Moodtips
MoodTips is an interactive public art installation designed for office workers in the Zhangjiang community of Shanghai. By combining facial expression recognition with emotional response suggestions in a lighthearted and accessible way, the project transforms emotional states—which are typically private and often overlooked—into experiences that can be perceived and responded to.
Through a continuous process of input and output, emotions are temporarily preserved and concretized in the form of tangible physical traces. As these traces accumulate within the public space, they create a subtle emotional flow and a sensory connection between people.
The project aims to introduce a momentary emotional exchange into a monotonous and high-pressure daily routine, allowing positive emotions to emerge, flow, and be quietly shared within a collective environment.
Year of Completion:
2026
Awards and Exhibitions
Participating in the "Re-Coded Being" Art Exhibition
Responsible for:
Research, conceptual design, device development and setup, and experience design
Video
Background
Zhangjiang is a technology park located in Shanghai's Pudong New Area, primarily composed of technology enterprises and universities. At the invitation of WISON, a key enterprise in the area, our team created a localized interactive experience artwork for it.
After living and studying in Zhangjiang for a period of time, we observed that in this efficiency-oriented environment, many commuters and students are under significant psychological pressure for long periods, making it easy for their morale to drop. As members of this community, we hope to take this opportunity to bring a touch of vitality and care to this area.
Research
Interview
Situational Observation
Concept
Therefore, we decided to design a participatory installation to help people in the Zhangjiang community relieve negative emotions through a quick and engaging interactive experience, while also strengthening the sense of connection within the community.
Design:
Interaction Design:
The core experience begins when the participant stands in front of the installation. The system captures and assesses their current emotional state through facial expression recognition technology, and generates a personalized MoodTip for that participant accordingly. Each MoodTip contains the identified emotion type along with a brief emotional suggestion.
But where exactly do these emotional suggestions come from?
The emotional suggestions are co-provided by community participants. We have set up multiple input points within the community, with the cafe serving as the primary location, making it convenient for people to stop by and write more leisurely and seriously. Participants scan a QR code to access a web interface and leave emotional suggestions for strangers. In this anonymous exchange, individuals support each other through their shared emotional experiences.
UI Design
The emotional role system consists of five types: Happy, Sad, Numb, Energetic, and Angry. These emotional roles are printed on the MoodTip. Based on the results of facial expression recognition, the system will randomly assign participants different role variants corresponding to their current emotional state.
Technical Development
The project consists of three main modules: a web-based input system, a sentiment analysis system, and an output execution system, which together form a complete closed loop of MoodTip generation and delivery.
Web-based input system
On the input side, the project uses Framer to build and deploy the website. Users submit mood suggestions through the webpage, which are then stored in a Google Sheet to serve as the project's content database. Subsequently, GPT reads the data from the sheet, performs semantic analysis on each MoodTip to determine its most suitable corresponding mood state, and assigns the appropriate mood label to it.
Sentiment Analysis System
On the emotion recognition side, the overall interaction logic and data coordination are built by TouchDesigner. The real-time video captured by the camera is transmitted to the OpenCV and MediaPipe models, which are used to analyze facial features, expression changes, and related emotional cues.
On the input side, the project uses Framer to build and deploy the website. Users submit mood suggestions through the webpage, which are then stored in a Google Sheet to serve as the project's content database. Subsequently, GPT reads the data from the sheet, performs semantic analysis on each MoodTip to determine its most suitable corresponding mood state, and assigns the appropriate mood label to it.
Execution Output System
On the output side, the system retrieves matching MoodTips from a categorized database based on the recognized emotions and triggers an automatic printing process through Folder Mill. MoodTips are output in the form of physical prints, thereby completing the transformation from digital input to tangible feedback.
On the input side, the project uses Framer to build and deploy the website. Users submit mood suggestions through the webpage, which are then stored in a Google Sheet to serve as the project's content database. Subsequently, GPT reads the data from the sheet, performs semantic analysis on each MoodTip to determine its most suitable corresponding mood state, and assigns the appropriate mood label to it.
Overall, the system constructs a continuously operating path of emotional flow: "emotional input from others → AI's semantic understanding → real-time emotion recognition → physicalized emotional feedback," allowing emotions to flow and be redistributed among different media and participants.
Production process:
Final Outcomes and Exhibition
Other Projects
Moodtips
MoodTips is an interactive public art installation designed for office workers in the Zhangjiang community of Shanghai. By combining facial expression recognition with emotional response suggestions in a lighthearted and accessible way, the project transforms emotional states—which are typically private and often overlooked—into experiences that can be perceived and responded to.
Through a continuous process of input and output, emotions are temporarily preserved and concretized in the form of tangible physical traces. As these traces accumulate within the public space, they create a subtle emotional flow and a sensory connection between people.
The project aims to introduce a momentary emotional exchange into a monotonous and high-pressure daily routine, allowing positive emotions to emerge, flow, and be quietly shared within a collective environment.
Year of Completion:
2026
Awards and Exhibitions
Participating in the "Re-Coded Being" Art Exhibition
Responsible for:
Research, conceptual design, device development and setup, and experience design
Video
Background
Zhangjiang is a technology park located in Shanghai's Pudong New Area, primarily composed of technology enterprises and universities. At the invitation of WISON, a key enterprise in the area, our team created a localized interactive experience artwork for it.
After living and studying in Zhangjiang for a period of time, we observed that in this efficiency-oriented environment, many commuters and students are under significant psychological pressure for long periods, making it easy for their morale to drop. As members of this community, we hope to take this opportunity to bring a touch of vitality and care to this area.
Research
Interview
Situational Observation
Concept
Therefore, we decided to design a participatory installation to help people in the Zhangjiang community relieve negative emotions through a quick and engaging interactive experience, while also strengthening the sense of connection within the community.
Design:
Interaction Design:
The core experience begins when the participant stands in front of the installation. The system captures and assesses their current emotional state through facial expression recognition technology, and generates a personalized MoodTip for that participant accordingly. Each MoodTip contains the identified emotion type along with a brief emotional suggestion.
But where exactly do these emotional suggestions come from?
The emotional suggestions are co-provided by community participants. We have set up multiple input points within the community, with the cafe serving as the primary location, making it convenient for people to stop by and write more leisurely and seriously. Participants scan a QR code to access a web interface and leave emotional suggestions for strangers. In this anonymous exchange, individuals support each other through their shared emotional experiences.
UI Design
The emotional role system consists of five types: Happy, Sad, Numb, Energetic, and Angry. These emotional roles are printed on the MoodTip. Based on the results of facial expression recognition, the system will randomly assign participants different role variants corresponding to their current emotional state.
Technical Development
The project consists of three main modules: a web-based input system, a sentiment analysis system, and an output execution system, which together form a complete closed loop of MoodTip generation and delivery.
Web-based input system
On the input side, the project uses Framer to build and deploy the website. Users submit mood suggestions through the webpage, which are then stored in a Google Sheet to serve as the project's content database. Subsequently, GPT reads the data from the sheet, performs semantic analysis on each MoodTip to determine its most suitable corresponding mood state, and assigns the appropriate mood label to it.
Sentiment Analysis System
On the emotion recognition side, the overall interaction logic and data coordination are built by TouchDesigner. The real-time video captured by the camera is transmitted to the OpenCV and MediaPipe models, which are used to analyze facial features, expression changes, and related emotional cues.
On the input side, the project uses Framer to build and deploy the website. Users submit mood suggestions through the webpage, which are then stored in a Google Sheet to serve as the project's content database. Subsequently, GPT reads the data from the sheet, performs semantic analysis on each MoodTip to determine its most suitable corresponding mood state, and assigns the appropriate mood label to it.
Execution Output System
On the output side, the system retrieves matching MoodTips from a categorized database based on the recognized emotions and triggers an automatic printing process through Folder Mill. MoodTips are output in the form of physical prints, thereby completing the transformation from digital input to tangible feedback.
On the input side, the project uses Framer to build and deploy the website. Users submit mood suggestions through the webpage, which are then stored in a Google Sheet to serve as the project's content database. Subsequently, GPT reads the data from the sheet, performs semantic analysis on each MoodTip to determine its most suitable corresponding mood state, and assigns the appropriate mood label to it.
Overall, the system constructs a continuously operating path of emotional flow: "emotional input from others → AI's semantic understanding → real-time emotion recognition → physicalized emotional feedback," allowing emotions to flow and be redistributed among different media and participants.
Production process:
Final Outcomes and Exhibition
Other Projects
Moodtips
MoodTips is an interactive public art installation designed for office workers in the Zhangjiang community of Shanghai. By combining facial expression recognition with emotional response suggestions in a lighthearted and accessible way, the project transforms emotional states—which are typically private and often overlooked—into experiences that can be perceived and responded to.
Through a continuous process of input and output, emotions are temporarily preserved and concretized in the form of tangible physical traces. As these traces accumulate within the public space, they create a subtle emotional flow and a sensory connection between people.
The project aims to introduce a momentary emotional exchange into a monotonous and high-pressure daily routine, allowing positive emotions to emerge, flow, and be quietly shared within a collective environment.
Year of Completion:
2026
Awards and Exhibitions
Participating in the "Re-Coded Being" Art Exhibition
Responsible for:
Research, conceptual design, device development and setup, and experience design
Video
Background
Zhangjiang is a technology park located in Shanghai's Pudong New Area, primarily composed of technology enterprises and universities. At the invitation of WISON, a key enterprise in the area, our team created a localized interactive experience artwork for it.
After living and studying in Zhangjiang for a period of time, we observed that in this efficiency-oriented environment, many commuters and students are under significant psychological pressure for long periods, making it easy for their morale to drop. As members of this community, we hope to take this opportunity to bring a touch of vitality and care to this area.
Research
Interview
Situational Observation
Concept
Therefore, we decided to design a participatory installation to help people in the Zhangjiang community relieve negative emotions through a quick and engaging interactive experience, while also strengthening the sense of connection within the community.
Design:
Interaction Design:
The core experience begins when the participant stands in front of the installation. The system captures and assesses their current emotional state through facial expression recognition technology, and generates a personalized MoodTip for that participant accordingly. Each MoodTip contains the identified emotion type along with a brief emotional suggestion.
But where exactly do these emotional suggestions come from?
The emotional suggestions are co-provided by community participants. We have set up multiple input points within the community, with the cafe serving as the primary location, making it convenient for people to stop by and write more leisurely and seriously. Participants scan a QR code to access a web interface and leave emotional suggestions for strangers. In this anonymous exchange, individuals support each other through their shared emotional experiences.
UI Design
The emotional role system consists of five types: Happy, Sad, Numb, Energetic, and Angry. These emotional roles are printed on the MoodTip. Based on the results of facial expression recognition, the system will randomly assign participants different role variants corresponding to their current emotional state.
Technical Development
The project consists of three main modules: a web-based input system, a sentiment analysis system, and an output execution system, which together form a complete closed loop of MoodTip generation and delivery.
Web-based input system
On the input side, the project uses Framer to build and deploy the website. Users submit mood suggestions through the webpage, which are then stored in a Google Sheet to serve as the project's content database. Subsequently, GPT reads the data from the sheet, performs semantic analysis on each MoodTip to determine its most suitable corresponding mood state, and assigns the appropriate mood label to it.
Sentiment Analysis System
On the emotion recognition side, the overall interaction logic and data coordination are built by TouchDesigner. The real-time video captured by the camera is transmitted to the OpenCV and MediaPipe models, which are used to analyze facial features, expression changes, and related emotional cues.
On the input side, the project uses Framer to build and deploy the website. Users submit mood suggestions through the webpage, which are then stored in a Google Sheet to serve as the project's content database. Subsequently, GPT reads the data from the sheet, performs semantic analysis on each MoodTip to determine its most suitable corresponding mood state, and assigns the appropriate mood label to it.
Execution Output System
On the output side, the system retrieves matching MoodTips from a categorized database based on the recognized emotions and triggers an automatic printing process through Folder Mill. MoodTips are output in the form of physical prints, thereby completing the transformation from digital input to tangible feedback.
On the input side, the project uses Framer to build and deploy the website. Users submit mood suggestions through the webpage, which are then stored in a Google Sheet to serve as the project's content database. Subsequently, GPT reads the data from the sheet, performs semantic analysis on each MoodTip to determine its most suitable corresponding mood state, and assigns the appropriate mood label to it.
Overall, the system constructs a continuously operating path of emotional flow: "emotional input from others → AI's semantic understanding → real-time emotion recognition → physicalized emotional feedback," allowing emotions to flow and be redistributed among different media and participants.
Production process:
Final Outcomes and Exhibition