I did some brief experiments with Runway ML last year for the pink performance and was a bit underwhelmed - the results were weird rather than compelling. I see a bunch of AI video on Instagram which is far more compelling and coherent. However, inspired by that, and the playful urge to make something a bit creepy and disturbing to include in the five minute video, I cracked it out again...
This time I decided to try turning a piece of my static art into a video. I'm currently running a traditional render of my fabric mask bursting into flames, and it's stupidly slow, so the idea of generating video in a few clicks is compelling!
I chose "Pain and Acceptance", and asked Runway ML to take the image and generate a video where the two protagonists look at each other emptily. I'm not bothering to show them all, but the first attempt featured a significant camera movement, which was not what I wanted, but also caused Runway ML to significantly distort the two figures and replace the writing on the wall with asemic writing. I discovered a setting to stop the camera moving, plus specifying in the text prompt that it was a static camera. The result was the video below:
Overall I'm happy with it - mostly as a toy - there's plenty of things I'd love to change, but for now my curiosity is satisfied! Also I'm not sure how to change them!
No comments:
Post a Comment