Part 9 – Morphing Figures

Generated Latent Space Walk videos from Machine Learning Model based on images of me practicing yoga. These videos show the morphing of one generated ‘pose’ to another resembling transitions within yoga as well as the subtle trembling of the body when holding a pose or breathing.


Tuesday 13th April

During my tutorial with Jen yesterday she mentioned a feature in Runway ML that I didn’t know about and it lets you add a green screen to videos. I decided to go back to my trained machine learning model in Runway based go images of me practicing yoga and I exported another short Latent Space Walk 15seconds. I then added the greenscreen by selecting the parts of the images I wanted to keep and exported the videos.

After exporting the green screen outcome I opened it in Premier Pro and used the Video Effects > Keying > Ultra Key to remove the green background and added a white background instead.

Editing the green screen outcomes in Premier Pro.
Using Video Effects > Keying > Ultra Key to delete the background.

I imagined that when I show this work I want to play it in a loop and I didn’t want it to jump to the start again as I enjoy how it smoothly transforms from one form to the next just like the movement in yoga practice so I copied the clip and reversed it so that it would smoothly transfer back to back.

I made a quick mock-up of a possible display for the work. I would like to go through this process a few more times to get different latent walks exported so I can maybe display 4 different videos together. I would also like to keep the works at a small scale if I was showing these I would maybe consider playing the videos on small monitors and covering them with a thick perspex or glass and experimenting with different ways of displaying them. I defiantly want to keep tot the 1:1 ratio format for these particular works.

Since the mock-up above was made using one video I have been using Runway to export 3 more videos hopefully with 4 varying outcomes in total running at 36 seconds staged together. The reason for doing 4 outcomes is to explore the generated latent space to show variety and difference between the shapes and forms .

The final 4 latent space walk videos.

These videos show the morphing of one generated ‘pose’ to another resembling transitions within yoga as well as the subtle trembling of the body when holding a pose or breathing.

The machine and software, Runway ML, I used had no knowledge of yoga and the anatomy of the body which is why these results are fascinating. By only training the model on 777 images of me while I practiced yoga, these latent space walk videos resemble movement and transitions which are very similar to yoga almost morphing from one surreal pose to another. To me this is an illusion of a machine trying to understand human mindfulness and movement.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s