In a bid to revamp its reputation in the creative community, Adobe is rolling out Firefly Image 3, the latest iteration of its generative AI model, during the Max London conference this week.
Long criticized for its underwhelming performance compared to competitors like Midjourney and OpenAI’s DALL-E 3, the Firefly image generation model is getting a significant upgrade with Image 3. Promising more realistic imagery and enhanced capabilities, Image 3 is poised to address the shortcomings of its predecessors.
The new model, now integrated into Photoshop (beta) and Adobe’s Firefly web app, boasts improved scene understanding, lighting, and text generation abilities. It aims to deliver crisper typography, more detailed iconography, and lifelike depictions of various subjects, including dense crowds and people with diverse moods and expressions.
Firefly is now natively integrated in the following Creative Cloud apps:
- Photoshop: Generative Fill and Generative Expand (both recognized by TIME as Best Inventions of 2023)
- Illustrator: Generative Recolor and Text to Vector Graphic (beta)
- Adobe Express: Text to Image, Text Effects, Generative Fill, and Text to Template (beta)
- Adobe Stock: Text to Image and Expand Image
- Substance 3D: Text to Texture in Substance 3D Sampler and Generative Background in Substance 3D Stager
- and available as a standalone web app at firefly.adobe.com.
Users around the world are leveraging Firefly as a partner to dream up new ways to be creative, differentiate their work, and build their businesses.
Initial tests suggest that Image 3 indeed represents a notable improvement over Image 2. Samples showcase enhanced lighting and intricate scene details, indicating a step forward in Adobe’s quest for superior generative AI.
New AI-powered features
Behind the scenes, Adobe is leveraging a larger and more diverse training dataset, sourced from uploads to Adobe Stock and licensed content. However, the inclusion of AI-generated images in the training data has stirred controversy, raising concerns about potential copyright infringements.
Despite assurances from Adobe regarding content moderation, challenges persist in ensuring a flawless user experience. Nonetheless, the introduction of Image 3 brings a host of new features to Photoshop, including a style engine for color and pose variation, and precision editing tools like Generate Background and Enhance Detail.
Meanwhile, enhancements to the Firefly web app, such as Structure Reference and Style Reference, offer additional avenues for creative exploration. With these upgrades, Adobe is aiming to empower users with greater control over their creative endeavors, all while maintaining competitive pricing and robust copyright protection measures.
As Firefly Image 3 makes its debut, the creative community eagerly awaits the opportunity to put its capabilities to the test, heralding a new chapter in Adobe’s ongoing commitment to innovation in generative AI technology.
I asked Adobe if, with all the upgrades, Firefly image generation pricing would change. Currently, the cheapest Firefly premium plan is $4.99 per month — undercutting competition like Midjourney ($10 per month) and OpenAI (which gates DALL-E 3 behind a $20-per-month ChatGPT Plus subscription).
Adobe said that its current tiers will remain in place for now, along with its generative credit system. It also said that its indemnity policy, which states Adobe will pay copyright claims related to works generated in Firefly, won’t be changing either, nor will its approach to watermarking AI-generated content.