Seeing Katherine Swynford: A Method-Led Visual Interpretation

Evelyn circa 1930

Tomb effigy of John Beaufort Wimborne Minster 

William Dugdales record of Katherine’s tomb prior to English civil war

 

Joan Beaufort/Neville 1430 Neville Book of Hours

Katherine Swynford in 1396

This image is a method-led visual interpretation informed by the Dugdale drawing of her tomb brass, Beaufort family imagery, and forensic methodology. It does not claim to show her true appearance, but to test how far a credible image can be produced from fragmentary evidence.

Clearly, we cannot — and never will — know what Katherine Swynford looked like. That uncertainty is absolute. The question behind this project was therefore not how to recover a lost face, but whether AI could be used in a disciplined way to produce an image that might offer something credible, cautious, and intellectually interesting from a handful of fragmentary sources.

The starting point was limited, and intentionally so: a seventeenth-century drawing of Katherine’s tomb brass recorded by William Dugdale; verified representations of her daughter, Joan Beaufort; the tomb effigy of her son, John Beaufort; and, as a final comparative check, a photograph of my grandmother Evelyn, Katherine’s eighteenth great-granddaughter. None of these sources can provide a likeness. Taken together, however, they offered a framework within which questions of proportion, restraint, and plausibility could be explored.

Using AI in this way required far more constraint than freedom. Considerable effort went into pushing the process away from aesthetic guesswork and towards method. The system was repeatedly challenged to engage with current forensic and anthropological approaches, to respect clear evidential boundaries, and to refuse the temptation to resolve uncertainty through invention. The methodology came first; the image followed.

The temporal focus of the image is Katherine Swynford in 1396, the year of her marriage to John of Gaunt. This moment was chosen carefully. It places Katherine late in life, newly elevated in status, and acutely conscious of scrutiny. The image does not attempt to represent youth or idealised beauty, but a mature woman at a precise historical juncture, shaped by experience rather than symbolism.

One strand informing this interpretation derives from the broader late medieval effigial tradition. This does not imply the existence of a surviving effigy of Katherine Swynford — none exists — nor does it provide facial detail. Instead, it establishes a visual discipline. High-status female commemoration in the later fourteenth century prioritised composure, symmetry, and recognisable social identity over individual likeness (Binski 1996; Saul 2011). Faces were generalised, expressions neutral, and emotional display deliberately suppressed. These conventions inform the tone of the image — its stillness, proportional balance, and restraint — rather than any specific features.

A second strand comes from comparative family imagery, most importantly the surviving representations of Katherine’s daughter, Joan Beaufort. These images are not treated as portraits to be copied. Rather, they provide a framework within which familial plausibility can be considered. Medieval family commemoration often relied on repeated proportional relationships rather than individual likeness (Camille 1992). Joan’s representations therefore allow broad facial relationships — such as facial width, eye spacing, and jaw proportion — to be tested without collapsing mother and daughter into the same face.

The tomb effigy of John Beaufort constitutes a third, separate strand. As a male effigy, it cannot inform soft-tissue detail or female facial structure. Its value lies elsewhere. It reinforces a family visual language characterised by restraint, symmetry, and structural clarity, typical of high-status Beaufort commemoration (Binski 1996). In this project, John Beaufort’s effigy functions as a control rather than a source, supporting proportional coherence while explicitly refusing to dictate appearance.

Distinct from these historical strands is the cautious use of a known female descendant. This element was introduced not to assert resemblance, but to test facial balance, asymmetry, and structural plausibility across generations. In forensic and anthropological practice, modern relatives are sometimes consulted not to predict likeness, but to explore the range within which features may plausibly fall (Wilkinson 2004). The descendant’s photograph was therefore used analytically rather than illustratively: as a means of checking proportion, facial harmony, and the interaction of features, rather than as a template to be reproduced.

This strand carries no historical authority, and its limitations are acknowledged openly. Its role is preventative rather than generative. It resists over-smoothing, idealisation, and abstraction, ensuring that the image remains recognisably human and individual while stopping well short of claiming genealogical certainty.

Throughout, the project was guided by the discipline demonstrated by the Richard III facial reconstruction, not as a technical model but as an intellectual one. That reconstruction showed how powerful an image can be when it privileges method over spectacle and transparency over resolution (Buckley et al. 2013; Appleby et al. 2014). The influence here lies in the attitude adopted: a willingness to stop short of certainty and to allow ambiguity to remain visible.

Equally important are the constraints that shape what this image refuses to do. There is no crown or coronet, because Katherine Swynford was a duchess by marriage, not a queen. There is no later medieval fashion, no symbolic regalia, and no attempt at idealisation. Age is allowed to register. Dress is restricted to a simple linen veil and wimple appropriate to elite women in the 1390s, consistent with manuscript and effigial evidence (Scott 2007). The image avoids drama because drama is not evidence.

What emerges is not an answer to the question “What did Katherine Swynford look like?” but a more modest and more honest proposition: what a responsible image might look like if it takes its cue from evidence, method, and refusal rather than invention.

If the image feels understated, even austere, that is intentional. It reflects an approach in which historical responsibility is valued over visual appeal — the opposite of what is often produced by unstructured, speculative AI image generation.

In that sense, Dugdale’s illustration remains central to the project. Not as a source of facial detail, but as a reminder of how easily images acquire authority, and how carefully they must be handled. The image presented here is not intended to replace Dugdale’s drawing, but to sit alongside it: a transparent, method-led response to its presence and its limits.

⸻

References

Appleby, J., et al. (2014).

The reconstruction of Richard III: anatomical, archaeological and historical perspectives. Antiquity, 88.

Binski, P. (1996).

Medieval Death: Ritual and Representation. London: British Museum Press.

Buckley, R., et al. (2013).

The King in the Car Park: New Light on the Death and Burial of Richard III. Antiquity, 87.

Camille, M. (1992).

Image on the Edge: The Margins of Medieval Art. London: Reaktion Books.

Saul, N. (2011).

English Church Monuments in the Middle Ages. Oxford: Oxford University Press.

Scott, M. (2007).

Medieval Dress and Fashion. London: British Library.

Wilkinson, C. (2004).

Forensic Facial Reconstruction. Cambridge: Cambridge University Press.