Autobotography Intent – IanD

Autobotography Intent – Talking to Myself

Abstract

I want to create a digital version of myself that I can talk to. I will use the Lyrebird API to replicate my voice, hook it up to a chatbot that is trained on my digital persona (dialogue on social media and chat applications), and finally rig it up to a hyperrealistic model of myself created with photogrammetry.

Synopsis

I’ve seen various technologies emerge in the last couple years that can, with a great amount of effort, completely represent a person’s likeness in a virtual environment. The Lyrebird API can listen to only a minute of human speech and artificially synthesize it to say virtually anything in the source’s voice. It’s not perfect, but it captures much of the intonation and nuance of human speech. This would be the first goal of my project: to create a digital version of my voice that can be used by anyone.

The next step would be to create a chatbot that responds like my “digital self.” I would harness conversation data stored on social media in comments, posts, and chats, organize it, and then train a neural network on that data. To be honest, this will probably take the longest and may not be possible given the time restraint. Data collection will be  It may be worth using a pre-existing chatbot

Finally, I would create a physical representation of myself in the digital world. To do so, I would have photos taken of myself radially around me from my chest-up. After about 60 or so images, I would feed the photographs into Autodesk Recap Pro, a software that can construct 3D models from a rotational series of 2D images. It utilizes structure from motion technology to generate a point cloud that constructs the 3D model. Recap Pro also maps a hyperrealistic texture to the model.

Once these steps are complete, I can integrate them all in Unity.

The purpose of this project is not only to “flex the tech” and show off what a crazy college student can do with free software these days, but to create jarring representation of “self” and what exactly defines our personalities. It serves, especially in light of recent Facebook findings, as a reminder of how much of ourselves we imbue in online personas on a daily basis.

Budget

Software: Lyrebird, ReCap Pro Student License, and Unity – Free

Time:

LyreBird voice creation – 1 hour

Creating a chatbot – 40 hours

Photogrammetry to Model – 25 hours

Head rigging in Unity – 10 hours

Coding in Unity – 20 hours