Adobe Firefly: A Complete Guide to Adobe’s AI Image Generator

I. Introduction

Introducing Adobe Firefly, the new AI image generator from Adobe

Adobe Firefly is an artificial intelligence image generator designed by Adobe. It is part of Adobe's range of generative AI models, all dedicated to creative expression. In fact, Firefly uses artificial intelligence to produce impressive images and text effects, all based on the “prompts” or ideas we give it.

It's a tool designed to enable designers to bring their ideas to life more efficiently and without any constraints.

It competes with other tools on the market such as Midjourney, Dall-e from OpenAI and Stable diffusion.

You can access Adobe Firefly at this URL: https: //www.adobe.com/fr/sensei/generative-ai/firefly.html

Adobe Firefly

You can register for free:

adobe firefly login page

Next, I selected “Text to Image” to access Adobe's AI image generator.

integration into Adobe's full suite of applications, with an emphasis on the unique aspect of its commercial use

Firefly is designed to be integrated directly into Adobe's suite of applications, including Creative Cloud, Document Cloud, Experience Cloud and Adobe Express. This integration enables users to leverage the power of generative AI in their existing workflows, increasing productivity and creative expression for all creators.

II. The different modes of Adobe Firefly

Adobe Firefly offers several modes to meet different creative needs. These modes include Text to Image, Generative Fill, Text Effects, Generative Recolor, 3D to Image and Extend Image.

Text to Image mode

This mode enables users to generate images from text descriptions. Compared to other AI image generation tools, Adobe Firefly stands out for its ability to generate high-quality, commercially viable images.

Firefly is better at creating realistic scenes than imaginary, fairy-tale or science-fiction scenes.

This is mainly due to the fact that Adobe Firefly's training dataset consists mainly of its Stock Images library.

Testing difficult “text to image” prompts for Generative AI:

As a first step, we're going to try out a number of prompts, including the difficulties known to image-generating AIs, in order to test the tool's generation capabilities.

All you have to do is click on “Text to image”, then write your prompt in the dedicated space.

Please note that the prompt must be in English only:

Prompt test 1“a smiling japanese man with a glass of scotch in his right hand and greeting with his left hand” This prompt shows how well Adobe Firefly handles :

  • Right from left
  • Respect for all prompt requests
  • The number of fingers on a hand (one of the biggest challenges for image-generating AIs)
  • The number of teeth in a mouth (another major challenge even for Dall-e and Midjourney)
  • Whether the images contain inconsistencies that make them directly identifiable as AI-generated images.
Limitations adobe firefly

Result:

  • Right from left: 3 out of 4 images are OK in this respect (15/20)
  • Respect for all prompt requests: the first and third images respect the prompt perfectly, but in the second image the character doesn't say hello with his hand, and in the fourth the glass is in the wrong hand. This is normally a 10/20 score, but it's worth noting that we still have 2 images that correspond exactly to our request, which is more than enough to find what you're looking for. Note also that all the images respect the Japanese origin of the character, as well as the Scotch glass. I give it a 14/20
  • The number of fingers on one hand (one of the biggest challenges of image-generating AIs): At first glance it looks ok, but in fact image 1 has 6 fingers on the left hand and a kind of mixture on the right. Image 2 seems to have 4 fingers. Image 3 seems to have a fusion with a sixth finger. Image 4 has the right number of fingers. With only one image (which doesn't perfectly respect the basic prompt) that has the right number of fingers and can therefore be used, it's an 08/20 for me.
  • The number of teeth in a mouth (another major challenge even for Dall-e and Midjourney). A single character mice his teeth, and the result is OK, but if we look at the details we'll see that the number and size of the teeth isn't really natural, but invisible without zooming in. No note because only one image out of 4.
  • Do the images contain inconsistencies that make them directly identifiable as AI-generated images? In addition to the hands, we have quite significant deformations on images 1 and 3 (a pity, as these are the most faithful to our request). Otherwise, the look and eyes are really OK. 11/20

My opinion on the result of this Prompt: If a photo taken by a “normal” human is a 20/20 in 100% of cases on the realism aspect, here we get a 12/20. On the other hand, this is a “One shot” test, so I'm sure it's possible to rework the prompt and test several examples, sometimes just regenerating the prompt gives better results.

This test was carried out in June 2023, and the progress made by generative AI in recent months, particularly on faces, is quite incredible. It's possible that I'll do this test again in 2 or 3 months and I'll get a 15 or 16 easily.

Prompt 2 test – “A gas station with a sign that says “Nick's Gas Station””This prompt lets us see how well Adobe Firefly handles handwriting management, which is also one of the known difficulties of major players such as Dall-e, Midjourney or even Stable Diffusion.

Prompt exemple Adobe Firefly

Result:

Here, not a single image manages to reproduce the text in the signs, and we can also see that the other texts are not comprehensible either. We do, however, have some very well-integrated letters. So, for the time being, a 08/20 for this subject.

So that's it for the main weaknesses of the model at the moment, we'll now look at some interesting examples of art that can be created with Adobe Firefly.

Examples of using Adobe Firefly

A scene from life: “Peruvian woman, captured in her traditional attire full of vibrant colors and intricate patterns, working diligently on her loom set against the backdrop of the rustic adobe house with alpacas grazing in the sunlit yard, radiating a sense of tranquility, hard work, and harmony with nature, in the style of Street Photography, taken with a Nikon D850, utilizing a 35mm lens with an aperture of f/1.8, ISO 400, to capture the vividness and fine details of the scene.”

We can see that there are many more elements in this prompt, so to get an image close to what we want, it's important to be precise.

Here the prompt is created like this:

[1] “Peruvian Lady“: This is the main subject of the prompt.

[2] ” captured in her traditional attire full of vibrant colors and intricate patterns, working diligently on her loom”: Here, I've given a detailed description of the Peruvian lady, describing her clothes and what she does.

[3] “set against the backdrop of the rustic adobe house with alpacas grazing in the sunlit yard”: I then described the setting of the scene, set in the backdrop of a rustic house with alpacas grazing in the sunlit yard.

[4] “radiating a sense of tranquility, hard work, and harmony with nature”: This part describes the mood of the scene. I chose to portray a sense of tranquility, hard work, and harmony with nature.

[5] “in the style of Street Photography”: I chose the style of street photography for this image, as it matches the description of the scene.

[6] “taken with a Nikon D850, utilizing a 35mm lens with an aperture of f/1.8, ISO 400, to capture the vividness and fine details of the scene”: Finally, I gave details of how the photo should be taken, specifying the camera, lens and appropriate settings.

A fusion between a lion and a horse: “Horse-lion amalgamation, featuring a horse's strong haunches and the fierce, golden-maned head of a lion, racing across a beach with the turbulent sea in the background, in a scene filled with dynamic energy and excitement, in the form of a realistic photograph, shot with a Nikon D850 DSLR in burst mode to capture the creature's movement, with a wide aperture for a shallow depth of field, focusing on the subject against the blurred background.”

A futuristic city : “Futuristic City, an advanced urban landscape at dusk, where colossal buildings glowing with neon lights stretch to the heavens and the streets teem with autonomous vehicles. Amidst the concrete and metal giants, a park filled with luminous flora offers a respite, with verdant trees bearing radiant fruits and glowing flowers. This city of tomorrow stands under a twilight sky painted with hues of purple and orange, in an atmosphere humming with the symphony of technological advancement and a sense of sophisticated serenity. Captured in a photography style, through a fisheye lens to exaggerate the perspective and enhance the grandeur of the cityscape, taken from a high vantage point during the ‘blue hour', with appropriate aperture, ISO, and shutter speed settings to balance the low light conditions and the vibrant city lights.”

House in the middle of the forest: “House in the middle of the forest and a big river across the house, an old timber cabin with a mossy roof and smoke curling from the chimney surrounded by towering ancient trees, settled against the backdrop of a clear, wide river reflecting the dappled sunlight, instilling a sense of solitude and serenity with the calming sounds of flowing water and rustling leaves, captured in the style of Photography, Landscape photography with a Canon EOS 5D Mark IV utilizing a wide-angle lens for a broader perspective and a smaller aperture to keep both the house and the forest in focus.”

A “stock image” of a doctor and patient: “Doctor and Patient, a meticulously composed scene featuring a medical professional in a white coat with a stethoscope around his neck, talking to a patient sitting across from him in a well-lit, modern healthcare clinic, within an atmosphere of trust, empathy, and hope, captured in Photography, using a Canon EOS 5D Mark IV in Portrait mode, with ISO 100, aperture f/1.8, shutter speed 1/200 to capture high-quality, sharp and detailed facial expressions and a bokeh background.”

Other Adobe Firefly modes:

  1. Generative Fill: This mode allows users to change elements directly in an image, for example changing the background, adding elements to a model, changing a color… In short, the possibilities are almost infinite. We'll look at this in detail in a future article.
  2. Text Effects: This mode allows users to create texts with a generative fill according to a prompt. It's quite simple, but difficult to describe. This is what I'm using for the featured image in this article
  3. Generative Recolor: This mode lets you change the colors of vector images, providing a quick way of testing different variations of a vector image.
  4. 3D to Image: This mode transforms 3D elements, which you position according to your desired perspective, into a 2D image, offering unique perspective and creative possibilities. Not yet available at time of writing.
  5. Extend Image: This mode lets you “extend” an image, i.e. create a continuation of an image on the ribs above. This mode is not yet available on the online application.

IV. A few examples of how to use the tool

When it comes to graphics, design professionals will undoubtedly be the first to use Adobe Firefly. The “Generative Fill” feature, for example, can save an incredible amount of time when it comes to testing different variations of a design, especially as Firefly is now integrated directly into Photoshop. You save time, boost productivity and discover design ideas you couldn't consider by hand.

In advertising and marketing, designers and marketers can use tools like Adobe Firefly to concoct tailor-made ads and content for their target audiences. Custom stock images can be created. These images can be adapted relatively easily, and A/B tested with 10 or 20 different visuals according to user data, to maximize engagement and conversion rates. Not bad, eh?

Interms of website development, developers can use AI to generate images and graphics for website design. This helps reduce costs and time spent searching for or creating images.

In content creation for social networks, AI can be used to create attractive and engaging visual content for social networking platforms, including infographics, memes and other shareable images.

In print and publishing, AI can be used to create images for print media, such as magazines, newspapers and books. This could well be a game-changer in the creation of visual content for these industries.

In games and entertainment, game developers and filmmakers can use AI to create images and graphics for their productions. This could result in more realistic graphics and more immersive experiences for viewers.

In fashion and clothing design, AI can be used to create digital mock-ups of clothing designs. This enables designers to visualize their creations before they are produced, saving time and resources.

Finally, for education and training materials, teachers and trainers can use AI to create personalized images to illustrate their points. This could help increase engagement and retention of information among students or trainees.

V. Conclusion

Here's my little analysis of Adobe Firefly: It's a great AI-powered image generator, a real novelty made in Adobe. It lets you create high-quality images and produce breathtaking text effects thanks to AI. What's cool is that it's integrated directly into the entire suite of Adobe applications, so for creative types, it's a great way to use all its high-tech features without leaving their usual work routine.

Firefly has several modes on the menu: Text to Image, Generative Fill, Text Effect, Generative Recolor, 3D to image and Extend Image, to meet all kinds of creative needs. Each mode brings its own specific features, enabling users to create custom images, explore new ideas and even automate certain tasks.

Admittedly, Firefly isn't perfect yet. For example, it still struggles a little with handwriting precision and strict adherence to prompts. But frankly, it's already doing a great job of creating super-realistic, commercially usable images. And let's not forget that AI is evolving at breakneck speed. So, advances in this field will certainly boost Firefly's skills even further.

When you see what Adobe Firefly can do, you realize that its potential is enormous in many fields: graphics, advertising, web development, content creation for social networks, printing and publishing, games, entertainment, fashion, clothing design, education, training

In a nutshell, Adobe Firefly opens up a wealth of creative possibilities thanks to AI. It simplifies the image creation process and opens up new avenues for creatives, marketers, developers and other pros. With future enhancements and integration with existing Adobe applications, Firefly has the potential to become a key player in the future of visual creation.

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *