Adobe just opened up its Firefly Custom Models to the public, letting creators and brands train AI image generators on their own artwork. The Verge AI reports that the tool, now in public beta, allows users to build personalized models that mimic specific artistic styles and character designs.
The pitch is straightforward: feed the system your assets, and it learns your visual language. Then use that trained model to generate new images that stay consistent with your brand.
What Firefly Custom Models actually do
The tool preserves visual details across AI-generated images, including:
- Stroke weight and line quality from illustrations
- Color palettes that match your brand identity
- Lighting styles for photography consistency
- Character features that stay recognizable across generations
This matters most for teams producing high volumes of content. Instead of starting from scratch every generation, you get a reusable foundation that keeps things visually coherent across projects, briefs, and campaigns.
Once trained, your custom model becomes part of your workflow. You can generate new ideas aligned to your aesthetic, reuse the model across projects, briefs and campaigns and produce at scale without losing what makes your work distinctive.
Privacy and IP protections
Adobe is making two things clear here. First, custom models are private by default. Images you upload to train your model won’t feed into Adobe’s general Firefly training data. Second, users must confirm they have the rights to everything they upload before training begins.
But how far does that protection actually go? According to The Verge AI, Adobe spokesperson Frankie Tobin pointed to one concrete safeguard: Firefly automatically checks uploaded images for Content Authenticity Initiative credentials. If a creator has opted their work out of AI training, the system blocks those assets from being used.
Beyond that, it’s mostly an honor system. Users click through a consent modal certifying they have the necessary rights and permissions. That’s a familiar approach, but it puts the legal burden squarely on the user.
How this compares to the competition
Adobe has positioned Firefly as the “commercially safe” option in the AI image generation space. Its base models train on licensed and public domain content, unlike competitors that likely scraped copyrighted works. Custom Models extends that pitch: your training data stays yours, your outputs carry fewer legal risks.
What stands out here is the enterprise angle. While tools like Midjourney and Stable Diffusion offer fine-tuning capabilities, Adobe is betting on workflow integration. If you’re already inside the Creative Cloud ecosystem, this slots right in.
Who can use it and when
Firefly Custom Models first appeared as a private beta at Adobe Max last year. Now it’s open to everyone as a public beta. Adobe hasn’t detailed specific pricing for the custom model feature separately, but it lives within the broader Firefly ecosystem.
This is a smart play from Adobe. The company isn’t just selling an image generator. It’s selling consistency at scale, which is exactly what brand teams and content studios need. Whether the IP safeguards hold up under real-world pressure is another question entirely.
More details are available in the original report from The Verge AI.