Terence Broad
@Terrybroad
Followers
320
Following
2K
Media
49
Statuses
282
Artist-researcher: https://t.co/0xdnD7Nb0S
London, England
Joined March 2025
Very happy to see this work from nearly 10 years ago getting this kind of recognition in 2025 🙏.
0
7
34
What was surprising however when all 1000 variations were played back at high speed the image of my likeness becomes present in the mind of whoever is viewing it .
Selfie experiment (2020). 1000 variations of myself, created using network bending.
0
0
1
Blazing-fast image creation – using just your voice. Try Grok Imagine.
284
557
3K
I dialed down the aggressiveness of the network bending so that the face was largely intact, but the features that generate facial features just far enough away that my likeness was (hopefully) not recognisable enough in any individual image
1
0
0
This time I used a selfie of myself and used StyleGAN inversion to find my doppleGANer
1
0
3
This work was created by using the same workflow that I had developed on the 0171 commission and haunted variations series:.
Using the respective latent for each band member, I disrupted and distorted the faces as they were being generated. Mostly to the middle layers of the network to preserve some of the likeness in the faces.
1
0
3
Selfie experiment (2020). 1000 variations of myself, created using network bending.
2
2
26
RT @halecar2: @verse_works It has been a delight to work with Rafaël! But what you guys just showed of @Terrybroad was amazing! Imagine a s….
0
1
0
RT @solos_gallery: Terence Broad.(un)stable equilibrium 1:1 #031. Sold for $405. There are no experiment 1:1 works listed for sale https://….
0
2
0
Which can be seen on all good streaming services:
open.spotify.com
Artist · 8K monthly listeners.
0
0
2
Though the band found these images compelling, it wasn't quite the right vibe for a synthpop/electro band. We ended up using the same method with different starting images to make these four EP artworks for the band.
1
0
4
A selection of these images were sequenced into GIFs to make the works Haunted variations 1 & 2, that were minted on tezos in 2021.
1
0
1
Using the respective latent for each band member, I disrupted and distorted the faces as they were being generated. Mostly to the middle layers of the network to preserve some of the likeness in the faces.
1
0
3
I cropped out each face and used the recently released Image2StyleGAN algorithm to find the latent vector that were the closest match of each of the band members in StyleGAN latent space:
1
0
1
In early 2020 I was asked to make a commission for a series of EP artworks for the band 0171, who gave me this press shot for the upcoming release as a starting point. I used this as an opportunity to play with the network bending library that I was developing at the time 🧵👇
1
3
23
RT @solos_gallery: “Colour is not a data source here, but rather a means of communication: a form of meditative dialogue between two neural….
0
3
0
RT @verse_works: Terence Broad.(un)stable equilibrium 1:5 #035.Presented by Luba Elliott & SOLOS. Owned by @ethspresso .
0
4
0
In 2021 I made a shorter variation of the same work and minted it on tezos:
0
1
7
In 2020, I took this second intervention and used in a latent space interpolation to make the 9 minute looping video disembodied gaze:
1
0
3
When researching network bending I developed an algorithm to analyse and cluster parts of the model together, such as finding the set of neurons that generates eyes. When turned off, the eyes are removed. The second image comes from when everything is turned off BUT the eyes.
2
8
45
RT @monkantony_tez: added 2 more 2016–17 AI art show artists to the list . 🦾 @artcenternabi's Why future still needs us: AI and humanity (N….
0
1
0