Skip to content
Go to file

Latest commit


Git stats


Failed to load latest commit information.


This project aims to generate 3D models and scultpures from natural language; at this stage in the process (June 11th, 2015) we've limited ourselves to poetry but, in the future, the algorithm could input any English text and output a 3D sculpture.

This work began with Jordan Scott and Aaron Tucker in 2014 and was continued at Ryerson University in Toronto, Canada with support from Ryerson's Center for Digital Humanities (CDH) as well as Ryerson's Digital Media Experience Lab (DME), with specific help from Namir Ahmed and Tiffany Cheung. Further work was done at 2015 Digital Humanties Summer Institute (DHSI) in Victoria, Canada, helped along greatly by Shaun Macpherson, Nina Belojevic, Jentery Sayers and in particular Devon Elliot.

##Methodolgy of Translation

###Model 1

To begin with, we started fairly simple: we created a set of rules in which 32 characters (26 letters, 6 punctuation marks)mapped to 32 numbers (a=1, b=2) across the x, y and z axis. From there, each word was cut up into units of 3 and those three characters were then mapped to their corresponding coordinates on their respective axises.

For example the word "the" translates into "t" =20 on the x axis, "h" = 8 on the y axis and "e" = 5 on the z axis; these three coordinates are then combined to make a point in 3D space.

The first iternation of this project invloved Aaron modelling ("building from nothing") rather than sculpting (starting with a set object and subtracting). The included translation of "First Fig" by Edna St. Vincent Millay was made by first translating the poem into units of three, then mapping those coordinates in Rhino. We then connected each point on the X-Y axis, then connected the X-Y points with another line up to the Z axis. This process created the included "wireframe" versions that can be seen in their respective folders. He then moved these "frames" into Sketup and began connecting those points to create "faces" and closing it off into a 3D model, pictures of which are also included.

At the 2015 DHSI, the first prototype was created in response to a poem generated by Margaret Konkol (New College of Florida) & Andrew Pilsch (Texas A&M) which they called Stein in Six Dimensions which cut up and recombined Gertrude Stein's "A Box" from her collection Tender Buttons. A photo of the original text as generated by their software and an Arduino board can be found under the appropriate folder. Aaron then followed the same process as described above when translating "First Fig" and, after exported as an .obj file, was printed using a MakerBot: The Replicator.

###Sculpture 1

The second sculpture was generated using Processing with the extremely generous and inventive help of Devon Elliot. In this iteration, we used the basic "Geodesic" sketch from the HE_mesh library and input the basic x-y-z coordinates used for the first model. The sketch then "carved" a shape from the geodesic based on those coordinates. This sculpture was different than Model 1 in that it was computer generated (rather than Aaron going through and connecting the points) and points to a much more useful and interseting version of the project. Many many thanks to Devon Elliot!

##The Future

We are hoping to move this project away from the early "hand coding" that Aaron did in the first prototyping and create a more automated workflow using Processing. Using this and a combination of libraries included in Processing will allow for the poems to move closer to scultpures: ideally each physical "poem" will start as a 32 by 32 by 32 block and each 3D point in space (as generated by the XYZ axis mapping defined earlier) will "carve" away a part of that block. Once this process is streamlined, we are hoping to generate a website that will allow for user input that would then render any English text as a 3D sculpture.


A machine that translates poems (text) into 3-D models and sculptures



No releases published
You can’t perform that action at this time.