DAY 3 SUMMARY

 

Ben Durack got the group started on Midjourney and demonstrated how it could be used within a design process.

 

We rounded off the week with a workshop exploring AI generated music led by Graduate In Residence Saoirse Horne.

Workshop 4 -  Text to Image within a Design Process


Ben Durack, lecturer in 3D Design, demonstrated how to use text-to-image AIs, such as Midjourney, and Viscom, as part of a design process.

 

The activity involved creating photography-like images of objects with Midjourney. 

 

Time was spent getting familiar with text prompts and how they work.

 

Example prompt: "a lounge chair in a Scandinavian minimalist style making use of premium natural materials, in a living room setting. Hyper realistic, 4K redshift, high contrast"

 

Example prompt: "a Knolling photograph of a collection of jagged materials to give the feeling of a Scottish forrest on a dark green background with rock forms. Minimal. --ar 4:5"

 

(Knolling, or flat lay, photography involves arranging objects on a flat surface.)

 

The images in the slideshow to the right show how images evolved over the course of the workshop through prompts and selection for re-iteration.

Workshop 5 -   Let's play with Open.AI's Jukebox


Gray's School of Art Graduate-in-Residence Saoirse Horne devised and led a group activity using the Jukebox sample library.  


The session began with a group listening excercise in which participants rated the sample's likeness to 'human music'.

 

Saoirse then shared some examples of experimental computer and tape loop music, and discuss the role of the musician in bringing 'heart' and 'feeling' to compositions. You can see the discussions which arose from this mapped in the diagram below/right.

The workshop concluded with participants co-creating some music together using samples downloaded from Jukebox. 

 

The group named the song "AI WHISPERS (I wanna be tall)". This was in reference to the above discussion and the term "angel whispers" - which refers to the strange noises heard when a radio isn't tuned properly. The section in brackets refers to the sounds the AI "singer" makes during the song, which sound a bit like the phrase "I wanna be tall".

 

CONTRIBUTORS

Helen Scarlett O'Neill

Jim Hamlyn

Catherine M. Weir

 

SAOIRSE HORNE

Links & References

 

Jukebox is available online in collaboration with google at this link: 
https://colab.research.google.com/github/openai/jukebox/blob/master/jukebox/Interacting_with_Jukebox.ipynb

 

You can also download the software and look through the library at this link: https://openai.com/research/jukebox

 

(Please be aware you may require add-on's on google to increase CPU and Python for your computer to use the software.)

 

Articles of interest: 
https://pitchfork.com/features/overtones/google-text-to-music/

https://www.newyorker.com/magazine/2014/11/10/looped