MISMATCHING
NATURE





Deep Learning
2023

Deep Learning | Animation | Speculative Design

mentored by Ludwig Zeller
collaborated with
Akshada Bhat, Paulina Ojeda

MISMATCHING NATURE
an AI-driven visual journey to highlight the urgency of the climate crisis
Mis-Matching is a visual AI exploration that focuses on raising awareness of climate change by focusing on ecological mismatching. In this phenomenon, species fall out of sync due to changing environmental conditions. We created an AI-driven animation featuring five plants, initially growing in harmony but later misaligning to illustrate this disruption. The visuals, generated with DALL-E and Stable Diffusion, are paired with a soundtrack generated with the sound AI tool AIVA that intensifies as the mismatch occurs. A voiceover, produced with ChatGPT and ElevenLabs, explains the concept, blending scientific insight with emotional appeal.


Tools 
Stable Diffusion, ChatGPT, Kaiber, AIVA, Eleven Labs, After Effects


PROCESS
1. explorations in Dall-E
First the visual language of the project was explored in Dall-E.



2. prompting in Stable Diffusion
Then we extracted prompts from Dall-E and adjusted them for prompting in Stable Diffusion. At this point Stable Diffusion offered more control in the generation of images and the possibility to process images in batches. Therefore, we first created basic animations of the plants in After Effects based on the plants created in Dall-E. Subsequently, we extracted frames from these animations and used image to image generation in Stable Diffusion to return to the initial visual language explored in Dall-E.



3. animation
Finally, five plants were animated from the images created in Stable Diffusion and an AI generated voice over and soundtrack added to explain the phenomenon of ecological mismatching and support a dystopian effect in the final video.