, so A1111 extensions install the same way. The big difference: Forge, reForge and Forge Neo ship ControlNet and MultiDiffusion built in. Do not install the separate ControlNet or Tiled Diffusion extensions there; use the built-in versions.

The short list

"—" means the repository does not say either way; check its issues before installing on Forge.

Fix faces and hands: ADetailer

ADetailer detects a face, hand or person after the image is generated, masks it and inpaints it at higher detail. It is the usual answer to small, blurry or warped faces in full-body shots.

You can give ADetailer its own prompt (for example detailed face, green eyes) so the fix matches the character. If you also have ControlNet installed, ADetailer can use inpaint, scribble, lineart, openpose, tile or depth ControlNet models during the fix.

Control pose and composition: ControlNet

ControlNet guides the image with a reference: an OpenPose skeleton, a depth map, edges, lineart or a tile of the image itself. The A1111 extension's v1.1.454 (July 2024) added ControlNet Union model support; the repository has had no commits since then. On Forge, reForge and Forge Neo use the built-in ControlNet instead.

Different prompts per region: Regional Prompter or Forge Couple

Both stop colors and features bleeding between subjects (red hair on both characters, the shirt color leaking into the skirt).

• Regional Prompter works in A1111, Forge and reForge, and its September 2026 update added Forge Neo models such as Z-Image, Krea2 and Anima. Full syntax and templates: Regional Prompter guide.
• Forge Couple is built for Forge Classic and Forge Neo and targets different prompts at specific regions.

Random variation: Dynamic Prompts

Turns {red|blue|green} and __hair_color__ into a different prompt for every image, with combinatorial mode for "one image per option". The extension has not changed since May 2024 but still documents the syntax used by many tools. Syntax and ready-made wildcard files: wildcards guide.

Bigger images: Tiled Diffusion & VAE and Ultimate SD Upscale

• Tiled Diffusion & VAE generates or upscales large images (2K and above) with limited VRAM (6 GB or less) by working in tiles. Tiled VAE alone helps when decoding runs out of memory. It also has regional prompt control. Its license (CC BY-NC-SA) forbids selling the code; images you make are not restricted. Forge's built-in MultiDiffusion covers the same idea.
• Ultimate SD Upscale upscales in tiles with img2img, which lets you use a higher denoise (0.3–0.5) without many artifacts. With a 512×512 tile it runs on any graphics card. It has had no commits since March 2024.

For more ways to save VRAM, see the low VRAM guide.

Quality of life

• Tag Autocomplete suggests Booru-style tags as you type, which helps with anime models prompted in Danbooru tags.
• LoRA Block Weight sets a LoRA's strength separately for each block of the model, useful when a LoRA changes the style more than you want. Regional Prompter's README suggests it for LoRA problems in Latent mode.
• Infinite Image Browsing is a full image and video manager with search. It reads generation data from A1111, ComfyUI, Fooocus, NovelAI, SwarmUI and others, and runs as an extension or a standalone app.

Older extensions to think twice about

• AnimateDiff for WebUI: no commits since September 2024, and its author added a non-commercial license.
• Anything whose last commit is years old and has open issues about your web UI version.

An old extension is not automatically broken, but if it fails after a web UI update, a fix is unlikely to come.

How to install safely

1. Open Extensions → Available, press Load from, find the extension and click Install. Or use Install from URL with the repository link above.
2. Go to Installed and click Apply and restart UI.
3. Install one extension at a time, so you know which one broke something.

Extensions are Python code that runs on your computer with your permissions. Install them only from repositories you trust, like the ones linked above. If you run the web UI with --listen or --share so others can reach it, do not add --enable-insecure-extension-access: that flag turns the Extensions tab on regardless of those options.

Build your prompt

The prompt generator builds a prompt from subject, style and lighting choices that you can paste into any of these web UIs. It runs in your browser and does not generate images.

Primary sources

• AUTOMATIC1111 stable-diffusion-webui
• AUTOMATIC1111 wiki: Extensions
• AUTOMATIC1111 wiki: Command Line Arguments and Settings
• Stable Diffusion WebUI Forge
• Forge Neo / Forge Classic (Haoming02)
• reForge (Panchovix)
• ADetailer
• ControlNet for WebUI
• Regional Prompter
• Forge Couple
• Dynamic Prompts
• Tiled Diffusion & VAE
• Ultimate SD Upscale
• Tag Autocomplete
• LoRA Block Weight
• Infinite Image Browsing
• AnimateDiff for WebUI