By Pili Valdivia


“Lo que no hablas — Metamorfosis” is a sound reactive AI generative experiment done during the Audiovisual Workshop at Aalto in autumn 2023.

The video combines the sound-reactive visualisation of the song “Lo qua no hablas” by Rubio with StableDiffusion frame by frame image generation, to illustrate the journey of a caterpillar in it’s quest of mutation, and to acknowledge the grace and struggles of the process of growth.

The project was created within TouchDesigner, and uses the song’s audio spectrum to live generate reactive geometries that are used on each frame of the video as image-to-image inputs into StableDiffusion. These input images are combined with a text prompt resulting in AI generated output images that simultaneously enter a feedback loop to give continuity to the sequence.

All inputs can be edited and modified live during the recording process to influence and actively manipulate the final result, combining the automation of TouchDesigner and unpredictability of StableDiffusion, with the human live decision-making and emotions.

Resources and References


  • “Lo qua no hablas” by Rubio