/QIAN CHANNING MU/ Work Life About
Project
The Flower Vase Girl
Tools
Arduino
p5.js
MadMapper
Runway
MidJourney
Physical Fabrication
Video Production
Description
The Flower Vase Girl is an interactive sculpture reimagining the eerie legacy of the “flower vase girls,” a grotesque cultural spectacle from late 20th-century China. Using projection mapping, Gen AI, and sensor-based interaction, the project dynamically transforms the performer’s visage, confronting viewers with themes of objectification, suffering, and societal voyeurism.
Overview
The phenomenon of “flower vase girls” embodies a chilling intersection of commodified suffering, gender oppression, and societal voyeurism. These performances, which falsely depicted women contorted into porcelain vases, symbolize both cultural nostalgia and the exploitation of female bodies. Though later revealed as illusions using mirrors and props, their unsettling implications persist as potent cultural and feminist critiques.
Using projection mapping, generative AI, and sensor-based interaction, the project reconstructs this eerie juxtaposition of human fragility and objectification. It dissects layered fears—biological, societal, and gendered—revealing humanity’s troubling tendency to exploit suffering for entertainment.
Through three staged interactions, the installation confronts viewers with their complicity. The “flower vase girl’s” face transforms from silent sorrow to seductive allure, culminating in grotesque distortion—a stark metaphor for the commodification of women under the voyeuristic gaze.
By merging technologies such as Arduino, p5.js, and Runway with the haunting aesthetics of East Asian folklore, "The Flower Vase Girl" offers a multi-layered critique. It compels audiences to reflect on how historical oppressions echo into modernity and to question their own gaze and responsibility.
At its heart, the work asks: can empathy and awareness disrupt cycles of exploitation, or does voyeurism inherently perpetuate systems of objectification?
Backgournd
The "Flower Vase Girl" was a form of street performance popular in late 20th-century China, characterized by the illusion of women appearing to be confined within porcelain vases. Only their heads were visible, leading to widespread myths that these women had been physically altered to fit into the vessels, with some narratives suggesting they had been placed in vases as infants.
The illusion was created using mirrors and carefully arranged fabric to conceal the performers’ bodies, giving the appearance of a torso and limbs fused with the vase. Audiences paid fees to view these performances, which often included singing or interaction.
·
Construction and Design
This installation features a simulated “flower vase girl” created through a combination of physical and digital components:
- Model Composition: The structure consists of a plastic tube frame, a foam head, an aged porcelain vase, and an AI-generated face.
- Projection Mapping: A short-throw projector, concealed at the front of the vase, ensures uninterrupted visuals and avoids viewer interference.
- Interactive Sensors: Distance sensors powered by Arduino and p5.js adjust the projected facial expressions based on the viewer’s proximity, seamlessly integrating audience interaction. The sensors are discreetly embedded at the base of the installation.
- Physical Components: The foam head is enclosed within a vase-inspired structure supported by a semi-open plastic tube framework. Inspired by rural Chinese aesthetics, the setup features high-contrast fabrics and colorful tassels for additional visual context.
- Digital Design: Facial animations generated through MidJourney and Runway are mapped onto the foam head, creating a fragile, porcelain-like visage that amplifies the unsettling uncanny valley effect.
Coding
The sculpture uses sensors to detect the audience's distance and is connected to Arduino and p5.js to switch the projection content based on the changes in distance. At 3 meters, 2 meters, and 1 meter, the audience will see different transformations of the Flower Vase Girl.
This is a brutal, absurd unraveling of pain, where suffering is consumed, deconstructed, and re-packaged as “enjoyment.”
Main Design
The head of the Flower Vase Girl is a plastic head model I bought online, close to human proportions. I modified the neck by cutting it so it would fit inside the vase.
Frame Design
The frame for the Flower Vase Girl's display cabinet is made of plastic pipes. I then went to a fabric market and bought three 3x6 meter red velvet fabrics to wrap around the frame.
Face Design
I wanted to design a female face from late 20th-century China that exists somewhere between dream and reality—one that features a ceramic-like quality with uncanny valley effects, but still retains a sense of realism. To achieve this, I ran over a thousand iterations with the MidJourney model, tweaking different prompts and experimenting with various inputs and outputs. After countless attempts, I finally arrived at a face that I was relatively satisfied with. It still carries a slightly surreal feel, but perfectly captures that unsettling, almost-real quality I was aiming for.
Facial Expression and Content Refinement
Since I needed to eventually import the work into Runway for dynamic processing, it was crucial that the generated some facial features and background remained consistent across frames. However, the #remix and #edit functions in MidJourney didn’t quite meet this requirement (i cannot control the GenAI...). To solve this, I used Photoshop’s AI features along with Xingtu (a photo editing app on phone) to fine-tune specific areas. This allowed me to adjust facial expressions and backgrounds with greater precision, ensuring consistency while also adding richness and control to the visual effects.
Since I needed to eventually import the work into Runway for dynamic processing, it was crucial that the generated some facial features and background remained consistent across frames. However, the #remix and #edit functions in MidJourney didn’t quite meet this requirement (i cannot control the GenAI...). To solve this, I used Photoshop’s AI features along with Xingtu (a photo editing app on phone) to fine-tune specific areas. This allowed me to adjust facial expressions and backgrounds with greater precision, ensuring consistency while also adding richness and control to the visual effects.
Bringing Her to Life
To bring the Flower Vase Girl to life, I imported the generated image into Runway, set the first and last frames, and then played around with different prompts to prevent the visuals from getting too abstract. The last part of the piece has really fast changes in her face, so I generate a bunch of short videos and stitch them together to create smooth transitions and effects.
Since the videos generated by Runway always have some shaking, it could cause the face to shift outside the projected area. So, I added some few keyframes in the editing software to keep the head as stable as possible and avoid it drifting too much.
To bring the Flower Vase Girl to life, I imported the generated image into Runway, set the first and last frames, and then played around with different prompts to prevent the visuals from getting too abstract. The last part of the piece has really fast changes in her face, so I generate a bunch of short videos and stitch them together to create smooth transitions and effects.
Since the videos generated by Runway always have some shaking, it could cause the face to shift outside the projected area. So, I added some few keyframes in the editing software to keep the head as stable as possible and avoid it drifting too much.
Projection
I imported the video generated by Runway into MadMapper and used the #mask tool to select the Flower Vase Girl's face for the projection. Since the object I was projecting onto was a curved head model, I tried a few 3D scanning apps on my phone to scan the model and bring it into MadMapper, but none of them worked out. In the end, I went with adding multiple wrapping points to the projected face and manually adjusted it so the projection would fit the curved surface as accurately as possible.
To avoid the projector interfering with the audience's interaction, I went with an ultra-short throw projector, which only needs a 22cm distance. I also made a wooden board with “Five Yuan for a Song” written on it to cover the projector,
I imported the video generated by Runway into MadMapper and used the #mask tool to select the Flower Vase Girl's face for the projection. Since the object I was projecting onto was a curved head model, I tried a few 3D scanning apps on my phone to scan the model and bring it into MadMapper, but none of them worked out. In the end, I went with adding multiple wrapping points to the projected face and manually adjusted it so the projection would fit the curved surface as accurately as possible.
To avoid the projector interfering with the audience's interaction, I went with an ultra-short throw projector, which only needs a 22cm distance. I also made a wooden board with “Five Yuan for a Song” written on it to cover the projector,
On-site Adjustments
All the models, frames, and equipment were brought to a rented studio, where I did the on-site adjustments.
During the testing, I realized that the original plan to place the sensor under the vase didn’t work because of a wooden board obstruction. So, I ended up placing the sensor in a less noticeable spot under the table, covering the Arduino and Breadboard board with a red fabrics, which kept it discreet while still making sure the distance measurement function worked properly.
Credits
A project by Qian Mu
Sculpture made with Arduino, p5.js, Midjourney, Runway.ML, MadMapper, and physical
Index Next