Adobe Firefly: A Quick Overview Of The Latest Generative AI Model

A few years ago, artificial intelligence seemed like a thing of the future. But today you hear people use terms like generative AI, conversational AI, and Generative Pre-trained Transformer as conversation starters. A mind-blowing leap for AI, no doubt. 

The past few weeks, in particular, have been monumental in the advancement of this tech. 

  • The launch of OpenAI’s GPT-4.
  • The release of Midjourney V5
  • Addition of DALL-E based image generator to Microsoft Bing
  • The introduction of generative AI with Cloud Services through NVIDIA AI Foundations 
  • And finally, the recent addition has been Adobe’s Firefly. 

Adobe Firefly is a generative AI model similar to the well-known DALL-E, Midjourney, and other AI-based image generators. But how is Adobe Firefly different or similar to its counterparts? What does Adobe do differently with its generative artificial intelligence tool? We have all the answers right here. 

Adobe Firefly – an overview 

First things first. Unlike other AI image generators, Adobe Firefly is not a single tool but rather a family of generative AI models. These tools are all set to revolutionize the way you design, edit, and manage your designs in various workflows in Adobe Express, Creative Cloud, Experience Cloud, and Document Cloud. 

Adobe Firefly API is on the cards. But at the moment, the beta version of Adobe Firefly is rolled out as a web-based application. And it can be accessed directly through the Firefly website. And people are enthusiastically sharing the crisp and accurate results from it. 

With the initial launch right now, there are two main functionalities available in the beta and those are:

  • Generating images from text prompts (suitable for commercial use)
  • Text effects generation 

More cool features are in store and are likely to be released in the coming months. 

Even in terms of the availability and integration of this generative AI model, Adobe is planning to expand its scope. In addition to offering an API in the future, there are plans to make the model available in native Adobe apps as well. Eventually, creators using Illustrator, Photoshop, Experience Manager, and Adobe Express will be able to seamlessly integrate AI into their design workflow in order to expand their creativity. 

Now that can be a big leap considering how some creators now generate design elements on AI models and then import them into their design tool for further processing. With Adobe Firefly, you get to skip the whole process of export and import. 

A quick look at what Adobe Firefly can do at the moment

Generate images based on text prompts 

Text-to-image functionality needs no introduction to the generation that’s been exposed to DALL.-E, Midjourney, and Stable Diffusion. All it takes is to key in a text prompt that accurately describes the concept you want visually represented and the AI model churns out images in seconds. The image generation functionality on Adobe Firefly is pretty much the same. 

Similar to most other AI image generators Adobe Firefly creates four different versions of an image generated based on the prompt. One thing that sets Firefly apart is its interface which lets you change aspect ratio, effects, colors, lighting, and a whole lot of other things directly without navigating out of the window. As a result, you can apply custom effects to see if you like the image or go ahead with generating new versions. 

Once you like what you see, you can also generate variations of that particular image or download it. 

Source: Image generated on Adobe Firefly. Prompt: A side profile of a woman with rainbow swirls of smoke overlayed in the foreground 
Text effects 

At first glance, this looks like Word Art of the AI era but there’s more to it than meets the eye. All it takes is a combination of the text you wish to stylize and the kind of effects you wish to apply to it. 

The below Tweet shows the design generated for the prompt “diamond crusted”. 

Let your imagination run wild and Adobe Firefly quickly applies the most unusual and unexpected effects to your text in seconds. 

Within the text effects, you can again choose from four different variations of the effect, change the font style, text effect fit, background color, and other parameters before you download the text. While the applications are plenty this effect is pretty fun to use. 

The best part is that Adobe Firefly can work its magic not just on text but also on emojis. People have been sharing creative versions of unicorns and stars from the emoji library. 

Adobe Firefly features in the pipeline 

While the image generation functionality of Adobe Firefly is pretty impressive, you must have noticed that the model currently lacks some of the features that its counterparts are known for, like inpainting and outpainting on DALL.E-2 for example. But these are on the way. 

  • Inpainting – this lets you select specific elements in your design and remove them, replace them, or edit them. AI ensures that the surrounding elements are not distorted in the process. 
  • Extend image – this is similar to the Outpainting feature introduced in DALL.E-2. This feature will let you extend images intuitively retaining the theme and the scene. It comes in handy when you have to expand an existing design or alter its aspect ratio in order to optimize it for a different platform. 
  • Sketch to image – if you have a rough sketch and wish to create a realistic design out of it, this feature will make it possible. Whether it is something as simple as a logo or a detailed portrait, Adobe Firefly will be able to convert it into an image.
  • Video editing – from changing scenes to specific aspects of the video, a whole lot will be made possible with the upcoming features on Adobe Firefly. This will make content repurposing an easy job. 

In addition to the above-mentioned features, similar to what we have seen on other generative AI models, there are a few more exciting ones too. 

Upcoming features that give Adobe Firefly an edge over other generative AI tools 

When there’s a new entrant in a market, comparisons are prone to occur. People have already been comparing Adobe Firefly to Midjourney and DALL-E talking about how similar or how different these models are. 

While Adobe Firefly has only been released with two features at the moment, the upcoming functionalities of this tool are creating quite a buzz. These are the features that are currently not available on other popular AI image generators right now. What are these much-anticipated features of Adobe Firefly? Let’s find out: 

1. Recolor vector 

With the two functionalities out in Adobe Firefly next in line is the vector recoloring option. If you’ve tried recoloring vectors on tools like Adobe Illustrator you know that there are a few steps involved in applying the desired color in place of a particular color. Adobe Firefly will let you do this in seconds through a simple text prompt. 

2. Text to vector 

This is a feature that a lot of designers might appreciate. Creating vector images with AI image generators has not been practical. Midjourney for example, can generate vector-like artwork in a rasterized version. This again needs to be vectorized on tools like Illustrator. 

However, Adobe Firefly will let users directly create vectors from text. Which means that you get to create better-quality design elements without significantly increasing the file sizes. 

3. Text to brush 

Creating custom brushes is no easy feat and most creators know this. But then having a custom brush comes in handy when you need to make a quick few touch-ups. Or even to extend portions of your design or create patterns. For all this, Adobe is all set to bring a feature that lets you create brushes based on text prompts. These can then be used in your design project along with the other brushes in your Illustrator or Photoshop libraries. 

4. Text to template 

This will probably be one of the most useful features for brands. By giving Adobe Firefly a bunch of elements you can create templates for your brand. Again, like the other designs you can edit these templates and get them ready for your campaign or save them for future projects. 

Adobe Firefly looks different from other generative AI tools not just in terms of these features but also in terms of its approach. And one such is the use of an inclusive dataset to train the model so as to avoid bias in the results. Another aspect that people are talking about is Adobe’s decision to value creators and acknowledge intellectual property conditions in its training dataset. Let’s talk about this in detail now. 

Adobe Firefly – embarking on a responsible approach for training the AI model 

Ever since the emergence of generative AI tools, one of the biggest topics of discussion has been “art theft”. 

The line between “inspired” and “infringed” seems to be fading and there’s not enough being done about it. Because when it comes to generative AI we can say that most tools are in their developmental stages, their learning stages. 

For example, Erin Hanson is an artist known for her work in Contemporary Impressionism. The below image shows her artwork titled “Crystalline Maples”. And the image below that shows an AI-generated version mimicking her style. 

There have been many such images popping up on the internet showing AI-generated images that are scarily similar to the original artworks of various artists. The problem stems from the fact that these original artworks were used in the training dataset. And most artists are not in favor of this approach. That’s where Adobe is trying to take a different path with Firefly. 

According to Adobe, the training dataset for Firefly consists only of images that are available in the public domain and openly licensed ones. The first model was reportedly trained on the existing Adobe Stock database which again has clear licensing to avoid IP issues. It is a step toward ensuring that artwork generated on Adobe Firefly does not infringe on individual and brand IPs. 

Additionally, Adobe Firefly will include a “Do Not Train” option that lets artists opt out so that their artworks are not used in the training dataset. 

Where does Adobe Firefly fit into your creative workflow?

Marketers hearing about generative tools day in and day out feeling nervous about lagging behind but not knowing which way to go, we see you! There’s a lot happening. The real problem is not about finding the right AI tool but about knowing where to use them. 

So, let’s address that question for Adobe Firefly. Where does that fit into your marketing process? 

  • If you are creating your designs yourself, you know that executing the design you have in mind takes time and effort. After spending this time and effort if your design does not turn out to be what you had in mind, it can be quite disappointing. In other words, if the idea that looked so good in your mind does not look half as good in design, it can be a setback. In such cases, generative AI tools like Adobe Firefly help save the time and effort you would otherwise spend. These tools help you quickly visualize your idea. 
  • Or if you are working with a design team, then you can use these generative AI tools to generate visual representations of your ideas in seconds. No more hunting for the right reference images for your design brief. We know AI tools aren’t perfect but your designer or design team can help perfect your designs and make up for the gaps in these AI tools. 
  • Another way to use Adobe Firefly and other AI design tools is to generate design elements for your brand graphics. Scouring through stock libraries to find that one little icon that completes your design is not always easy. 

The future of content creation lies in AI integration 

Brands are using AI in:

  • Marketing to support automation
  • Customer support to help improve customer experience
  • Social media to perform better analytics

Even if you aren’t adopting these new technologies, your competitors are. 

The future of design, and content creation, on the whole, looks favorable to the incorporation of  AI into the workflow. For improved productivity. So, the key is to identify the right AI tools to invest in rather than spending on all of them and feeling overwhelmed. Based on what we see right now, Adobe Firefly is one such tool that’s worth investing your time and attention in. How would you use Adobe Firefly to enhance your design workflow? 

Or if you are looking to work with a design team that’s all set to embrace the new trends in the world of design, one that can turn your AI-generated design elements into brand-relevant graphics for your campaigns, choose Kimp. Sign up for a free 7-day trial to see how Kimp works.