This extension adds image-making nodes to Modly. The model makes realistic photos from plain text.
You do not need to know anything about AI. Pick a node, type what you want, press run.
| Node | What it does in one line | Do I connect anything? |
|---|---|---|
| Juggernaut XL | Makes a brand new image from words | No |
| Juggernaut XL Restyle | Redraws a whole image in a new style | Yes: image in |
| Select by Text | Picks out a part (like the sky) by typing its name | Yes: image in |
| Juggernaut XL Inpaint | Replaces only the white part of a mask | Yes: image + mask |
The three Juggernaut nodes share one model download. Select by Text needs one small extra model.
Note about Preview: the Preview node is not part of this extension. It is a built-in base node that comes with Modly, and you find it under Base in the node list — not under "Juggernaut XL". You do not download it from this extension, and it is not listed on the extension's page. See Node 5 — Preview below.
- Open Modly.
- Click Extensions.
- Click Install from GitHub.
- Paste this link and confirm:
https://github.com/iammojogo-sudo/juggernaut-xl_modly - Wait for setup to finish. Do not close Modly while it runs.
- Go to the Models page.
- Download Juggernaut XL.
- Want to select parts by typing words? Also download Select by Text.
- Wait until it says Installed.
About the popup: at the end of setup Windows asks for permission (a blue prompt that says Python / Unknown publisher). Click Yes. Then fully close Modly. This adds two small changes to Modly itself: the Preview node (a Modly base node, see Node 5) and sliders next to the number boxes. Neither is part of the extension's own node list. Modly reopens by itself when it is done.
- Clicked No by mistake? Close Modly, then double-click
add_preview.batin this extension's folder. - Want them gone? Close Modly, then run
remove_preview.bat. That restores Modly exactly as it was (removes the Preview node and the sliders).
- A computer with an NVIDIA graphics card (recommended).
- 6 GB of video memory minimum. 8 GB or more is much better.
- Some free disk space for the models.
- Mac works too, but it is slower.
Use this when you have no picture and just want one.
- Add the Juggernaut XL node to your workflow.
- In Prompt, type what you want to see. Example:
a golden retriever puppy sleeping on a red couch, photo - Leave everything else alone the first time.
- Press run.
- Look at the result with a Preview node (see Node 5).
That's it. You do not connect any wires to this node.
| Setting | What it does | What to put |
|---|---|---|
| Prompt | The description of the picture | Plain words. Name the subject, place, and lighting. |
| Negative Prompt | Things to keep out of the picture | Leave it empty. This model works better with nothing here. |
| Quality Steps | How long it works on the picture | 30 is fine. 20 is faster but rougher. 40–50 is slower and a bit nicer. |
| Prompt Guidance | How strictly it follows your words | 5. Lower (3–4) = more realistic. Higher (7+) = stranger, more forced. |
| Aspect Ratio | The shape of the picture | Pick Portrait for people, Landscape for scenery, Square for both. |
| Seed | The random number for this run | -1 = different every time. Type a number to get the exact same picture again. |
- Nothing looks like your words → your prompt is too short. Add details like place, light, and camera style.
- It crashes or runs out of memory → use a smaller image size, or close other programs.
- The picture is weird/blurry → raise Quality Steps to
40.
Use this when you have a picture and want it redrawn in a different style, like oil painting, anime, or winter.
This node changes the entire picture. It cannot keep one part the same. If you only want to change one area (for example the sky), stop and use Node 3 + Node 4 instead.
- Add the Juggernaut XL Restyle node.
- Drag a wire from your image output into this node's image input.
- In Prompt, describe the whole picture you want back, not just the
change. Example:
same dog, but in a snowy forest at night, cinematic photo. - Set Restyle Strength (see below).
- Press run.
| Value | What happens |
|---|---|
| 0.3 – 0.5 | Stays close to the original. Only small changes. |
| 0.6 – 0.8 | Changes a lot but you can still tell it is the same scene. |
| 0.9 – 1.0 | Almost a brand new image. The original barely survives. |
Start at 0.7. Too strong? Lower it. Too subtle? Raise it.
| Setting | What to put |
|---|---|
| Prompt | Describe the whole result you want. |
| Negative Prompt | Leave empty unless something keeps appearing that you hate. |
| Quality Steps | 30. |
| Prompt Guidance | 5. |
| Seed | -1 for random. Same seed + same settings = same result. |
Use this when you want to grab one thing in the photo — the sky, the car, a person's shirt — without painting over it by hand.
- Add the Select by Text node.
- Connect your image to it.
- In What to select, type the thing you want. Examples:
skythe bicyclethe woman's jacket
- Press run.
- The output is a black-and-white mask:
- White = the part you asked for.
- Black = everything else.
Plug that mask into the Inpaint node (Node 4) to actually change anything.
| Setting | What it does | What to put |
|---|---|---|
| What to select | The words that describe the target | Short and specific. sky beats the part above the trees. |
| Threshold | How picky the selection is | 0.5 to start. Too much selected → raise it. Not enough selected → lower it. |
| Invert | Flips white and black | No normally. Yes if you want to keep the thing and change everything else. |
| Expand (px) | Grows the white area outward | 8. Stops thin seams around the edges. |
| Feather (px) | Softens the edges | 4. Raise to 10–16 if the change has a hard visible line. |
- It selected the wrong thing → be more specific:
the red car, notcar. - It selected too much → raise Threshold.
- It selected only a small piece → lower Threshold, or turn Expand up.
- The change has a hard edge → raise Feather.
Use this to fix or replace just one spot — swap the sky, remove a person, change a shirt color — while everything else stays untouched.
- Add the Juggernaut XL Inpaint node.
- Connect the original picture to the first input, labeled Image.
- Connect a black-and-white mask to the second input, labeled
Mask (white = replace).
- Easiest mask: use Select by Text (Node 3).
- Any hand-made black-and-white PNG works too.
- In Prompt, describe what should appear inside the white area only.
Example:
bright orange sunset sky with clouds. - Press run.
Rule to remember: WHITE changes, BLACK stays.
| Option | Use it when |
|---|---|
| Connected mask (default) | You plugged a mask into the second input. This is the normal choice. |
| Transparency: replace subject | Your image already has a see-through background and you want to change the subject area. Works with Image Editor → Remove Background (a separate extension — see FAQ). |
| Transparency: replace background | Same, but you want to change the background. |
Leave it on Connected mask unless you are using a transparent PNG.
Careful: the Transparency options ignore the mask you connected. They only look at the see-through area of the image. If your image is a normal, fully opaque picture you get an error like
This image has no see-through area— even though a mask is connected.
- Have a mask connected? → set Mask Source = Connected mask.
- Want to use Transparency? → run Image Editor → Remove Background on the image first (that node lives in the separate Image Editor extension — see FAQ), so it actually has a see-through area.
| Setting | What it does | What to put |
|---|---|---|
| Mask Source | Where the mask comes from | Connected mask. |
| Prompt | What goes in the white area | Describe only that spot, e.g. wet cobblestone street. |
| Negative Prompt | Things to avoid | Leave empty. |
| Inpaint Strength | How hard that spot is redrawn | 0.9. Lower (0.6–0.7) keeps more of the old texture. |
| Quality Steps | Time spent on the fix | 30. |
| Prompt Guidance | How strictly it follows the words | 5. |
| Seed | Random number | -1. |
- Nothing changed → your mask has no white in it.
- Error: "This image has no see-through area" → Mask Source is set to a Transparency option but your image is opaque. Set it to Connected mask, or remove the background first.
- The whole image changed → white is covering everything. Fix the mask.
- The edges show a visible seam → go back to Node 3 and raise Feather and Expand.
This one is not from this extension. Preview is a base node that ships with Modly. You find it under Base in the node list, next to Modly's other built-in nodes — never under "Juggernaut XL". It is not downloaded, listed or managed as part of this extension.
Modly's canvas is built for 3D models, so flat pictures do not show up on it. The Preview node displays your image right inside the workflow.
- Add the Preview node from Base in the node list.
- Connect any image output to it (Juggernaut XL, Restyle, or Inpaint).
- Run the workflow. The picture appears in the node.
- It is a base node, so it is not listed under "Juggernaut XL" and it does not show up on the extension's page. It works with any image output.
- If the base Preview node is missing from your Modly, the extension's
setup can add it: click Repair on this extension, or run
add_preview.batwhile Modly is closed. - To take it back out, run
remove_preview.batwhile Modly is closed.
[Your image] ──► Select by Text ("sky") ──► mask ─┐
[Your image] ───────────────────────────────────► ├─► Inpaint ──► Preview
┘ prompt: "red sunset sky"
- Select by Text with
the person. - Set Invert to Yes (select everything else).
- Connect image + mask to Inpaint.
- Prompt for the new background, e.g.
busy city street at night.
[Your image] ──► Restyle ──► Preview
Set Restyle Strength to 0.7, prompt for the whole new look.
Juggernaut XL ──► Preview
No inputs. Just type a prompt and run.
- Keep the default settings at first. They match the model's recommended values.
- Number boxes (Restyle Strength, Prompt Guidance, Threshold,
Inpaint Strength) have a slider beside them: drag the slider to scroll
the value, or click the number and type any value — decimals like
0.5are fine and stay put while you type. - Screenshots below are small thumbnails — click one to open it full size.
- Leave Negative Prompt empty. This model does better without long ones.
- Out of memory? Use a smaller image.
- Want the same picture again? Put the Seed number from the last run back in and change nothing else.
Why did my restyle change everything? Because Restyle redraws the whole image. It cannot keep parts the same, even if your prompt says so. Use Select by Text + Inpaint for one area.
How do I change just one thing, like the sky?
Select by Text → type sky → connect its mask to Inpaint → prompt for what you
want there. Only the white part changes.
How do I make a mask myself? Any black-and-white image works: white on the part to change, black elsewhere. Save it as a PNG and plug it into the Mask input.
Why can't I say "keep everything the same"? Your prompt describes the whole picture. There is no way to attach words to one spot, so that instruction does nothing.
How do I remove a background? That is a different extension — it is not part of this one. Look for Image Editor in Modly under Extensions (Install from GitHub), or grab it from the Community Extensions list in the Modly Discord. Once it is installed, use its Remove Background node — the subject stays exactly the same.
Do I have to download the model three times? No. Juggernaut XL, Restyle and Inpaint share one download.
| Problem | Fix |
|---|---|
| Still says Install after downloading | Reload Modly. |
| Says the weights are missing | Download them from the Models page. |
| Out of memory / crash | Use a smaller image, close other apps, or use a GPU with more VRAM. |
| Windows "Unknown publisher" popup | Click Yes, then fully close Modly so the Preview node installs. |
| The base Preview node or the parameter sliders disappeared after a Modly update | A Modly update overwrites the patched app. Click Repair on this extension, or run add_preview.bat while Modly is closed. |
| Nothing happens when I press run | Check the node has a prompt and, for the others, that the wires are connected. |
- Juggernaut XL v9 by RunDiffusion
- CLIPSeg by CIDAS for text-based masking
- SDXL by Stability AI
- iammojogo
MIT