"This is a whispering dialogue between mysterious nature and human beings.
Come and wake them up in the night."
produced by: Sarah Song
"WHISPER" is a projection piece using voice input as its source of interaction. Particular words or phrases will be recognized and then trigger specific visuals which will then be projected onto the surfaces (cap) of mushrooms.
The reason why I choose mushrooms is because their spore print reminds me the pattern of iris in our eyes. Every mushroom has unique structure of gills which make every spore print unique just as the iris of human beings. "A forest is much more than what you see." says ecologist Suzanne Simard. And I believe fungal networks can not only allow plants to communicate, but also using its secret way to communicate with us.
The project of Friedrich van Schoor and Tarek Mawad in 3 years ago, Projections in the Forest, gave me great inspiration. I really love the concept of bioluminescence. The emission of light from a living organism is like they are sending message to the world. For this reason, I decided to use light projected on the mushroom (hopefully) in response to human voice, for example, the mushroom will “blink” when I say, “wake up” (refer to the video demo).
I used the ofxVoiceController addon written in the lab of week 19 for voice analysis. It works together with ofxMaxim and ofxRapidLib. It is relatively easy for me to collect training data and fast to train. In my piece, there are 5 classes in total, the first class is the recoding of the ambient sound (noise). Correspondingly, there are 5 visuals including cells, spore, gills, luminescent spot and an eye open video. To achieve the bioluminescent effect, I used mostly cold color in the generated images and a blueish eye video as well. Particle system were used to represent spores of mushroom and gills were represented by lines.
Inevitably, I encountered some difficulties:
Setting: Projection distance need to be set in 0.5-1.5 meters range in order to project clear visuals onto small surfaces (mushroom caps). It is impossible for me to carry all my stuff and equipment to a mushroom hunting right now, so what I finally did is I placed a bunch of mushrooms in the backyard for my projection.
Projecting on real organism: Mushroom caps are circular but vary in shape, it is difficult to project a smooth and precise boundary on them. I had tried to use ofxPimapper to create a multi grid surfaces so I can drag the anchor point for mapping but I couldn't use ofxVoiceController addon at the same time (did not solve the puzzle). Also, I failed to use OpenGL for texture mapping on circles. My final decision was to use ofFBO object and allocated them to different location instead. The result is a bit satisfied but not impressive to be honest.
Voice recognition and interactive machine learning methodologies are really complex. At the moment, this piece is just like a prototype. It still needs to be optimized and there are lots of possibilities to be expanded in the future.
Moving forward, I would like to first solve the texture mapping problem which is the most challenging aspect of my program. Furthermore, to improve the voice recognition model, I need to add more training samples and test some better features for more precise results.
Ideally this piece is to project on living mushrooms in the nature, but when it is put into an exhibition context, I would like to place the mushrooms in a glass terrarium and project in a dark space.
Special thanks to:
Addon & References
Some code was hacked from class exercises:
ofxVoiceController addon - From week 19 (Machine Learning) of Workshops in Creative Coding Term 2
BallSystem example (Voronoi) - From week 11 (Complexity) of Workshops in Creative Coding Term 2 by Theo Papatheodorou
Projections in the Forest by Friedrich van Schoor and Tarek Mawad