What happened
Adobe opened its Firefly Custom Models to public beta on March 19, 2026. Creators can now upload 10-30 of their own images to train a custom model that captures their specific illustration style, character design, or photographic look. Training takes 30 minutes to a few hours. The custom models join over 30 industry-leading models already available in Firefly, including Google's Nano Banana 2, Veo 3.1, Runway's Gen-4.5, Kling 2.5 Turbo, and Adobe's own Firefly Image Model 5 (now generally available).
Why it matters
Custom model training democratizes brand-consistent AI image generation. Instead of carefully crafting prompts to approximate a visual style, creators can now encode their aesthetic directly into a model. This is particularly valuable for maintaining character consistency across scenes, preserving illustration stroke weight and color palettes, and replicating photographic looks across campaigns. The integration of 30+ third-party models in a single interface also makes Firefly a model marketplace, reducing the need to switch between different generation tools.
Who should pay attention
- Design teams maintaining brand visual consistency across campaigns
- Illustrators and photographers who want to scale their personal style
- Frontend developers who generate UI assets or marketing visuals