Beginner’s tutorial series
Easy mode with Brew

Positive/Negative prompts

AI upscale your images

Fine-tune with Render

> Upload/edit images | advanced

Change visual styles

Make realistic people with AI

Commands & troubleshooting

Upload and altering images with AI

You can upload your photos and use /remix to make dramatic changes to your images. We have various kinds of ways to edit images, each with its own strengths and limitations.  Start with this quick video to see it in action:

Remix is useful for DRAMATIC changes to an image. It will not keep the face the same, it will move furniture, it will add things that were not there before. Usage:

/remix and your prompt plus a <concept>

…and that will completely change the aesthetics of an image. Very powerful!

For less destructive results, use the /edit command instead.

Examples of /edit:

Both /remix and /edit require a photo for input. You can attach a photo from your device or render an image and then alter it, like this:

/brew a photo of a read headed woman sitting next to a window

A render returns multiple images, so we need to select the one we want to edit first.

Select the picture by replying to it, as if you were going to chat with the photo. And in the reply box, type something like this:

/edit what if her hair was blue?

Use a verb in your prompt, and asking a question can sometimes lead to better results than barking a command. In both cases, be specific.

Edit does not use <concepts>.  It uses its own model.

You can also use positive and negative prompts to put more emphasis on what is important.

/edit change her hair to the color ((blue))

And those are the edit basics!  Take a break here and try it.


The power of /edit is controlled by a hidden parameter called /strength. Strength is set to 1 by default, which the maximum.  Reducing strength by intervals of 0.1 can give you milder results when it goes overboard.

Another tip:  Be as literal as possible. In this failed edit below, it’s clearer to say “behind him in the background” than not acknowledge him, or edit may incorporate him into the request.

Although it looks so cool, bottom left!



Both /edit and /remix use a (hidden) parameter called /strength, which control the intensity of the prompt that follows.

Consider this command to the image above:

/remix /strength:0.7 Magical Fairy

Here, we are saying that we’re putting 70% of strength into “Magical” fairy but that’s not the whole story.  If the image came from an AI render, it is also passing it’s guidance quietly.  The ebb and flow of guidance and strength is better explained with a photo, so check that out in the Guidance tutorial.


Let’s say I want to turn four of my very best friends into skeletons.

Next, select the picture by clicking on it.

On PC, that’s right click. On Mobile, long touch the picture until you see a “reply” box appear, as if you’re going to talk to the picture.  And that’s how we tell the AI what to do. The clarity order of my words has the most weight on my results, so I need to make sure I am telling the AI exactly what I want.

In the chat box that appears, type /remix and what you want to see happen next. Like this:

Now type what you want to see happen after the /remix command

and in a few seconds…

Oh, no, that’s a little too creepy!  I want cartoon skeletons, not fresh corpses! I wasn’t specific enough. I need to use clearer language.

And that’s where adding more descriptive words and concepts come in. I should have said “four cartoon skeletons, clip art, etc” so it gave me actual skeleton people instead. But this is what happens when you use Remix — big general changes.  We’re going to need an even more powerful command to make big sweeping general visual changes: switching models, or concepts.

Introduction to concepts

A <concept> is an overall aesthetic, usually one style but some concepts can do many styles depending on the trigger words used.  You can browse the many different options here. Now, let’s try it again with a concept like vector art, which produces flat shapes in the visual aesthetic that I’m looking for. I can get exactly what I want without writing a long crazy prompt about 2D cutouts and trying to minimize the shadows in the image.

/remix four skeletons <vector>

Much better. We support over 100 different concepts at the time of writing.

Using Negative Prompts

We touched on this already, but let’s review in the context of /remix.

(Positive) and [Negative] brackets tell the AI what to emphasize. The smooth brackets are positive and the [stabby-looking sideways vampire teeth] are negative, that’s how I remember them. Use them to bite and stab a word down from not appearing in your images like this: [six fingers] and use round soft brackets to say what you do want, like (epic boobs).

/remix David Bowie as the joker (facepaint))) [clown nose]

Here we are saying (facepaint) is important, but [clown nose] is not what we want. Use my silly example to remember which one is evil, negative. Dracula./re

“Joker” prompt by Pirate fellow Guillo W

You can use multiple brackets for emphasis, each one inching up by 20%. (((((epic boobs))))) gives us 100% seriousness on epic boobage.

Example 3: A very stubborn image

Now that we know some advanced manual commands, know what concepts are, and what positive prompts are, let’s look at another problematic result. One of our users wanted to turn this image into the gates of hell.  So let’s see how they did…

They try: /remix gates to hell

And they were disappointed that the results were underwhelming. What went wrong here?

First of all, the AI will try to warp the limited contents and colors into a hellish interpretation in a very gentle way without totally changing the image. And that’s not what we want. We want some colorful fire!

But first, let’s recap what we learned so far:

  • The order of your words is important. Put the most important words first in your prompt.

  • Be more specific. Instead of writing “a man with a dog face”, which can mean a man who is accompanied by a dog who happens to have a face, write a dog-faced man. A fiery place is easier to understand than the many interpretations of hell, so lead with fire and add hell later.

  • Use (positive) brackets to emphasize what you want.
  • Use a <concept> that’s most appropriate for your target aesthetic
  • Use strength and guidance as a last resort if the above wasn’t clear enough

In the case of a black and white photo that needs colorful fire, we also forced up the guidance (creativity) and strength (literal interpretation) to get this wild image:

Remix is your most powerful tool after Render. Have fun experimenting!

Tips & Troubleshooting:

Remix works best when you are reminding the AI what the contents are to begin with. Had I just said “/remix skeletons” without mentioning that I’m expecting four skeletons back, it may draw a graveyard or one big skeleton with dark bones. Be as specific and literal as possible, and put the most important words towards the front of the prompt for best results.

The bot is ignoring my remix commands

If you’re in a group chat, @PirateDiffusion_bot needs to be an administrator of that group or it cannot read your messages. Make this change by editing the group (top right menu) and set it as an administrator. This will reset the room and will also change the room ID, so please backup your stuff before doing this.

Check your members list. It needs to say “has access” or remix/edits won’t work as it cannot read the image to begin with.

Also, make sure compression is turned on when you upload the image, or the bot will think its an attachment. Drag your image towards the bottom on PC, or make sure this box is turned on:

Beginner’s tutorial series
Easy mode with Brew

Positive/Negative prompts

AI upscale your images

Fine-tune with Render

> Upload/edit images | advanced

Change visual styles

Make realistic people with AI

Commands & troubleshooting

pirate diffusion logo