Adobe Firefly Custom Models lets you train AI on your unique photographic style

Adobe Firefly Custom Models Enter Public Beta

Adobe has launched Custom Models into public beta for Adobe Firefly, marking a significant shift in how creators can use generative AI tools. The feature allows artists and brands to train AI on your own work, creating personalized models that generate images aligned with specific visual styles, characters, or photographic aesthetics 1

1

. According to Adobe's Deepa Subramaniam, the goal is to transform Firefly into "your all-in-one creative AI studio that brings together the industry's top AI models and the best multimodal creative tool" 1

1

.

The Custom Models feature directly tackles one of the most persistent concerns about generative AI: the unpredictability of AI outputs. By uploading your own images, you can create a model that captures specific elements like stroke weight, color palette, and lighting characteristics 3

3

. This approach enables creators to scale content production while maintaining cohesive visual identity across multiple projects, campaigns, and platforms.

Training AI to Align with Specific Brand Aesthetics

Custom Models are optimized for three primary creative applications: illustration styles, character design, and unique photographic style. For illustration work, the system analyzes and preserves distinctive visual elements such as stroke weight, fills, and color consistency 1

1

. Character design capabilities allow creators to reproduce the same subjects more consistently across multiple images, particularly useful for comic book artists or brand mascot development. The photographic style training enables photographers and visual teams to replicate specific looks across entire campaigns.

"No matter who you are, it takes years of investment building a visual identity. Maintaining that across media, campaigns, formats and platforms takes intention," Subramaniam explains. "Upload your assets, and Firefly analyzes and trains a model aligned to your aesthetic" 1

1

. The models remain private by default, ensuring that content created with them stays entirely under the creator's control 2

2

.

Expanding Firefly Into a Comprehensive AI Workspace

Beyond Custom Models, Adobe is transforming Firefly into a unified workspace for AI-driven content generation. The platform now provides access to more than 30 models from Adobe and partners including Google, OpenAI, Runway, and Kling 2

2

. Recent additions include Google's Veo 3.1, Runway Gen-4.5, Firefly Image Model 5, and Kling 2.5 Turbo, allowing users to generate content with one model, compare results with another, and continue editing without switching platforms 2

2

.

The platform also introduces new tools like Quick Cut, which transforms raw footage into rough edits within minutes, alongside capabilities to add or remove objects and extend scenes 2

2

. Adobe is experimenting with agentic AI approaches through initiatives like Project Moonlight, which can automatically turn random clips into polished edits, suggesting where AI-driven creation might evolve next 2

2

.

Delivering Brand Consistency at Scale

For teams producing high volumes of content, Custom Models offer what Adobe positions as a competitive advantage through workflow integration. Once trained, custom models become reusable assets that can generate new ideas aligned to established aesthetics across multiple projects and campaigns 1

1

. This capability addresses a critical need for brands requiring consistent on-brand visual assets without sacrificing what makes their work distinctive.

Several major brands, including Tapestry and Deloitte Digital, have already used Custom Models to scale on-brand creativity 3

3

. Adobe continues to emphasize that Firefly is trained on commercially safe data, making it suitable for commercial use 2

2

. The public beta release signals Adobe's commitment to addressing creator concerns about control and consistency while expanding AI capabilities for professional workflows.