Adobe Firefly: 30 AI Models and Custom Styles

Adobe Firefly Integrates Over 30 AI Models

Adobe has recently expanded the capabilities of its artificial intelligence platform, Firefly, by introducing custom models that allow users to create specific visual styles. By using between 10 and 30 personal images, users can train a model to consistently reproduce a visual style. This innovation is part of a broader strategy aimed at providing advanced creative tools to a wider audience.

The Firefly platform now brings together over 30 AI models from various providers, such as Google, Runway, OpenAI, and Kling 2.5 Turbo. This unique integration enables users to generate, refine, and compare results across different models, with unlimited image and video generation capabilities, according to Adobe's statements.

At the same time, Adobe is developing the Moonlight project, which focuses on conversational AI assistants. These assistants are designed to operate across multiple Adobe applications, allowing users to describe their goals in a chat while the agents perform specific actions using Adobe tools.

Custom Models for Users

With the introduction of Custom Models, Adobe now allows users to train their own reusable AI models. By uploading between 10 and 30 images in JPG or PNG format, users can capture a precise visual style. This feature is particularly optimized for three use cases: illustration styles, characters requiring a consistent appearance, and photographic styles to be replicated across numerous images.

The trained models are designed to preserve crucial details such as line weight, color palettes, lighting, and character features across generations. Once created, these models can be reused in various projects and campaigns. By default, the models remain private, and users retain ownership of all generated content.

When training a custom model, Firefly analyzes the uploaded images and suggests improvements to ensure style consistency. For example, it can identify images with mismatched color palettes or those that do not align with the desired style.

Adobe positions this feature as particularly beneficial for teams producing large volumes of content that need to maintain visual consistency. The company plans to refine this feature based on feedback from real-world workflows.

Over 30 AI Models in One Place

Firefly now consolidates over 30 AI models in a single environment. In addition to Adobe's Firefly Image Model 5, which is now generally available, the platform includes models from Google, such as Nano Banana 2 and Veo 3.1, as well as Runway Gen-4.5 and OpenAI. The Kling 2.5 Turbo model is the latest addition to this collection.

The idea behind this platform strategy is to allow users to generate with one model, refine with another, and compare results side by side. Adobe claims to currently offer unlimited video and image generation across all available models.

Moonlight Project: Turning Prompts into Conversations

Adobe is also continuing the development of its conversational AI assistants. The company is deploying these assistants across several products, including Photoshop, Express, and Acrobat. Users can describe their needs in a chat, and the agents perform specific actions using Adobe tools.

The Moonlight project automatically generates content ideas from uploaded photos, complete with formatting and descriptions. For example, the assistant can create three social media post suggestions from a dog photo album.

This initiative, first presented at Adobe MAX 2025 in October, is a conversation-based interface, powered by agents, designed to work across multiple Adobe applications. It is built to learn a user's style and leverage their own assets and libraries. The project is currently in private beta, and Adobe is now expanding access through a sign-up form.

Since MAX, the company has gradually expanded its offerings. In February, Adobe introduced Quick Cut, an AI editing feature that automatically transforms raw footage into a structured first cut. Around the same time, Adobe integrated Photoshop, Acrobat, and Express directly into ChatGPT, allowing users to edit images and documents by typing text prompts into the chatbot.