Oxia Palus bring TextMaster Collection to KnownOrigin


The team at Oxia Palus are resurrecting the world's lost art. Their previous NeoMasters™ work resurrecting lost artwork combined the technologies of spectroscopic imaging, artificial intelligence, and 3D printing to recreate lost artwork that an artist painted over. To coincide with the release of their latest NeoMasters™ with MORF Gallery at Focus Art Fair in the Louvre, Paris, 1-4th September, they're releasing a world first as NFTs on KnownOrigin, namely TextMasters.

TextMasters are of a fundamentally different nature to NeoMasters™, since they stem from archived textual records of destroyed artwork where no visual record exists utilising OpenAI’s DALL-E 2.

This technology allows us to not only generate output from textual records, but also sift through hundreds of possible variations in a short space of time. Whilst there exists an infinite range of possibilities as to what the original work might have been like, we have honed in on a select few which we believe give good insight into that modal space.

The first TextMaster release includes destroyed and missing art from the world’s most famous artist, including Leonardo da Vinci, Johannes Vermeer, Giotto di Bondone, Diego Velázquez, Tintoretto, Sandro Botticelli, Titian, Eugène Delacroix, and Jean-Antoine Watteau. Whilst there exists an infinite range of possibilities as to the original work, all TextMasters are released as 1/1, with only one ever generated for a particular lost piece.

TextMaster Collection:
The first three releases in the TextMaster collection will drop as follows:
1. Diego Velázquez TextMaster NFT – Cupid and Psyche. 1/1.
2. Giotto TextMaster NFT – Painting of the Virgin Mary. 1/1.
3. Leonardo da Vinci TextMaster NFT – The Medusa. 1/1.


TextMaster Collectors will also gain pre-public early access to future TextMaster and NeoMasters™ NFT releases. As part of their first TextMaster release Oxia Palus will be dropping 10 famous destroyed or missing works.


The first TextMaster to be released is ‘Cupid and Psyche’ by Diego Velázquez - A Diego Velázquez TextMaster. The Rokeby Venus is one of Diego Velázquez’ most infamous paintings, not only because it is his only surviving nude, but also because the motif was officially discouraged by the Spanish Inquisition. Three other nudes by the artist are however recorded in 17th-century Spanish inventories, but with no surviving visual representation. Velázquez painted two nudes for the Hall of Mirrors in Madrid’s Royal Alcázar Palace, ‘Cupid and Psyche’ and ‘Venus and Adonis’, but both were lost in the Royal Alcázar fire on 24th December 1734.


Following on from Velázquez, the second TextMaster to be released is ‘Painting of the Virgin Mary’ by Giotto di Bondone – A Giotto di Bondone TextMaster. Giotto is famous for producing the decoration of the Scrovegni Chapel, 1305, in Padua, also known as the Arena Chapel. The decoration is a fresco cycle which depicts the Life of the Virgin and the Life of Christ and is regarded as one of the masterpieces of the Early Renaissance. It is thought that Giotto painted another seminal piece of the Virgin Mary between 1297–1337, but the location of the painting is currently unknown. Records show that the work was given by the poet patron Petrarch to Francesca da Carrara, Lord of Padua, in 1370, after which the trail goes cold.


The third TextMaster to be released is ‘Medusa’ by Leonardo da Vinci, A Leonardo da Vinci TextMaster. The Medusa is described by da Vinci's biographer Vasari as being one of his earliest works, painted before 1500, and is thought to be a motif he painted twice: once on a wooden shield, and once with oils on canvas. Either or both paintings are believed to have been held in the collection of Cosimo I de' Medici of Tuscany, the second Duke of Florence between 1553-59, but was however lost in the second half of the 16th century.

We believe the newfound ability of resurrecting these paintings with the advances in AI that has emerged in recent years to be of utmost importance, and we hope that by starting with some of the most notable lost pieces in history we will be able to generate wider public interest in restoring our collective cultural heritage.


Following on from the first three TextMasters we'll be releasing destroyed and missing art from Johannes Vermeer, Diego Velázquez, Tintoretto, Sandro Botticelli, Titian, Eugène Delacroix and Jean-Antoine Watteau.


Oxia Palus' collaboration with OpenAI’s DALL-E 2 presented an opportunity to incorporate multiple modalities into art recreation. DALL-E 2 blends state of the art NLP models and large scale one-shot image generation models. This is significant because DALL-E 2 is a step towards general rather than narrow AI. Applying the first glimpses of more general AI models to artistic creation is an attempt to capture something ephemeral with computational means.

More specifically for art restoration, this presents us with the opportunity to leverage two modalities when attempting to restore a lost piece. Previously we have only been able to use image generation models and work on pieces where a reliable ground truth already exists. However the larger generality of DALL-E 2 allows us to expand reconstructive work into completely unexplored domains, such as cases where only textual evidence remains of the piece.

Having the DALL-E 2 partially trained on the relevant artist’s surviving work, and then prompting it with textual input to guide its generation allows us to explore what a piece without visual ground truth might have been like. The reliability of the output then rests on two domains: work with which the artist produced during the same period as the lost piece, and textual accounts of its appearance. Whilst this will most likely not recreate the exact original piece, it can give us a range of possibilities as to what the original piece might have been like, and so offers a new strand of insight into art history, aided by machine learning.

A practical application of this exploratory work is that many lost pieces are still intact, but merely misplaced throughout history. There are countless stories of lost paintings cropping up in private collections and most miscellaneous of places. By exploring and visually recreating the various possibilities of a lost piece we might be able to help locate such cases by the sheer fact that private collectors can realise that they might have come across the original piece by visual recollection.

We believe that DALL-E 2 is a progenitor for a new paradigm within visual creation across many domains. It is evident that we're closer to automating image generation with a reasonable quality assurance than we first would have thought. Whilst in its infancy, this technology will soon likely disrupt the visual creative industry and over time reorganise how many businesses operate with visual contractors. This shift opens the door to a new generation of visual artists who leverage both artistic and computational skills to realise their work. The positive side is that much of the lower level handiwork can perhaps be automated in many pipelines, which leaves more room for artists to focus on higher level conceptual domains. This invitation to evolve as artists, and further bridge the sciences and the arts both personally and culturally, is an incredibly exciting paradigm to be part of.


About Oxia Palus:

Founded in 2019, Oxia Palus has pioneered the use of artificial intelligence, spectroscopy and 3D printing to resurrect the world's lost art. Named in NVIDIA’s Top 10 AI stories of 2019 and winning CogX's Best Innovation in Creative Arts 2021, Oxia Palus has reached headlines globally for resurrecting lost art by the world's most famous artists, presenting the world's first NeoMaster exhibition in London in 2021. Oxia Palus was founded in January 2019 by George Cann and Anthony Bourached at University College London. Research Associates Jesper Eriksson and Ryan-Rhys Griffiths joined Oxia Palus in 2021. With potentially thousands of works of art hidden dormant beneath existing paintings, destroyed and missing, the journey in resurrecting the world’s lost art has only just begun.