Overview / Usage

Pictures of Jap Girls in Synthesis, is a machine that creates live, visual, poetry. The machine reads, or listens to poetry in a variety of languages. It synthesizes in real-time, visual images illustrating its understanding and interpretation of the text. This is done using state-of-the-art deep learning algorithms for speech-to-text, language translation, and the AttnGAN (arxiv.org/abs/1711.10485) for text-to-image synthesis. While the results are far from photorealistic, we get images of surrealistic quality, containing colors, shapes, textures, patterns and objects related to the words and their context, and creating a poetic experience. This project was created in a 30-hour hackathon, at Geekcon-Art Hamidrasha 2018.

Note:
The content of this page has been recently vandalized in the following ways:

1. The URL has been changed breaking previously published links
2. The project name has been changed
3. The title image has been changed
4. The overview text has been changed
5. Images have been deleted
6. The video URL has been deleted
   The author has not conceded to these changes nor has he been notified.
   These changes are in violation of the author's written agreement with Intel's AI developer community manager.
   Some of these changes may also constitute an unlawful manipulation of copyrighted artwork.
   This page is now automatically monitored for unauthorized changes.

Methodology / Approach

Using a lot of python glue and coffee, we hacked together overnight a novel speech-to-image demo, using off-the-shelf technologies.

Technologies Used

Google Speech-to-Text API
Google Translate API
AttnGAN

Repository

https://www.youtube.com/watch?v=4zdo-bPXszY&t=1s