Adobe’s AI image generator can now be trained on your own art
Creators can now feed their own images to Adobe's AI to generate content in their unique style.
Adobe has moved its Firefly Custom Models feature from private to public beta, giving creators a powerful new tool for brand consistency. The system allows users to upload their own image assets—such as character designs, illustrations, or photography—to train a private, custom AI model. This model learns specific visual elements like stroke weight, color palettes, lighting, and character features, enabling it to generate new content that matches the creator's established aesthetic. The tool is designed to help teams produce high volumes of on-brand assets without starting from scratch each time.
A key focus for Adobe is addressing ethical and legal concerns around AI training. The company requires users to confirm they have the necessary rights and permissions for all uploaded content. Furthermore, Firefly automatically checks images for Content Authenticity Initiative (CAI) credentials, preventing the use of assets where creators have opted out of AI training. This positions Firefly as a commercially safe alternative to models trained on scraped web data, aligning with Adobe's broader strategy of promoting ethical AI for professional creators.
- Public beta launch allows anyone to train a private AI model on their own image assets.
- Preserves specific artistic details like character designs, color palettes, and lighting for brand consistency.
- Includes proactive rights checks using Content Authenticity Initiative credentials to prevent unauthorized training.
Why It Matters
Enables brands and creators to scale content production while maintaining their unique visual identity and ensuring commercial safety.