Azay-le-Rideau, France
Musée Maurice Dufresne
A travelling guillotine of 1794, on its cart.
Episode 01 · Paris, 1792
Rebuilt from its own 1792 paperwork: how it was built, how it works, and why every part is shaped the way it is.

In March 1792 a surgeon wrote to a carpenter with a list of dimensions, and almost every line in that letter is an engineering decision. Each chapter here follows one in the video: jump to the part you came for, or back into the video at that point. Every claim is tagged with its source, like this: S03. Metres first, the 1792 French units in brackets.

A brief written for a machine: no human skill in the loop.

Two posts, a weight, a blade and a rope, from Louis's numbers.

One cycle, the physics of the drop, and a calculator.

Why fourteen feet, why the slant, why grooves.

A snag list, swollen grooves, and the fixes.

The numbers, and who they were.

Four scores out of ten. The first entry on the board.
As promised in the video
This is the model we built the episode with: twenty-one parts, every size from the 1792 paperwork. Spin it, drop the blade in slow motion, or watch all seventeen assemblies come off one by one.
Reconstruction from the 1792 specification; dimensions in metres are model values. Speeds ignore friction. The full model runs best on a laptop or desktop; phones get a lighter version.
Every part is tagged by how sure we are: attested reconstructed assumed illustrative. Turn on "Evidence colours" in the full model to see them.
Chapter I
A brief written for a machine
In eighteenth-century France, even your execution depended on your class: nobles got the sword, commoners the rope. And a sword is only as good as the arm swinging it. In 1766 the Comte de Lally needed several blows. S01

Proposed it
Doctor and deputy. Asked for one method for all classes.

Specified it
Surgeon. Wrote the requirement and the dimensions. It was briefly called the Louisette.
AI-generatedBuilt it
Harpsichord maker. No portrait of him is known, so this one is painted.


Chapter II
Two posts, a weight, a blade and a rope
Louis's letter to the carpenter survives, so we can build it from his own numbers. A machine of roughly that age survives too, in a museum in Venlo, in the Netherlands. S03 S27
And the whole thing is a kit. The joints are mortise and tenon, locked with removable pins, so it can be knocked down, loaded onto a wagon and rebuilt somewhere else. The carpenter's first quote even included horses and wagons to set it up and take it down for every execution. S04 S27
| ID | Part | Size in the model | Evidence |
|---|---|---|---|
| P01 | Scaffold deck, railing and 10-step stair | 4.0 × 3.8 m, 2.0 m high | reconstructed |
| P02 | Sole (base frame) | sills 0.20 × 0.20 m; 1.10 × 2.40 m | attested |
| P03 | Uprights (2) | 0.162 × 0.20 × 4.1 m; gap 0.325 m | attested |
| P04 | Grooves with copper liners | square, 0.027 m deep | attested |
| P05 | Crossbeam (chapeau) | 0.85 × 0.25 × 0.22 m | attested |
| P06 | Pulleys (2 + outer pulley) | ≈0.12 m; outer ≈0.08 m | attested |
| P07 | Side and rear braces (4) | 0.10 × 0.10 m | attested |
| P08 | Mouton with lugs, ring and crash bar | 0.32 × 0.30 × 0.08 m | reconstructed |
| P09 | Blade (couperet) + 3 bolts | 0.30 m wide; edge 45° | reconstructed |
| P10 | Hoisting rope | ≈0.02 m hemp, two falls | reconstructed |
| P11 | Rocker-arm release, rod and handle | pivot ≈4.0 m; handle ≈1.3 m | reconstructed |
| P12 | Lunette, lower half | 0.365 × 0.20 × 0.06 m; hole 0.13 m | reconstructed |
| P13 | Lunette, upper half | as P12; slides in its own grooves | reconstructed |
| P14 | Bascule (tilting plank) with 2 straps | 1.2 × 0.32 × 0.05 m | attested |
| P15 | Bascule table and carriage | top at 0.875 m; 1.7 m long | reconstructed |
| P16 | End of the grooves (impact point) | stops, top at 1.22 m | reconstructed |
| P17 | Head basket | ≈0.5 × 0.4 × 0.35 m | illustrative |
| P18 | Body basket | ≈1.8 × 0.6 × 0.6 m | illustrative |
| P19 | Paint | ox-blood red | attested |
| P20 | Straw bundle (demonstration target) | ≈0.12 m bundle | illustrative |
| P21 | Joint pins (mortise-and-tenon locks) | iron pins at each frame joint | reconstructed |


AI-generatedChapter III
One cycle, from the bascule to the reset
Gravity does the cutting. The engineering is in holding forty kilograms four metres up, putting the target in exactly the same place every time, and letting go cleanly.
E = m·g·h ≈ 40 kg × 9.81 m/s² × 2.51 m ≈ 0.99 kJ
v = √(2gh) ≈ 7.0 m/s (25 km/h)
t = √(2h/g) ≈ 0.72 s
To picture a kilojoule: a 25 kg bag of cement dropped from four metres, all of it focused onto one sharp edge.
In the 1860s an observer timed the whole process, from the first step to the basket lid closing, at fourteen seconds. S08
Chapter IV
Why fourteen feet, why the slant, why grooves
So why is it built like this? Three choices make it work, and every one of them goes back to Louis's brief.
Louis compared the weight to a pile driver's ram, "whose force increases with the height from which it falls". S01
Double the drop and you double the energy. But the speed only rises by about 40%, and a taller frame is harder to build, keep straight and reset. Fourteen feet was their balance. They didn't calculate it the way we can; they had a rule of thumb, and they tested it. Try doubling it in the calculator.
Louis's report gives the reason. Under a microscope a cutting edge is just a very fine saw, and a saw cuts by sliding, not pressing. You know this from your kitchen: press a knife straight down on a tomato and it squashes; slide it, and it glides through. S01
Angle the blade at 45° and for every centimetre it falls, it also slides a centimetre sideways. In the ideal case that halves the force needed. And because the edge meets the neck at one point and works across, only a little of it is cutting at any instant. S19
Everything is built for repeatability. The grooves make the blade fall in the same line every time; the bascule and lunette put the neck in the same place every time. That was Louis's brief: no human skill in the loop. S01
AI-generated
AI-generatedChapter V
A snag list, swollen grooves and the fixes
No machine works perfectly the first time, and this one had problems fast. In June 1792 an architect, Giraud, inspected the first machine, called it "made in haste", and wrote a snag list that reads like an engineer's. S07
Built to Louis's numbers, the lugs have only a few millimetres of play, and damp oak swells: just 1% on a 16 cm post is 1.6 mm (our model's estimate). In July 1792 the grooves swelled and the blade failed to cut all the way through. Metal doesn't swell, so metal-lined grooves became mandatory. S07 S28
On the earliest machines the rope stayed tied to the weight and fell with it. When it snagged, executions went badly wrong. And the energy left over hit the bottom of the grooves with something like twenty tonnes of force for an instant (about 200 kN if it stops within 5 mm; an illustration, not a measurement). Again and again it knocked the frame out of line. By the 1860s the weight landed on springs topped with rubber. S08 S27
Around 1870 France rolled every fix into one national design, the Berger model, and that's the guillotine it used until 1977. S09 S10

AI-generatedChapter VI
The numbers, and who they were
It did its job with terrible efficiency. On 31 October 1793 the Paris executioner and his team executed twenty-one deputies in thirty-six minutes. S15
During the Reign of Terror, 16,594 official death sentences were carried out across France in just over a year, most of them by guillotine. It took a king, a queen, the chemist Lavoisier and, in the end, the revolutionaries who had used it, Robespierre included. S11 S12 S15
About eight in ten of those people weren't nobles at all. They were craftsmen, farmers, ordinary people.Of those condemned, 31% were artisans and 28% peasants S29
A machine designed to make death equal also made it fast, cheap and easy to repeat. France kept using it until 1977, and abolished the death penalty four years later. S11
Chapter VII
Four scores out of ten
Every machine on this channel gets four scores out of ten: how new the engineering was for its time, how well it did its job, how much it changed its world, and how much of its engineering lives on.
It's the first entry on the board, so for now it's in first place, and last place. Which machine should challenge it next? Tell us in the video's comments.
Ranked by total · ties share a place · the rulers run 0 to 10 in half-point steps
A few machines and their ancestors survive in museums. Displays change, so check before you travel.

Azay-le-Rideau, France
A travelling guillotine of 1794, on its cart.

Nuremberg, Germany
The Dillenburg machine, used until 1813: the wooden frame and its grooves.

Edinburgh, Scotland
The Maiden of 1564, the ancestor from chapter I.

Geneva, Switzerland
The Geneva guillotine; its frame was restored in the early 20th century.
Venlo, Netherlands
A machine of about 1792, roughly the age of the one we rebuilt. S27
Marseille, France
A Berger guillotine of 1872, rarely shown; it was displayed in October 2025. S24
The primary sources are Antoine Louis's own letters of 1792, transcribed in Hector Fleischmann's La Guillotine en 1793 (1908). Where sources disagree, the video and this page say so.
Archive pictures come from Wikimedia Commons; each links to its file page and licence.
The paintings and engravings marked "AI-generated illustration" were made with Google's Nano Banana and Veo, in the style of period history painting and the Encyclopédie. They show the world around the machine, never the machine itself; the guillotine you see is always our 3D model. Only public-domain art was used as a reference.
A reconstruction of the Paris machine of 1793 from Antoine Louis's 1792 specification, with gaps filled from Du Camp's 1860s measurements and replica research. Dimensions in metres are model values; speeds ignore friction; the square around it is illustrative, not a survey. Built with three.js (MIT); the light viewer is Google's model-viewer (Apache 2.0).
EB Garamond, IM Fell French Canon and IBM Plex Mono, under the SIL Open Font Licence, served from this site. Licence files.