Utah Teapot: Artifact, History, Datasets
The Utah teapot is an infamous test object in computer graphics.
The teapot is honored by the University of Utah Teapot Rendering Competition (TRC), the final project of the CS 6620 (raytracing) course and open to all who have taken the course previously[1].
A lot has been written about the legacy of the teapot and the people involved, but I want to pin down the early history and data.
The Virtual Teapot

Figure 1
: Some of Martin Newell's work, on graph paper, modeling the teapot (image source).Martin Newell, then a graduate student under Ivan Sutherland at the University of Utah, wanted to show the usefulness of Bézier patches in 3D modeling[2]. While at tea with his wife, Sandra Newell, she suggested Martin model their teapot, which is indeed a perfect application. In addition to being a curved surface, the teapot also is not topologically trivial, can cast self-shadows, does not require a texture to look good (although one can also easily be added), and contains positive, negative, and saddle curvature. Martin modeled the teapot by hand on a Tektronix storage tube. The earliest publication record I could find was with the entire teaset in his 1975 PhD thesis (see Figure 28 (pg. 84) and Figure 29 (pg. 86)).
Martin had an interactive display, and was showing someone, possibly from some funding agency, that the Bézier patches could be scaled interactively. Jim Blinn found that a shorter teapot was more aesthetically pleasing. (Note that there is a myth that the shorter teapot is the result of Blinn compensating for the non-square pixels of an Evans & Sutherland framebuffer, but Blinn has specifically debunked this[3][4][5].)
"Everybody, including me, liked the teapot squished" - Martin Newell
Versions of the teapot apparently circulated within a "close-knit DARPA-Net computer graphics research community"[6]. It first showed up on the mainstream research scene, evidently in the squished form, in Texture and Reflection in Computer Generated Images [Blinn and Newell, SIGGRAPH 1976], with the earliest paper accessible online being the version of this paper published in [Blinn and Newell, Commun. ACM 1976].
Teapot Datasets
The teapot was modeled as a collection of cubic Bézier patches. Bézier patches are good for curved surfaces, but are not commonly supported by software today. Note that, because they are fundamentally curved, there can be no exact tessellation of the teapot into triangles.
There are many datasets of various ancestry. The earliest I know of was published in The Origins of the Teapot [CG&A, Crow 1987], alongside Pascal display code.
There is also a page at Holmes3D.net. This has several files. There are digitized versions of the CG&A article data and un-scaled versions of same. There is also another dataset with different units of unstated origin, and the teaspoon and teacup "from the web".
There is also Steve Baker's wiki, which has an embedded dataset of unknown origin but language suggesting he thinks it's original. Also notable is the claimed scaling of 1.3.
The archival site funet.fi/pub offers two versions of the teaset.
There are of course many other datasets that have been published, but seem to me to be derivatives of these. Even these are doubtless derived from each other somehow; I am in the process of untangling this.
The Physical Teapot
The Newells' teapot was purchased from a ZCMI department store and manufactured by the Melitta company[7]. The Newells donated their teapot to the Computer History Museum in 1984.
The Newells' teapot appears to be a c. 1974 manufacture of a (possibly revised) 1961 design, itself a revision of a 1959 design, and is due to Lieselotte "Lilo" Kantner, or perhaps Jupp Ernst[8][9].
According to Jever Castle Museum[10], the teapot was available in the 'Minden' line of products. The line originated in 1954 as earthenware, added an option for porcelain in 1956, added the vital teapot in 1959, introduced a 2nd-generation teapot in 1961, and discontinued the teapot in 1963. The teapots seem to come in two sizes (the Newells' was the larger). This information is corroborated by Museum-Digital's entry for the smaller 2nd-generation teapot, though is mildly contradicted by Friesland itself[11].
However, this is not the complete story for dates, and in particular the teapot was produced again later, and at some point the design was tweaked to result in subsequent generations[12]. The Jever Castle and Museum-Digital links above both vaguely suggest the Newells' teapot was 2nd-generation. However, the Computer History Museum lists the Newells' teapot as being made c. 1974, which does not imply but does seem to indicate a later generation[13][14].
Jupp Ernst is credited with most of the Minden line. However, the late Lieselotte Kantner is often cited for the teapot. Kantner was long-time design head of the relevant section of Melitta, and appears to have joined Melitta in 1959. The timing does make it possible.
The relevant Melitta section was given its modern name 'Friesland Porzellan' in 1982, and the subsidiary became independent in 1995. By the time the teapot became iconic in computer graphics, it took some time to connect the Melitta company to Friesland Porzellan. In turn, Friesland Porzellan didn't know they happened to still be making an icon. When the reconnection was made, they donated a teapot to the Oldenburg Computer Museum and renamed the product from 'Haushaltsteekanne' ('household teapot') to 'Utah-Teekanne' ('Utah Teapot').
Until quite recently, you could still actually order one of these, and this is where my Utah teapot comes from[15]. Unfortunately, the company suffered a disastrous fire in June 2023 and are still rebuilding. Regrettably, they have no current mention of the teapot on their website.
It is not clear if the physical Utah teapot will, or can, be made again. We all emphatically hope for a return of the teapot, and wish the best of luck to Friesland Porzellan in rebuilding!
Notes
Thanks to Kyungjoon Lee for suggesting some historical sources.