In the shifting waters of the Gulf of Maine one of the fastest-warming marine regions on Earth a team of researchers at MIT is developing an innovative tool to visualise and protect vulnerable ocean life. The initiative, called LOBSTgER (Learning Oceanic Bioecological Systems Through Generative Representations), is blending cutting-edge artificial intelligence with underwater photography to create new forms of visual storytelling that bridge science, technology, and art.
Launched through MIT Sea Grant, the LOBSTgER project is co-led by award-winning underwater photographer and visiting artist Keith Ellenbogen and mechanical engineering PhD student Andreas Mentzelopoulos. Together, they’re exploring how generative AI can amplify field-based visual data to reveal unseen or impossible-to-capture scenes from beneath the waves.
Photography Meets Machine Learning Beneath the Surface
The project centres on Ellenbogen’s meticulously curated photographic archive, drawn from years of diving in the cold, low-visibility waters of New England. From lion’s mane jellyfish and blue sharks to elusive American lobsters and ocean sunfish, Ellenbogen’s images combine artistic vision with scientific precision, documenting the Gulf’s biodiversity under challenging conditions.
“This work requires multiple dives, missed opportunities, and unpredictable conditions,” Ellenbogen explained. “But these challenges are part of what makes underwater documentation both difficult and rewarding.”
Mentzelopoulos has translated this visual database into a suite of custom-built latent diffusion models, trained from scratch using original code. These models can produce entirely new, yet scientifically grounded, images or enhance existing photos by correcting lighting, recovering obscured detail, or simulating scenes that are virtually impossible to capture in the field.
“Our goal isn’t to replace photography,” Mentzelopoulos emphasised. “It’s to build on and complement it, making the invisible visible and helping people see environmental complexity in a way that resonates both emotionally and intellectually.”
A Hybrid Method for Telling Ocean Stories
LOBSTgER’s approach represents a rare convergence of marine biology, computational science, and artistic expression. The AI is not just creating synthetic images; it’s internalising fine-grained details of light, colour, and behaviour from thousands of Ellenbogen’s real-world photographs. From refracted sunlight to the glint of sediment, the resulting imagery captures both the aesthetic and ecological truths of underwater life.
The process is twofold. Ellenbogen gathers data in the field, often in unpredictable conditions, while Mentzelopoulos refines the AI models in the lab, training them to replicate both the subject matter and the emotional tone of those encounters. The models can operate unconditionally, producing new scenes from learned data or conditionally, enhancing or modifying specific input photos.
This hybrid workflow has practical implications. It accelerates the curation of marine imagery, aids in visual clarity in difficult shooting conditions, and provides a means of storytelling that blends fact and imagination, much like early photography did in the 19th century when it brought distant environments into public consciousness for the first time.
Why the Gulf of Maine and Why Now?
The Gulf of Maine is warming faster than 99 percent of the world’s oceans. This dramatic shift threatens marine life and destabilises ecosystems at a pace that outstrips traditional monitoring tools. By enhancing how we visualise these changes, LOBSTER offers a new way to educate and engage the public.
The project builds on Ellenbogen’s broader mission, Space to Sea: Visualising New England’s Ocean Wilderness, which documents the region’s marine life in unprecedented visual detail. Now, with the help of AI, these images don’t just show what’s there; they also hint at what could soon be lost.
Art, Science, and the Future of Ocean Communication
What makes LOBSTgER stand out is not just its technology but its philosophy. The team sees photography and AI not as conflicting tools, but as a continuum where one captures the real, and the other extends our ability to see, imagine, and infer.
As AI-generated visuals become more prominent across media, LOBSTgER sets a precedent for ecological accuracy, artistic sensitivity, and ethical application. It’s a blueprint for using technology not simply to impress, but to inform, educate, and inspire action in the face of environmental urgency.
In a world where oceans are changing faster than ever, LOBSTgER is more than a research tool; it’s a visual language for the Anthropocene. As the project evolves, the team plans to expand its methods to marine ecosystems beyond the Gulf of Maine, offering a glimpse into the future of scientific storytelling where data, design, and devotion to the natural world come together beneath the waves.
