An AI reconstruction brings forgotten sculptors into view. It also asks who gets to invent the missing studio

There is a peculiar form of disappearance that happens when history keeps the face of an artist but loses the room in which she worked. A portrait survives; perhaps a letter. A few works reach a collection. The studio, with its materials and habits and the ordinary evidence of a professional life, is gone. A new Paris exhibition is using generative AI to imagine some of those missing spaces, while insisting that the women who occupied them were artists rather than decorative figures in someone else's account.[1][2]

Visionaries, which opened at the Petit Palais on 20 October, centres on new plaster portraits by contemporary sculptor Prune Nourry. The exhibition highlights eighteen historical women sculptors working in Paris around the late nineteenth and early twentieth centuries. Google Arts & Culture contributed to a ten-minute documentary in which generative imagery animates versions of lost studios. The project offers a revealing use case for AI in museums: making an absence visible without pretending to turn speculation into recovered fact.[1][2]

There was research before there was reconstruction

Google says three art historians spent a year gathering archival documents, including letters, journals and photographic portraits, for the project. Those materials are not simply prompts. They are evidence of people whose professional lives have often been hidden behind the better-documented reputations of male teachers and contemporaries. The women featured include Nancy Elizabeth Prophet, Jane Poupelet and Camille Claudel—artists with distinctive practices and histories, not interchangeable characters in a period scene.[1][2]

The challenge was a shortage of photographs of the studios themselves. Google's technologist-in-residence used generative and image-editing techniques and a custom tool built with Gemini to create three-dimensional environments and simulated camera movements. These environments allow a viewer to move through a plausible workspace rather than staring at a still image. They remain constructions. No algorithm can recover the exact placement of an absent sculpture or determine whether a forgotten workbench stood beside a particular window if the archive never recorded it.[1][2]

That epistemic distinction is the heart of the story. A restoration of a damaged but documented photograph aims to recover something the source contains. A recreation of an unphotographed studio necessarily combines documented clues with interpretative choices. Both may be emotionally effective; they make different truth claims. Museums should tell visitors which details come from the record and which were supplied by a contemporary artist or model.[1][2]

One image shows the opposite of recovery

A sequence in the documentary alters an archival photograph of Antoine Bourdelle's studio. Female students gradually disappear from the image, leaving the men. The effect is deliberately artificial. Its purpose is to dramatise how women can vanish from the story even while their presence was captured by the camera. In this case, openly visible manipulation may communicate a historical argument more honestly than a photorealistic room whose provenance is unclear.[1][2]

There is a potential hazard in rendering historical imagination with convincing visual detail. The more lifelike a recreated studio looks, the easier it is for a screenshot to detach from its caption and circulate as proof of something it never documented. A museum film is encountered with surrounding explanation; a clipped image on social media may not be. Clear on-screen labels and durable archival citations are therefore part of the artistic work, not an administrative afterthought.[1][2]

Nor should the restoration story become an advertisement in which a modern technology company receives credit for the historical artists' achievement. The artists made the works; historians found the documents; Nourry shaped the exhibition; digital tools supplied a new visual language. Keeping those contributions distinct is how the technology can serve, rather than overshadow, the history it helps interpret.[1][2]

The work is what survives the room

There is another way to assess success besides counting exhibition views. Does the project help a visitor remember an artist's name, find surviving examples of her sculpture, understand the training restrictions she faced or notice how museum catalogues have treated her? A beautiful reconstruction without those connections would risk substituting spectacle for recovery. A strong exhibition leads a viewer from the imaginative scene back to the verifiable work and its maker.[1][2]

The possibility is nevertheless compelling. Artificial imagery can give an overlooked archive a spatial, navigable presence, enabling people to picture labour that conventional collections barely represented. Used with disciplined sourcing, it can encourage curiosity and make absences harder to ignore. Used carelessly, it can fill those absences so seamlessly that viewers forget they were ever there.[1][2]

The most respectful outcome is neither to ban reconstruction nor to present it as time travel. It is to invite visitors into an openly interpretative studio, then show them the surviving documents and sculptures that made the reconstruction worth imagining.[1][2]

Report an error or suggest a correction