Adobe Foundry wants to rebuild Firefly for your brand — not just tweak it
Hoping to attract more business teams to its ecosystem, Adobe launched a new model customization service called Adobe AI Foundry, which would create customized versions of its flagship AI model, Firefly.
Adobe AI Foundry will work with enterprise customers to redesign and retrain the architecture Firefly Models specific to the client. AI Foundry build models differ from Firefly custom models in that Foundry models understand multiple concepts compared to custom models with only one concept. These models will also be multimodalwhich offer a broader use case than custom Firefly models, which can only record and respond to images.
Adobe AI Foundry models, based on Firefly, know a company’s brand tone, image and video style, products and services, and all its intellectual property. Based on this information, the models generate content for whatever use case the company wants.
Hannah Elsakr, vice president of GenAI New Business Ventures at Adobe, told VentureBeat that the idea to create AI Foundry came about because enterprise customers wanted more advanced custom versions of Firefly. But no matter how complex enterprise needs are, Adobe will take charge of the redesign rather than handing the reins to customers.
“We will retrain our own commercially secure Firefly models with the corporate IP. We keep that IP separate. We never take that back into the base model, and the enterprise itself owns that output,” Elsakr said.
Adobe will deploy the Foundry version of Firefly through its Firefly Services API solution.
Elsakr likened AI Foundry to a consulting service, as Adobe will have teams work directly with enterprise customers to retrain the model.
Deep tuning
Elsakr calls Foundry a deep-tuning method because it goes beyond just refining a model.
“The way we’re thinking about it, maybe more in layman’s terms, is that we’re surgically reopening the Firefly-based models,” Elsakr said. “So you get the benefit of all the knowledge in the world from our image model or a video model. We go back in time and pull in the IP of the company, such as a brand. It can be images of a shooting style, regardless of what they are licensed for. Then we retrain. We call this continuous pretraining, where we consider the model to determine some things differently. So we are literally retraining our base model, and that’s why we call it ‘deep tuning’ instead of ‘fine tuning’.”
Part of the training pipeline involves Adobe’s embedded teams working with the company to identify the data they need. The data is then securely transferred and recorded before being tagged. It is fed into the base model and then Adobe begins a pre-training model run.
Elsakr claims that the Foundry versions of Firefly will not be small or distilled models. Often companies’ additional data extends Firefly’s parameters.
Two early customers of Adobe AI Foundry are Home Depot and Walt Disney Imagineering, Disney’s research and development arm for its theme parks.
“We are always exploring innovative ways to improve our customer experience and streamline our creative workflows. Adobe’s AI Foundry represents an exciting step forward in embracing cutting-edge technologies to deepen customer engagement and deliver impactful content across our digital channels,” said Molly Battin, senior vice president and chief marketing officer at The Home Depot.
More customization
Companies often turn to refinement and model adaptation to bring large language models with their vast external knowledge closer to the needs of their business. Fine-tuning allows business users to use models only in the context of their organization’s data, so the model doesn’t respond with text that is completely unrelated to the business.
However, most organizations do the fine-tuning themselves. They connect to the model’s API and start retraining it to respond based on their ground truth or their preferences. Several fine-tuning methods exist, including some that are possible with just one prompt. Other model providers are also trying to make it easier for their customers to refine models, for example Open AI with his o4-mini reasoning model.
Elsakr said she expects some companies will have three versions of Firefly: the Foundry version for most projects, a custom Firefly for specific single-concept use cases, and the base Firefly because some teams want a model that is less encumbered by corporate knowledge.
