During the rise of artificial intelligence (AI), Adobe keeps up with the innovations and creates a new feature that is changing the way we edit visuals. The new Generative Fill feature easily generates objects and creates content from simple text commands.
In this article, we’ll take a look at how Generative Fill works and what can you do with it.
Generative Fill review summary
Generative Fill delivers impressive results transforming text prompts into visual objects and textures that quickly enhance pictures. Although it’s currently only in the Photoshop Beta version, it can soon become a new benchmark for image editing.
Table of Contents
What is Generative Fill?
Generative Fill is an AI-powered feature in Photoshop that enhances images with the help of text prompts. It can add, remove, or extend existing content and is currently available in the Beta version of Photoshop.
Powered by Adobe Firefly, Generative Fill analyzes existing content and produces new, realistic visuals based on the context. The feature is designed to assist graphic designers and photographers by automatically generating or removing content.
The company is keeping pace with the growing demand for AI technologies. Soon the editing process in Adobe Photoshop will be less time-consuming and more automated.
Photoshop is not the first software to include AI technologies in image editing. For example, Luminar Neo has a lot of AI features as well. However, it seems that Generative Fill will be far more advanced.
The Generative Fill feature can be enabled by clicking Window and selecting Contextual Task Bar in the Adobe Photoshop Beta version. You can download Beta version of Photoshop from the Creative Cloud desktop app.
Adobe announced the Generative Fill feature in May 2023. It is currently available in Beta version of Photoshop and should be officially available in the second half of 2023, as stated on the official Adobe website.
Give it a try: by getting Adobe Creative Cloud plan and installing Photoshop (Beta) desktop version, you can already try Generative Fill.
Features and capabilities
When I first heard of the newly announced Generative Fill feature by Adobe, I knew I had to test it as soon as possible. Their sample video of the editing process had me thinking: “No way”.
During my experimentation with Generative Fill, I found out that it can perform these tasks:
- Select object
- Remove background
- Remove unwanted objects
- Generate visual content
- Extend images
- Create new objects
The process is very straightforward – select an area and describe what you want to see. And Generative Fill will provide a unique result following your instructions.
Adobe Photoshop’s Generative Fill can soon become a new standard in the graphic design industry, offering an efficient and flexible solution for generating fascinating visuals and textures.
Generative Fill is integrated into Photoshop
Adobe has taken a smart step by integrating Generative Fill into Photoshop. It is powered by their newly announced product Adobe Firefly.
By integrating Generative Fill directly into Photoshop, Adobe has eliminated the need for users to rely on third-party plugins or external applications. This integration ensures a streamlined workflow, allowing users to experiment and harness the potential of Generative Fill without leaving Photoshop.
Add or remove objects
Whether it’s eliminating unwanted elements, such as photobombers or distracting background objects, or adding new elements seamlessly into a composition, Generative Fill proves to be a game-changer.
I gave Generative Fill a test. First, I wanted to remove the little blurry house from the background.
Content-Aware Fill tools existed in Photoshop and Lightroom before, but now it’s integrated inside the Generative Fill taskbar, making it easy to access. We could see its capabilities to be improved too.
Minor corrections are needed here, but let’s say I’m satisfied with the result.
Next, I decided to replace the trees by simply selecting them with Lasso Tool and generating a prompt.
I used a short and simple text prompt – “Replace the trees with more green ones”. Here’s the result:
Looks pretty good. The trees look really natural, however, the mountains on the left side got distorted.
I also wanted to add an object on the left side to create a feeling that the woman was looking at it. I provided this prompt:
“Add a fish in the water, make its colors match with water, and make it low opacity”.
Let’s look at the final object it generated.
Impressive results on extending images
Finally, I wanted to extend the image on the sides to create a panoramic look.
What I like about this feature, is that you just need to create the empty space where you want the image to be extended, and select that empty space. After that, you click on Generative Fill and select “apply” without entering any prompt. This way, the tool understands that you want to generate extended content for the empty space.
Here’s the result:
I truly think this is impressive. This feature also makes it easy to change the aspect ratio of an image without cropping it.
However, I must notice that this feature could take a long time to load if your computer is slow and the image is large in size.
I also checked many other articles to see how this feature is tested by other people. I quickly noticed that other creators got even better results, with Generative Fill enabling them to generate content for even much larger empty zones.
New layer for each task
Generative Fill automatically generates a new layer mask for each task performed. These layers are called Generative Layers and are added to the layers taskbar.
I would call this a very positive feature allowing to selectively reveal or conceal parts of the generated content.
By automatically creating a new layer mask, Generative Fill ensures that you have the freedom to experiment and iterate, enabling you to achieve professional-grade results with ease.
Here’s one more example of an image edited with Generative Fill only.
Generative Fill Prompts
Here is a list of prompt examples for the Generative Fill feature:
|Add an object||Add [subject name or short description of what you want to be added]|
|Add or replace background in your selected area||[background description]|
|Replace object||Replace [unwanted object] with [preferred object]|
|Remove object(s) in selected area||Select area and leave text prompt empty (acts as content-aware fill)|
|Extend image||Select empty area around image (selection area should overlap a little bit with your image) and leave text prompt empty|
Text prompt examples to insert objects
To insert an object, you can use this kind of prompt:
“Add [subject name or short description of what you want to be added]”
However, the word “add” is not necessary – once you’ve selected an area in your image where you want to insert something, just write what object you want to be added and Generative Fill will do all the work.
Often, even 1 word is enough. For example, if you write a text prompt “house”, it will insert a house in your selected area.
Here are some text prompt examples to give you a better idea:
- “Sunflower field”
- “Rustic wooden bridge”
- “Vintage typewriter on a desk with scattered papers”
- “Airplane leaving contrails in the sky”
- “Swinging hammock between two palm trees”
Text prompt examples to change background
- “green grass with yellow flowers”
- “shingle beach”
- “crystal-clear lake with reflections”
- “snowy mountains with night sky”
- “milky way in the night sky”
Text prompt examples to replace objects
- “replace pen with a pencil”
- “replace clouds with a rainbow”
- “replace cup of coffee with a cup of tea”
- “replace sunflower with coneflower”
- “replace bike with a red motorcycle”
Remove objects or expand images
- When you want to remove a distracting object – select the area around it and leave the prompt empty.
- If you’ve selected an empty space around the image and want to expand it, just leave the prompt empty – in that case, it understands that you want it expanded.
How much does Adobe Generative Fill cost?
Currently, Generative Fill is only available for Creative Cloud subscribers. Everyone, who has a Creative Cloud subscription, can access the Generative Fill feature by downloading the Beta version of Photoshop.
In the future, we expect Generative Fill feature to be available with all Adobe plans that include official Photoshop, with the cheapest one currently starting at $9.99 per month.
What Generative Fill is lacking?
While using Generative Fill I noticed a few downsides and inaccuracies in the outputs:
- Sometimes it generated visual content that was not entirely accurate. For example, when I asked Generative Fill to replace the sky, it generated a damaged sky with black spots.
- I experienced slow loading for large images. When I asked to extend an image that was about 5500 x 3600 pixels in size, it took a while till it managed to generate extra content for it. It may require a powerful computer system to maintain smooth performance.
- Sometimes you might be not happy with the result it created, so you will need to re-generate the output or add more information to your prompt.
- At the moment, Adobe’s Generative Fill only supports English prompts. But it should support other languages in the future too.
Tips for using Generative Fill
- Carefully select the area. I noticed that the accuracy of your selection plays a crucial role in achieving the desired results. Luckily, the Generative Fill taskbar includes an option to adjust your selected area by adding a border, feather, or expanding the selection.
- Prompts don’t have to be long, they can be only a few words. There is no need to provide context or ask the AI to take on the role of someone. Just tell it what you need.
- Understand your end goal. If you write “add greens” and you hope to see a large tropical forest added to your picture, you might get disappointed. Try to write exactly WHAT you want and HOW you want it.
- Try again or add more details to your prompt if you didn’t achieve the expected result. While Generative Fill in Photoshop is remarkably advanced, it may not always produce the exact output you were expecting on the first attempt. Experiment with different selection techniques, tweak the prompt or input parameters or add more details to the content you provide to Generative Fill.
Adobe is planning to officially include Generative Fill feature in Photoshop in the second half of 2023.
To access Photoshop Beta, you need to have a Creative Cloud subscription. Once you are a subscriber, open the Creative Cloud app, select Beta apps and install Photoshop (Beta).
Beta version is almost the same as the original Photoshop, additionally it provides new features that should be available in original version in the future.
Text prompts in Generative Fill are a way to provide specific instructions or descriptions to the algorithm in Adobe Photoshop. By describing the desired outcome or providing additional context, text prompts help the Generative Fill feature understand your intentions and generate more accurate and relevant visual results.
Currently, Photoshop’s Generative Fill supports text prompts only in English. However, new languages should be added in the future.