How can you be anybody in Zoomspace? Very recent developments in deep-learning, allow creating synthetic media with unprecedented quality and ease. This includes doing facial reenactment in real-time, provided with only a single image of the desired avatar. In this hands-on participatory tutorial, we will use accessible on-line colab notebook to create deep-fake videos and live avatars using your own materials. No prior knowledge needed. I will also introduce and demo the first available purely online solution for live real-time avatars from the webcam in the browser.
I am maintaining an open source repository at github.com/eyaler/avatars4all and I am giving talks, workshops and tutorials on these techniques. I am actively mentoring and collaborating with artists, journalist, educators and activist to allow them to access the technology.
How will you deal with varying numbers of participants in your session?:I have given over 150 talks and tutorial in the last years, offline and online from 0 to 100's of people.
What is the goal and/or outcome of your session?:takeaways:
1. Understanding the contemporary technological possibilities for media synthesis and manipulation, namely deep learning based deepfakes and puppeteering
2. Getting hands-on familiarity with accessible on-line tools, allowing to create head or body deep-fakes and live avatars without need of prior knowhow.
3. Discussing the moral, social and ethical implications and prospects, considering fake-news, privacy, self-image as well as creative and activist usage
A new-media artist and machine learning researcher. I hold a PhD in physics. I create interactive installations, computer generated art, algorithmic poetry, online interventions. I teach in the community.