This demo builds on the Word-As-Image for Semantic Typography work to support Arabic fonts and morphing whole words into a visual representation of a semantic concept. It is part of an ongoing project with the ARBML community to build open-source Arabic tools using machine learning.
The GIF you see above are 7 outputs of the program when given the Arabic name of an animal in text and asked to morph it into a visual representation of the actual animal. The input of course can be any text and the output any concept. I am planning on adding support to different Arabic fonts in the future and support for other languages as well.