Comfy Console

Comfy Console is a free tool that makes AI images on your own computer. It is a simple screen that sits on top of a program called ComfyUI. You do not have to understand ComfyUI to use it.

1 · What this tool is

Comfy Console makes AI images. You type words that describe a picture, press a button, and the computer draws it for you.

• It runs on your own computer. Your pictures and your words never leave your machine.
• There is no account, no login, and no cost.
• There are no limits on how many pictures you make.

Because everything runs on your own machine, what you create is yours — and your responsibility. Please don’t generate anything illegal.

2 · What you need before you start

You need three things. If you are missing some of them, the tool still opens — it will just hide the features it cannot use, so you only ever see buttons that work.

Thing 1 — ComfyUI (required)

ComfyUI is the free engine that draws the images. Install it first. The easiest way is the official ComfyUI portable version for Windows — the Full Setup guide walks you through it step by step.

Thing 2 — At least one model (required)

A model (also called a checkpoint) is the "brain" that knows how to draw. Without one, no pictures can be made. Download one checkpoint file and put it in this folder:

ComfyUI/models/checkpoints

Any SDXL checkpoint works. For anime-style art, a good choice is a WAI Illustrious checkpoint. You can find checkpoints on Civitai — download the file that ends in .safetensors.

Thing 3 — Extra parts, for the extra features (optional)

The basic tool (type words → make a picture) works with just Thing 1 and Thing 2. Some advanced buttons need extra parts. If you do not install these, those buttons simply will not appear — nothing breaks.

• Detailers (fix faces, hands, feet) need the ComfyUI-Impact-Pack and Impact-Subpack add-ons, plus their detector files.
• ControlNet (copy a pose or shape from a picture) needs the comfyui_controlnet_aux add-on and the controlnet-union-sdxl model.
• Upscaling (make pictures bigger and sharper) needs the ComfyUI_UltimateSDUpscale add-on and an upscale model.
• Erase (make something vanish from a picture) needs the comfyui-inpaint-nodes add-on and the big-lama model.

Add-ons are installed with a tool called ComfyUI Manager. You do not need to do this now. You can add them later, one at a time, whenever you want that feature — the Full Setup guide shows how.

3 · How to start it (in order)

Do these steps in order, from top to bottom.

1. Start ComfyUI first. ComfyUI must be running before the Console can talk to it. Start it with the setting that lets the Console connect (this is called the CORS setting). If you have a launcher file that includes --enable-cors-header "*", use that one.
2. Wait for ComfyUI to finish loading. It is ready when the ComfyUI window shows its normal screen and stops scrolling text.
3. Open Comfy Console. Double-click the comfy-console.html file you downloaded. It opens in your web browser.
4. Look at the top-right dot. When the Console connects to ComfyUI, the dot turns green and says connected. Now you can make pictures.

4 · Make your first picture

1. Pick a model. In the Settings panel on the far left, find the box labeled Checkpoint. Click it and choose your model from the list.
2. Type what you want. In the Prompt panel in the middle, in the box labeled Character, type a short description. For example: 1girl, red hair, blue dress, smiling.
3. Press Generate. Click the big Generate ▸ button at the top. You can also hold Ctrl and press Enter.
4. Wait. A bar fills up while the picture is being drawn. This takes a little while. This is normal.
5. See your picture. When it is done, the picture appears in the large Preview area on the right.

5 · The tabs, explained simply

Along the top bar, next to ⚙ Settings, there are a few tabs. A tab is a mode. You click one to switch what the tool is doing. (On Workflow and PNG the Settings column steps aside, since those modes don’t use it; clicking a tab also re-opens the Prompt panel if you’d hidden it.)

• ⚡ Quick — the normal mode. Type words, make a picture. Use this most of the time.
• 📂 Workflow — for advanced users only. It lets you load a ComfyUI file. You can ignore this.
• 🔍 PNG — drop a picture you made before, and it reads back the words and settings used to make it.
• 🩹 Edit — fix or change a picture. Drop an image here to remove a flaw, erase an object, make it bigger, or fix hands and faces.
• 🎓 Train — make your own character. You give it a set of pictures and it learns to draw that character on command. This one is an optional extra that needs a one-time install first — see Part 6 of the Full Setup guide. Until you install it, this tab stays hidden, and that is fine — you do not need it to make pictures.

6 · The main features, one at a time

You do not need any of these to make pictures. Turn them on only when you want them. Each one is a checkbox — click it to turn it on, click again to turn it off.

LoRAs

A LoRA is an add-on that teaches the model one specific thing — often a specific character or art style. Add one in the LoRAs box, and set its strength (a number, usually around 0.8). When you pick a LoRA, its trigger word (the word that turns it on) appears below it. Click the word to add it to your description.

Detailers (Face, Hands, Feet)

In the one Detailers box, turn on Face, Hands, and/or Feet. Each finds that part of your picture and re-draws just it to make it cleaner. They run in order (face → hands → feet) and make pictures take longer, so use them for pictures you want to keep.

ControlNet

This copies the shape of a reference picture — like a pose — while your words decide how it looks. Turn it on, choose a Type (Pose, Depth, Lineart, Canny, or Scribble), and drop a reference image. Your picture will match that shape.

Upscaling

Ultimate Upscale makes the picture much bigger and adds real detail. It is slow, so use it on pictures you already like. Keep it turned off while you are still experimenting.

The Edit tab

Drop a picture into the Edit tab, then pick a mode:

• 🖌 Paint-fix — paint over a flaw (like an extra finger) and it re-draws just that spot. The Avoid box lets you list things to keep out, and Soft blend (on by default) feathers the change so it doesn’t look pasted on.
• 🧽 Erase — paint over something you want gone (like a watermark on a plain background) and it fills it in from the surroundings.
• 🩹 Clone / Heal — the reliable way to remove something, with no AI: Alt-click a clean nearby patch to copy from, then drag over the flaw to paint that clean area over it. This is the tool for a stray mark, mole, or seam that Paint-fix keeps re-drawing. Press Save ▸ when it looks right.
• ⬆ Upscale — make any picture bigger and sharper.
• ✦ Fix Face / Hands — clean up faces, hands, or feet on a picture you drop in.
• ⛶ Extend — add more space around a picture (for example, to un-cut feet that got cut off).

In the Edit tab you can drop your image anywhere on the picture area, zoom with Ctrl+scroll, and undo a brush stroke with Ctrl+Z.

Forever mode

The ∞ Forever button keeps making new pictures on its own, one after another, until you press Stop. You can make each one different in three ways, all at once:

• 🔄 rotate characters — cycle through your saved characters, each filed into its own folder.
• 🔄 randomize size (in the Resolution box) — tick the sizes you like and it picks one at random each picture.
• 🎲 wildcards — each 🎲 rolls a fresh pose, outfit, camera angle, lighting, and more every time (see just below).

🎲 Wildcards — automatic variety

Next to each tag menu (Outfit, Pose, Camera, Art Style, Lighting, Time & Weather, Colour…) there is a 🎲 button. Press it and it drops that whole menu into your prompt as a “roll,” and every picture then picks a random one. Stack a few and Forever gives you endless variety hands-free — a random pose, from a random angle, in random lighting. The tool deals every option once before it repeats any, so you will not keep seeing the same few.

⭐ Keep & Explore — “more like this one”

Made a picture you love? Press ⭐ More like this. The tool locks that exact picture — same character, same pose, same everything — and keeps making fresh versions of just that, one after another. It is how you turn one lucky shot into a whole set of them. Press ⭐ again (or Stop) to go back to normal.

7 · Handy extras (small things that help a lot)

You do not need any of these, but once you know they exist they save a lot of clicking.

• Your work comes back — the tool remembers your prompt and your settings (model, size, detailers, and the rest) and restores them the next time you open it. Nothing to press.
• Save a character — the Character / Prompt saves box (in the Prompt panel) stores your whole prompt and its LoRAs under a name. Load it back with one click. Recipes (at the top of the Settings panel) save just your quality settings, so you can mix any character with any settings.
• Pull fields from another character — open ⇄ Pull fields, pick another saved character, tick which parts you want (outfit, pose, LoRAs…), and pull just those into what you are working on. Add to appends instead of replacing — great for building a character out of pieces of several.
• Copy button 📋 — on every prompt box and the checkpoint, copies that text to your clipboard.
• Refresh button 🔄 — next to Checkpoint / LoRAs. After you drop a new model into ComfyUI’s folder, click it to load the new file without restarting the app.
• Backup / Restore — in ⚙ Settings. Save all your Recipes, Characters and settings to a file, and load them back on any computer. Do this before moving the file.
• Save folder 📁 — sets which folder your pictures save into (it auto-fills from the character you load), so your library sorts itself.
• Keyboard shortcuts — Ctrl+Enter makes a picture. Ctrl+↑ / Ctrl+↓ on a highlighted tag makes it stronger or weaker (turns it into (tag:1.1)).
• Seed buttons — ♻ reuses the last seed (to make a small change to the same picture); 🎲 sets it back to random.
• Drag & drop — in the Edit tab you can drop an image anywhere on the picture area, not just the little box.
• ⤓ Update button — top-right of the app, next to Settings. It quietly checks if a newer version exists. If one does, the button lights up; click it and it downloads the new version to your Downloads folder. Then you close the app, replace your old comfy-console.html with the new one, and open it again.

8 · If something does not work

Here are the most common situations and exactly what to do. None of these mean the tool is broken.

9 · Words this page uses

ComfyUI

The free engine that actually draws the images. Comfy Console is a simple control panel for it.

Model / Checkpoint

The "brain" that knows how to draw. You need at least one.

Prompt

The words you type to describe the picture you want.

LoRA

A small add-on that teaches the model one thing, like a specific character or style.

Trigger word

The word that turns a LoRA on. The Console shows it for you.

ControlNet

A tool that copies the shape or pose of a reference picture.

Detailer

A tool that re-draws just the face, hands, or feet to make them cleaner.

Upscale

Making a picture bigger and adding detail.

CORS setting

A setting that lets the Console talk to ComfyUI. ComfyUI must be started with it.

Log

The text area at the bottom of the Console that tells you what is happening.

Seed

A number that decides the random starting point. The same seed with the same settings makes the same picture. -1 means "pick a new random one each time."