I recently got access to the Adobe Firefly Beta, so I thought I’d review it. Adobe Firefly is a new family of creative, generative AI models that will be integrated into Adobe products. Generative AI is artificial intelligence that can create new content from scratch, such as images, text, or video. When released, Adobe claims that Firefly will offer new ways to ideate, design, and communicate while significantly improving creative workflows. Some of the features that Firefly will offer are:
- Text to image: This feature allows users to generate realistic and original images from natural language descriptions. Users can also edit the generated images using Adobe tools.
- Text effects: This feature allows users to create stunning text effects from natural language commands. Users can customize the text effects’ font, color, style, and background.
- Recolor vectors: This feature will allow users to change the color scheme of vector graphics using natural language instructions. Users can also adjust the vectors’ hue, saturation, and brightness, but that functionality is yet to be released.
Firefly is in beta and will be available for Adobe Creative Cloud subscribers soon. The new service from Adobe aims to enhance the creative process and give creators practical advantages while adhering to ethical standards and transparency in generative AI. Firefly is part of Adobe Sensei, which is Adobe’s AI and machine learning platform.
The Good
If you are looking for an easy-to-use generative AI platform, Firefly fits the bill, especially compared to using something like Midjourney. It has a simple interface and a menu-driven approach with filters familiar to someone who is used to using Adobe Photoshop. In addition, I like the ability to create images with multiple aspect ratios via the dropdown (you can do this in Midjourney, but you need to know which options to add to the prompt); I used the Widescreen Aspect (16:9) to create the blog header for this post.
Also, the rendering was speedy, and I got results from my prompts in under a minute. I like that Adobe is trying to be transparent by disclosing the source of the image and that it’s been generated by AI; you’ll see a prompt that notes this before you download any of your images. In addition, you will get a content summary, including that it was generated with AI as part of your content credentials.
The Bad
The only images you can access in Firefly are from Adobe’s stock art collection. You won’t be able to upload your images to edit. However, that will change after the beta period. I wanted to generate a picture of Woodford, our chief canine officer, riding in his 2022 Ford Bronco. Here’s where the context about the dog breed and the knowledge of the car model broke down. I suspect this is a model limitation that should improve over time.
Summary: Adobe Firefly Beta is Good for a Pre-Release
My Adobe Firefly Beta review? Firefly wasn’t bad and looks promising, but it still has a long way to go. I suspect as they go live and have more tagged images and training data, the ability to generate a broader range of accurate images will improve. I also wonder if the users’ data will help with the training just as ChatGPT did with their free tier. Overall, the Adobe product seems more polished than what you would see from Midjourney, Stable Diffusion, or OpenAI’s DALL-E. The ability to integrate into Adobe products will likely help with productivity. Though for now, Midjourney seems to provide a much better result. There has been a lot of fanfare about AI replacing humans. I suspect skilled graphic designers will be more productive using Firefly. This will allow them to offload time-consuming tasks and spend more time tweaking and polishing. For now, Adobe Firefly is worth watching.
