Nhõng Nhẽo
This is a game that teaches the meaning of a Vietnamese word through forcing the player to role play a cat in order to hopefully embody the behavioral meaning, which can catalyze an experience of understanding the word internally.
produced by: Tammu Do
Introduction
The Vietnamese word is one of those foreign words that is difficult to translate into english. This behavior includes body language and sounds, and most of the time a combintion of both. The initial game concept was ambitious in that I wanted to use a skeleton tracking, mfcc audio input, and machine learning. One condition of the project was to focus on interaction though so this concept was fine for holding a personal promise of goals in the future, however given this condition and time, I decided to scope down to one prominent characteristic of the behavioral word, the voice. Nhong Nheo can be recognized by the tone of voice, specifically how the verbal message is carried out. This tone is most recognizable when the verbal message peaks at the end of the sentence.
Concept and background research
This game idea grew from participating in a game design work shop with game designer, Federico Fasce. He asked his participants to create a game about a word that is difficult to translate into english, whose meaning could not be justified by the english vocabulary. There’s a lot of dark and sorrowful words I could think of in Vietnamese that would fit this mould, but because, much of my work has been serious in the last couple months, I wanted to go towards something playful. Also, much of my practice as an artist and designer involves alot of investigations and critical analysis on my identity, place in the world, and my community. Usually, there is so much interest and attention on east asian cultures, and too often stereotypical views or ‘knowledge’ of south east asian cultures. Speaking on that, I wanted to be careful on how I decided the visual language and tone of the project. There are many asian tropes of the submissive and whiny asian women, and sometimes when I have explained this word before to others, some folks have already racially gendered the word to asian females, when this is not true at all in Vietnamese culture. I have seen people of all genders performing this role.
Much of my initial research into cultural games began at looking sources in China, Korea and Japan, specifically for my inital concept of using the body to embody the behavior. Korea’s games particularly interested me since they have a similar word called “Aegyo” (Aegyo, 2019), which describes a a cute and/or childish behavior. I would argue that they are not the same, because somehow the Vietnamese counterpart seems to also describe a texture or even, timbre of the ‘cute’ sound that is not exclusively attached to ‘childish’ but in many interesting ways, like the raspy or depth quality of a cat meow.
After some analysis of these differing qualities, I was able to strategize what kind of embodiment role I wanted to place on offer for the user: a cute cat. My presumption as not only someone who spits on the gender binary, but also someone who has seen men, women, and folks in between speak back to cute cats gave me this: regardless of gender, everyone can meow like a cute cat if they wanted to. Usually gender associations with being sweet to an animatl are less structured as to compared to other structural societal systems. There’s so much reserach in how animals can provide a safe space for male prisoners to rebuild their humanity, and show a gentleness that is usually not afforded with the pressures of hypermasculinity (Deaton, 2005). There are many programs in which inmates are given the opportunity to learn how to train service dogs (Deaton, 2005). To be clear though, these measures are not all pure good intentions. Prisons, training, and labor has a whole slew of problems part of the Prison Industrial complex that serves modern day slavery, but that’s another conversation. The reason why I am drawn towards this idea is because, I’ve seen even the ‘toughest’ men in my life, like my brother melt like a cheeseball petting our cat, Yo-Yo, and to be frank, I need to see this gentle like behavior more. So many women need to see and experience this more from the men in their lives. So what started out as a cultural idea became also a very personal one in which I wanted to make a game that encourages not only ‘cute’ behavior but sweet behaviors in PUBLIC. Obviously, this doesn’t mean that this game will be the saveall tool that will make your brother turn off his macho performativity. But I feel that having games that provide at least a moment for you to be human, could at least start a path for me to research and design playful moments that could dispel and destabilize gender.
Keeping the end in mind, I had to also negotiated with myself what kind of visual language would serve my goals. I didn’t want to avoid ‘warm’, or usually feminine associated colors out of the fear that the word would again be wronly associated with only a specific gender, but at the same time, I understand this very much happens frequently. So how do I serve my own interests in saying that colors are literally for everyone, without downplaying my choice color palette that I personally find cute and fullfilling. And if I go for the usually associated gender neutral colors of greys and blacks, the warmth of the game would not be captured, not to mention the problematics of gender netural visual languages somehow always leans towards ‘boy’ clothes. I feel this will be an ongoing design challenge that will always walk alongside gender. For now, I decided my warm colors would serve my purposes of indicating to the player how well they are emboding the behavior or rather how cute they are behaving. I chose a color palette whose saturations would aid in the gamified visual language. The Coolors App helped me randomize a certain hue range (Coolors, 2019).
Some of the body capture game installations I saw online embraced their different typographic, graphic design composition, and color modern traditions. In the case of one game created by, Bashi-Ta, they embraced strong graphic compositions with contrascting colors that a lot of Japanese graphic design started and (actually influenced western design practices) (Bashi-ta, 2018).
I decided my graphic art style would serve my interests for now. And researching the Vietnamese modern graphic design scene would be for another day. I went on to draft my vector art before programming.
Youtube cat videos were my friend, since I haven’t spent a good amount of time with a cat friend in a while. The depth and intonation of these meows are remarkably different from one another, and that was a cue to me that I should also consider the different vocal ranges we have. Sure, most of us can meow, but the ranges are different, even the high pitched cute meows. I had to remind myself this later again during the process of aquiring the final training data.
Tools: Rebecca Fiebrink’s openframeworks VariousAudioInputs app to extract audio features as inputs, Wekinator, Wekinator Input helper, and a Nhong Nheo cute app I created as the output sourcem (Fiebrink, 2019). Source link in bibliography.
Technical
The graphics and animations I decided on were based on sweet Vietnamese desserts, desserts that have a variety of textures, tarty,citrus, sour, and other different flavors. Textures particularly for this game was important to me, and I wanted to give players that understanding of how ‘cute’ can have a distinct texture. The way I designed the art was structured like this: the cuter your meow is, in this case the cutest meow was loud and higher in pitch, the sweeter and more textured sweets you would get. In other words, if your meow is loud, higher in pitch, and somehow offered different textures or depth in your m-e-o-o-w-w, then you’d get a bowl of mochi balls in ginger syrup, more flavor and complexity. If your meow was flat and lacked intonation, your reward would be just a pet on the head. I also referenced two existing heart processing sketches that I wanted to use with my game, since hearts and cuteness are easily associated (referenced in bibliography with links).
Given the time, I was only able to craft three levels of cuteness (excluding the default brown screen which is meant to only classify ambient noise), and even though meows are all different, my players are human and from my user resesarch, many have different versions of what a cute meow means to them, which means a longer process of accomodating for that difference. Sometimes in these situations, the designer has to give a general template, enforce the rules, and pray the players play along and try to guess and aim for the cutest score, which is the whole point of playing. At the end, I hope most people were able to grasp a better understanding of the word and behavior.
My first functioning workflow involved using only three audio inputs from Rebecca Fiebrink’s VariousAudioInputs app, since I have had some luck with simple color change experiments with
processing. Since, I decided that training accuracy was not a priority for me, I also wanted to see how little inputs I could get away with to still create a strain free, playful experience. I started out testing with KNN model, which has been friendly to me in the past with more simpler audio experiments. I found the noise was too much especially in a space similar to a gallery/exhibition. I switched over to Decision trees, Naive Bayes, and discovered I needed more training data in order to get a better gameplay experience that was not too difficult. However, the gem solution after much testing and feature configurations, using the Wekinator Input helper, MFCCs and RMS audio inputs served me the best in classifying 'cute' meows while still allowing for reasonable tolerance of vocal range. At the end, gathering more training examples from people in the higher and lower ranges served me well since most of my training data involved my voice range, which I believe is some where in the middle-high.
Future development
My playtest with cohorts was rewarding in that I found more people were willing to meow than I expected, and they were eager to get the cutest score. I did have classification issues in my early play tests due to having not enough of a range of data, but that was quickly resolved with bringing in a few people who were in the baritone area of ranges. Regarding the cute animations, computational art, I was hoping to grab the raw data through osc from the feature extractor program from Rebecca Fiebrink’s Various Audio OF app in order to create audio wave animations to further give users the feel and understanding of their own voice and how they are scoring. I had issues trying to even get the source code working, so this will be on the books for future development. In general, I want to give more real-time animation feed back particularly representing the input soundwaves. Also, more cute graphics or emojis wouldnt hurt, at least as one ideation. I also hope to build on this theme and this word, Nhỏng nhẽo as the foundation of a new series of work centered on this behavior. Given more time, I would like to experiment more with different words. One ideation I am excited about is to develop this game with more words in which the longer a player engages with the game, the words change, and they slowly start saying more political words and perhaps sentences affirming their gentleness and kindness.
Another idea that came to fruition from researching how humans respond to cute and other postive stimuli, I wanted to explore ‘cute aggression’ and particularly dimorphous expressions of emotion. Dimorphous expression is described in a study focused on this very topic, as “both positive and negative expressions occur simultaneously in a disorganized manner, which leaves witnesses to rely on the context of the situation to interpret them (Aragón, 2015).“ Exploring how to make games that open up this space of embracing complex emotions is fascinating to me. One real example, I thought of right away: my best friend always punches me in the shoulder when she laughs at my jokes and thinks I’m funny. I asked her why she does that, and she said she thinks I’m cute and doesn’t know where ot put her ‘emotion.’ Remembering this small, harmless example but a real one sparked a lot of curiousity in my new endeavor of making games that subtley encourage more postive emotions and behaviors.
Self evaluation
Building an intuitiveness with wekinator and its input helper can be easy with love and time, but I have to keep reminding myself that me, myself, and I will only be representative in the data, if I don’t go out during early testing periods to train on at least two or three people. As per usual, as Phoenix Perry says often: “Test often, test early.” And now I must engrain inside me: “Train early, train often. (within reason).” Also towards the last stride of this iteration of the project, I realized I had to add more training data than I assumed was necessary.
References
Aragón, O. R., Clark, M. S., Dyer, R. L., & Bargh, J. A. (2015). Dimorphous Expressions of Positive Emotion: Displays of Both Care and Aggression in Response to Cute Stimuli. Psychological Science, 26(3), 259–273. https://doi.org/10.1177/0956797614561044
Bashi-ta. (2018, September 17). ゴロゴロくん / 寝転んで遊ぶインタラクティブコンテンツ. Retrieved April 15, 2019, from https://www.youtube.com/watch?v=wxlTeBf4d_c&fbclid=IwAR1JEz0Vcac9vXIvpT3elIEH1VoD8tVaEa9HWeWqn-Q9Nuk99gnF_97xRM8
Coolors. (n.d.). Retrieved April 20, 2019, from https://coolors.co/d6f6dd-dac4f7-f4989c-ebd2b4-acecf7
Deaton, C. (2005). Humanizing Prisons with Animals: A Closer Look at “Cell Dogs” and Horse Programs in Correctional Institutions. Journal of Correctional Education, 56(1), 46-62. Retrieved from http://www.jstor.org/stable/23282783
Korean Aegyo: The Seven Levels. (2019, February 03). Retrieved April 15, 2019, from https://www.90daykorean.com/korean-aegyo/
Romano, Andrea, De Polavieja, Gonzalo G., Rubolini, Diego, Caprioli, Manuela, Boncoraglio, Giuseppe, Ambrosini, Roberto, & Saino, Nicola. (2011). Sex-Related Effects of an Immune Challenge on Growth and Begging Behavior of Barn Swallow Nestlings. PLoS ONE, 6(7), E22805.
Various Audio features extractor, and simple colour change reference example source:
Fiebrink, R. (n.d.). Retrieved April, 2019, from http://www.wekinator.org/examples/
Wekinator and Wek Input helper:
Fiebrink, R. (n.d.). Retrieved from http://www.wekinator.org/downloads/
Third party Heart Graphics and animations:
simple Heart processing: https://www.openprocessing.org/sketch/660715
Heart wings aimation: https://www.openprocessing.org/sketch/377004