Adobe Firefly: tutorial

With the arrival of Dall-E, , Midjourney and Stable Diffusion Adobe – the giant of digital content creation – saw the writing on the wall: its position could be seriously threatened.

Because if a few lines of text could create any kind of image (photorealistic, illustration, digital product interfaces…), why would we use its suite of tools?

So it got to work on its own AI-powered content generation tools and initially launched two (with several more on the way):

• Generative fill in Photoshop.
• Adobe Firefly: the text-to-image generator (and a few other things).

In this article I'm going to talk about the latter.

About how it has become an extremely useful tool for my day-to-day work.

I'm going to explain how I approach some fairly common use cases.

So, if you also carry out similar tasks, you can add it to your arsenal and weigh it against the alternatives.

So, let's start at the beginning.

What is Adobe Firefly?

Firefly is the name Adobe has given to its generative artificial intelligence, which lets us start from text (a prompt) to create images from scratch, retouch or modify existing ones, create videos and edit them easily, or create vectors from sketches.

As you can see, it's quite comprehensive and goes beyond what alternatives such as Stable Diffusion or Midjourney currently offer, to name the two most popular.

Another advantage of it being from Adobe is that it's integrated into its editing tools, such as Photoshop or Illustrator..

But for people who don't use those tools, it also has its own web platform, so anyone who wants to use it can do so independently.

And that's the platform I'm going to explain.

Uses of Adobe Firefly

First of all, the first thing you need to do is open the tool and create a free account if you don't already have one.

Once inside, you can see that, as of today, the tool has six types of use. Four already work and two will be available soon:

Let's go through each one and see what it can offer us.

Text to image

The first is perhaps the best known and most common.

It's the direct competitor to Midjourney or Dall-E: an image generator driven by prompts.

In other words, we write some text -in English- and the tool turns it into several images, so we can choose the one we like or keep changing the text until we get what we want.

How to use it

When we enter this feature, an infinitely scrolling carousel of images generated with the tool opens.

If we hover over one of them, we're shown the prompt that generated it:

Although we can click any of them to open the generator, it's worth spending a few seconds looking for an image that has a style or tone similar to what we're looking for – even if the image itself depicts a different concept –, because it will save us time on editing.

Let me clarify with an example:

Let's say I need a diver (the underwater kind) as the central motif for an illustration of the seabed.

Step 1: selecting the initial image

I look around and see this:

For the illustration I have in mind, this works for me:

• The frontal angle.
• The fantasy realism style of the image.
• The lighting and color tones.

So I click it and the generator opens, where I can see and edit the prompt used, along with the image styles:

Step 2: adjusting the prompt to get a base image

The first thing I can do is modify the prompt so it matches my idea.

If I'm not quite sure how to explain in English what I want, I can use an online translator such as DeepL for that:

Next, I paste it into Firefly and generate the results:

Pshhhhh…

Yes, it's what we asked for, but frankly they're far from what I had in mind. So it's time to refine the image generation a little.

There are mainly two things I can do:

#1. Using a reference image

To do this, from the page with the lousy results, I go back using the browser arrow until I reach my original space koala.

At this point, I choose the image I like best and use it as the model from the three-dot icon that appears in the upper-right corner when I hover over the image:

Despite the facial distortion, the results are starting to get (a little) better for my purpose:

A few interesting things here:

• The most obvious is that Firefly has used the shape of the koala's ears to design the air tubes.
• The images are more realistic than fantastical because I told it I wanted "photo" content. We'll come back to this point now.
• I'm choosing a neutral background so I can later show you a complete use of the tool, but in reality, if I only needed the diver illustration, I'd create the background right here.
• We can choose how much weight the original reference image has using the slider that appears when we click that image. In general, 50% will give the best results:

In any case, since I'm still not happy with the results, I'm going to explain the second improvement method.

#2. Modifying the prompt

It's the other option for improving the initial results. And it's much simpler.

If I change the prompt above to: "Cyberpunk diver in bright blue submarine suit, face covered, on a neutral background", the results are much better for my purpose:

Now we're talking.

The two images on the right, especially the top one, could work for me and I could use them as a reference..

Which is, in fact, what I'm going to do. It will be my starting point..

Step 3: selecting the format

Essential for two reasons:

1. It has to fit my illustration.
2. When exporting, Firefly adds a watermark to the image that will mess up the composition (if we haven't planned for it in advance).

Later I'll explain how to fight the second issue, but for now let's stick with the first.

In the right-hand panel we have all the content options.

The first is choosing the format (ratio):

We'll choose whichever one best suits the composition we have in mind.

You can choose or change it later, but you should know that the generated image will be modified (in our example, the diver). In other words, the format doesn't just change: when the image is regenerated, it won't be identical and will change to a greater or lesser extent, so I recommend choosing it now.

I'm going to keep the square format, which is very versatile.

Step 4: choosing the image type

Just below the format is where we start giving the image the tone we want.

By default, based on the prompt, Firefly interprets the connotations we want the image to have: that "cyberpunk" in my example guides the result, for instance.

To begin with, we'll decide whether what we need is:

#1. A photo

As you can see, it doesn't change the original style ("None") very much.

#2. A "graphic"

Which is the illustration or "digital" design version of the previous image.

#3. An artistic style.

Compared with the previous one, this illustration has a much more pictorial or artistic character. More "analog", if you prefer.

As you can see, we're already narrowing down the type of image we want much more. But hold on, there's quite a bit more.

Step 5: defining the style

In my case, I'm going to use photography because, as I said at the beginning, I'm looking for a fantasy-realism style, so I want the initial, base image to be realistic.

Now it's time to apply the fantastical tone, so let's do a little magic.

First, click the "All" button to see all the possibilities, and there are plenty:

You have different "movements", "themes", "techniques", "effects", "materials" and "concepts" to play with.

And you can combine them with one another.

So, do it.

Try things.

Experiment.

Have fun.

That said, as a tip, I'd suggest applying the different selections one at a time so you can see the impact each one has on the image.

That's how you keep some control, because if you apply several at once, you won't know which one is giving the image the effect you like:

One thing you need to know, though, is that the styles you apply to the image will be VERY heavily affected if you're using a reference image.

This is the result of using the same prompt and the same effects without using a reference image:

Step 6: color and tone, the finishing touch

We almost have our protagonist, but we want to polish the image even more.

For that we have the last section of the panel, where we can play with colors, light and composition (where the image is taken from, the type of lens, etc.):

Tip:

You should know that these details, especially composition, will have more impact on our image:

• If we've added a real background, unlike in this example where I'm looking for a neutral background.
• And, above all, if we aren't using a reference image, which removes many of these nuances.

Step 7: downloading the image

Once I have the image I want, all I have to do is download it from here:

But to my surprise, when I open it I see this:

Ohhhh.

All this time working to get an image I can't use.

Well, that's just great.

Don't worry, there are several solutions to this, and I'll explain them in a moment.

But first, let's look at other use cases, because we'll run into the same problem there too.

Generative fill

If we go back to the Adobe Firefly home page and choose the Generative fill option, we'll be able to upload one of our own images (which wasn't possible with image generation), or work with one of the images they suggest:

Use case 1: removing elements from an image

I'm going to upload my diver and select the remove tool:

Now, with the brush in the lower toolbox set to "Add", I'm going to remove the elements I don't want from the image:

I may have to repeat this option a couple of times:

Until we finally get something decent and usable:

Don't worry about the red blotch in the background; we'll make it disappear in a moment.

Obviously, anyone who knows Photoshop has more possibilities, but for those who don't, this feature will be very useful.

Because although in this case I want to remove the watermark, in any photo from our latest vacation I can remove unwanted "extras".

Or in some selfie group shot, I can be very mean and pretend that person Idontlikethatmuch was never there.

I'll leave that with you.

Use case 2: changing an image background

Right, once I've removed the watermark as I just explained (more options later), I click "background" to remove the background:

There we go, now we have the image with no background:

And I add whatever background I want by writing the prompt. For example, an underwater background:

I put it in the jungle:

Or in New York:

Not bad, huh?

Creating illustrations with backgrounds directly

Now that you know these possibilities, I should point out that it would actually have been muuuuch easier to create the background with the image generator, for example as in these two examples:

Whose prompt is:

"Diver in a bright blue submarine suit, with a black diving suit covering his face completely, against a dark and terrifying background of the deep sea.".”

In bold you can see how I tell it that I want the image background.

Faster and easier, so depending on what I need I'll choose the long, manual method or the fast, automatic one.

Creating a complex illustration or composition

Being able to remove backgrounds or add others to our images comes in handy in different situations.

One of them is when you need to create a complex composition with different elements and backgrounds.

In our example, imagine I want to add two more divers to the right or left of the main character and a giant squid behind them. Plus the underwater background, of course.

So, what I'd need to do is:

1. Create each element separately with Firefly as we saw in the previous step.
2. Remove the background from each one as we've seen.
3. Create the composition in Photoshop and save it.

To get something like this:

An "illustration" that any professional illustrator will clearly see has enormous room for improvement.

But I've seen worse video game covers.

And, most importantly: it was made in less than ten minutes.

Seriously.

Starting from scratch.

At least to me, the result / time ratio is amazing..

And Firefly lets us carry out two of the three steps -creating all the elements for a professional illustration / composition and isolating them from the background- VERY quickly.

For the first time in history, our creativity really is the limit..

Not our design or retouching skills.

Not the time available.

Our imagination. Plain and simple.

Color variations in vectors and text effects

Firefly has two other features that I haven't used yet.

One is creating color variations in vector illustrations..

Its use, again, is very simple: upload an SVG and write the prompt or use the options panel:

The other available feature is creating text effects.

I suppose I'll use it at some point for a banner or another format, but for now all I can tell you is that to use it, you just write the word or text and select the effect.:

That wraps up Firefly's use cases, but a promise is a promise…

How to remove the Firefly watermark

Before starting, I think it's worth repeating that images generated with Adobe Firefly on its website are NOT licensed for commercial use.

Can we do it? Yes, but at our own risk.

Right, besides that, you need to know that whenever we download an image from the platform, this watermark will be added in the lower-left corner.

Whether we create an image from scratch or use generative fill to remove or change backgrounds.

But there are ways around it if we want to use the images commercially.

Let's start with the easiest.

#1. By cropping the image

Obvious, right?

I download the full image and then crop it in Photoshop or similar so the watermark is left out:

If you remember, in step 3 of the tutorial we defined the format. And back then I told you there were two factors to consider when choosing it:

• The format we wanted for the image.
• The watermark.

Because, for example, if we want a square image for an Instagram post, we can choose a horizontal format ("landscape" or "widescreen") and then crop it like in the image above.

The watermark will be outside the final composition.

Enlarging the crop

If, after cropping, the image size is too small for what we need, we can always use Waifu to scale it up.

It's a tool I use constantly for this kind of thing, because it doubles (or quadruples) an image with far less quality loss than the current competition.

And on top of that, it's free and very easy to use:

1. Select the image.
2. Choose the settings.
3. Download.

#2. By editing the image

Perhaps the most obvious method and the first one that comes to mind.

But of course, there are a couple of different approaches with this technique.

Using Firefly itself

As we did in the generative fill tutorial with the diver's arm.

The easiest method, although it won't always work.

Using Photoshop (or similar)

If you know how to edit, you may start here directly:

Copy, paste, clone stamp, content-aware fill, etc.

In addition to our usual arsenal, subscription versions of Photoshop add Firefly's generative fill, so you get both methods we're talking about in a single tool.

Ideal.

#3. Take a full or partial screenshot

Let's move on to more imaginative but very simple methods.

Once you've created the image in Firefly, instead of clicking download and having the watermark added, you can do something else: take a full or partial screenshot of the browser window.

Once you have this screenshot, you can refine the crop in Photoshop or similar, with two advantages:

• The image you get this way already has a higher resolution than the downloaded image.
• What's more, you'll be able to use the full image, without a watermark in the corner.

This is always useful, but even more so if you have a 4k monitor, because the screenshot resolution will be huge and will even work for print designs (which require higher resolution than digital formats).

#4. Element inspector

A method similar to the previous one, but a little less manual and more techie..

It consists of, once the image we want has been generated, instead of taking a screenshot as before, finding the original image file in the code so we can save it to our PC.

First, open the inspector by right-clicking the shaded area next to the image, because if I do it on the image itself it won't open (Adobe wants to make things a tiny bit harder for us, understandably):

Next, find the image's source file. You'll need to know a little HTML for this, but it isn't difficult:

I right-click the image path in the HTML and select "open in new tab":

This opens the image in a new browser tab, at a more than decent resolution and without a watermark. Then I can save it to the computer normally:

This is the last technique I know for downloading Adobe Firefly images without a watermark.

As you can see, they're different, and I use one or another depending on the situation. Get used to all of them and the watermark will never be a problem for you again.

Now that we've given the tool a thorough look, we can start drawing a few conclusions. But first…

Advantages of using Adobe Firefly

I see fundamentally three advantages to this tool compared with its competitors.

To begin with, one of the main ones is that it's free..

And there is no limit on generation or downloads. At least for now.

It's not that Midjourney is excessively expensive -quite the opposite- but if you only want very occasional images, well, you save a bit.

The second advantage I see is that it's powerful and simple at the same time..

In other words, getting good results is relatively easy and the side panel helps.

Other tools such as Stable Diffusion are very powerful – especially in their latest versions – but have a steeper learning curve. With Firefly, by contrast, you can get very good results from the start with simple prompts and the side panel.

The final clear advantage I see is that, thanks to this side panel, it's possible to generate a distinctive style for images. If I always use the same styles and effects -a combination that generates images with a certain personality- I can maintain consistency throughout an entire project (images for a website, illustrations or communication pieces for a client).

Disadvantages of Adobe Firefly

But of course, it wasn't all going to be positive here.

This AI has its weak points too. And they're not minor, exactly:

To begin with, the web platform doesn't allow commercial use of generated images.

You can do it, of course, but you risk a fine. So it's up to you.

In this respect, the second complication comes from the watermark added to every creation.

I've explained the ways I know to get around this problem, but it's still a nuisance.

The next downside I see is that, in many cases, it still doesn't handle generating faces and hands very well, producing otherwise correct images with those areas distorted. So if that's the case, you have to modify the prompt and pray.

In this respect, although it's powerful, I think it is still one small step behind the latest Midjourney model, which, as well as improving these deformities in fingers and eyes, generates images with slightly higher quality.

In fact, if we compare it with that tool, we can see that Firefly lacks certain features, because it can't:

• Generate web interfaces.
• Generate logos that look decent.
• Copy the style of well-known artists .
• Use an external image as a reference.

These are all very useful features that Midjourney has and Firefly, for now, doesn't.

Finally, a couple of minor details:

I've found that certain complex concepts such as "monocle" can't be recognized by Firefly within a composition. And it isn't because the prompt is complex: if you replace that word with "glasses", it recognizes and uses it without any problem.

I've run into several cases like this, and finding alternatives is awkward or, in some cases, impossible.

That said, it can use these concepts when written on their own, so I assume it's only a matter of time before it recognizes them.

Finally, it can also be said that, for now, you can only use prompts in English.

This isn't something that only happens with Firefly, but it does happen, and if we want to use complex prompts and aren't very comfortable with Shakespeare's language, we'll have to turn to external translators.

Final conclusions

After explaining the basic uses of the tool, with its advantages and disadvantages, my conclusion is that if the disadvantages aren't a problem -and they won't be for many people-, this is an excellent tool.

One of those that changes the rules of the game.

But of course, we'll have to wait for Adobe to release its commercial version for individuals -it already has an option for businesses- to see the price and decide whether it's really worth it or whether the paid competition (Midjourney) or the free option (Stable Diffusion) offer better value.

For now, I recommend that you start playing around with it, because you may genuinely be in for a very pleasant surprise.