Profile
Saved portraits
Person Gallery
More portraits
At a glance
Corpus in numbers
Medium
Technique mix over time
Composition of printmaking techniques among dated portraits, per decade (100% stacked). The lithograph band appears only after ~1800 — a built-in dating clock.
Rankings
Top places & most-portrayed people
place_best is a single field per object — the best available place, chosen by a waterfall: strict production place where known, else broad production place, else the publisher's/printing seat (a proxy), else the catalogue origin field. Publisher seats actually make up the majority (20,150 of 37,224), so read this as "the place associated with the print", not strictly where it was engraved.
Dating
Object date precision & range widths
Only ~47% of portraits carry a parsed object date. The catalogue often gives a bounded range rather than an exact year: the visible site may show "um 1700", but the source metadata stores bounds like 1690–1710 (wider when the year ends in a zero). Range widths are shown for the dated-but-not-exact objects. Very wide ranges are frequently an artefact — often the object was only bounded by the engraver's lifetime (≈62% of ranges ≥50 yrs have start = creator birth & end = creator death).
Sources
Holding institutions
The Portraitindex aggregates from many collections. The Österreichische Nationalbibliothek (ONB) alone supplies ~50% of the corpus, followed by the Herzog August Bibliothek, Germanisches Nationalmuseum and Staatsbibliothek zu Berlin. Image delivery mirrors this: onb 50% · direct 40% · hab 10% · Heidelberg <1%.
Time Series
Portrait metadata over time
Counts
Social composition of the corpus
Prevalence plots each tag as a share of all portraits that decade — the lines are independent, may overlap (a ruler is also counted under nobility) and need not sum to 100%. Social composition is a genuine partition: every portrait is assigned to exactly one stratum by dominant status (ruler → clergy → nobility → non-noble military → other/commoner), so the stack covers 100% of portraits, including the large ordinary-sitter residual. Absolute counts shows raw numbers. (The earlier “100% stacked” view mixed overlapping, non-exhaustive tags and is retired.)
Map
Territory layer and portrait places
Treatment panel
Panel cities: treated vs untreated
Prompt Results
Alternative display prompt shares
New Prompts
Inspect pilot classifications
Prompt
Exact instruction sent to Gemini
Examples
No field selected
Three
Religion, clothing, and surround pilots
Prompt
Exact instruction sent to Gemini
Examples
No field selected
Frankfurt National Assembly
Paulskirche portrait comparison
No selection
No Paulskirche portrait loaded
Selected portrait
Spectrum overview
Portrait examples by faction position
Explore
Compare two portrait categories
Overlap 0
Category setup Change filters
PyFeat detected
Portraits with more than one person
Based on PyFeat face counts. This is a visual signal and should be read as detected faces, not a manually verified person count.
PyFeat
Facial expression and detection series
1800 Scan
Largest group contrasts
Auxiliary Person Data
lobid / GND manual review sample
0 / 0
No lobid sample loaded
Portraitindex main metadata
lobid GND JSON
Show raw lobid JSON
Sidebar QA
Show Conflicts
0 / 0
No conflict selected
Sidebar vs. Website
Metadata comparison
These are dataset-level statistics across all portraits, not values for the portrait currently shown above.
Duplicates / Across data sources
Within the same portrait
TBC
Model vs. human / logical residuals
Prompt Residuals
TBC
Blind labelling
Hand-check
Pick one field and label just that item on random portraits (1700–1900). The model's own answer is hidden so labels stay unbiased. Each field is checked 100×; your labels are backed up and never overwritten.
Protected Area
Diagnostics access
Location Matching
Treatment crosswalk audit
Available place variables
This shows which place-like fields exist and how much coverage they have. These are candidates for treatment/location matching.
Most common unmatched places
These are the names that the archived exploratory crosswalk did not match to the city panel. They are the best manual-check targets.
Metadata Overlap
Place fields
Archive
Archived prompt pilots
Examples
No archived prompt selected
Full corpus
The militarisation curve
Share of portraits with military elements, by decade (full corpus, decade-level like the curves on the other P: tabs). Uniforms and orders/decorations rise sharply after ~1790 while armor disappears — the militarisation of the portrait runs through modern insignia, not through the old knightly iconography.
Gemini
Military element shares over time
Gemini
Military element shares: group comparison
Treatment
Treated vs. untreated over time
Full corpus
The secularisation curve
Share of portraits with any religious display, by decade (full corpus). Religious self-presentation is stable near 40% through ~1750, then collapses across the Enlightenment to under 10% after 1800 — driven almost entirely by the loss of religious text (name/office inscriptions) and clerical attire.
Prevalence & who carries it
Field rates and social groups
Clergy carry religious display at 83% vs 16% for everyone else — the prompt cleanly separates the clerical estate. Women (9.7%) rarely carry it, since the signal is dominated by male clerical and scholarly portraits.
Treatment
Treated vs. untreated over time
Male portraits
The great male renunciation
Luxury textiles (embroidery, brocade, silk, lace, fur) fall from ~24% of male portraits before 1760 to ~8% after 1790, while worn orders and sashes climb from ~8% to over 30% and the plain sober coat spreads. Men trade ancien-régime finery for uniforms, decorations and dark coats across the revolutionary decades — the “great male renunciation.” (is_male corpus only.)
Male portraits
The dress-elaboration curve
Distribution of the dress-elaboration score (1–5) over time. Level 3 (“respectable”) dominates; the very plainest (1) and full court dress (5) stay thin. Use “Levels” to collapse to low (1–2) / medium (3) / high (4–5), and “Group” e.g. to compare kings vs civilian men.
Prevalence & dress score
Field rates, dress elaboration and social groups
Binned: simple (1–2) / medium (3) / complex (4–5)
Nobles dress far more elaborately (mean dress score 3.41; 50% at level 4–5) than non-nobles (2.77; 8%). Clergy are the soberest estate. Dress score uses the full 1–5 scale.
Gallery
Example portraits by dress level (1–5)
Three example portraits for each dress-elaboration level, from level 1 (plain sober coat) to level 5 (full court / ceremonial dress).
Treatment
Treated vs. untreated over time
Full corpus
The staging curve
Portrait surrounds simplify sharply across the Enlightenment: mean elaboration falls from ~2.8 (1740s) to ~1.6 (1820s), and the share of “elaborate” surrounds (level 4–5) collapses from ~20% to ~1%. The ornate baroque frame, allegory and heraldry give way to a plain neoclassical ground — steepest in the 1780–1810 revolutionary window.
Distribution & who carries it
Elaboration levels and social groups
Binned: simple (1–2) / medium (3) / complex (4–5)
Nobility carry the most elaborate surrounds (mean 2.69; 22% at level 4–5) vs non-nobles (2.41; 12%). The staging of status via allegory and heraldry is a noble idiom that fades with the baroque.
Treatment
Treated vs. untreated over time
Gallery
Example portraits by elaboration level (1–5)
Three example portraits for each surround-elaboration level, from level 1 (bare ground, all attention on the sitter) to level 5 (overwhelming ensemble of frame, allegory, heraldry and inscription).