All work

Grumble

An app where students report small problems at school by speaking them. Speech becomes text, lands in a category, and bubbles show how big each problem is.

Speech-to-text, categorising, recording and storageduoFeb – May 2025Flutter, OpenAI API (Whisper, GPT-4o)Finished

Grumble's category screen, with bubbles that grow with the number of complaints.

The problem

Students have no place to share the small but annoying problems at school, like dirty toilets or bad food. Our idea: let them report a problem anonymously, and lightly, by just saying it.

What I built

  • Speech to text, and sorting. The app records a complaint, Whisper turns it into text, and GPT-4o puts it in one of the existing categories. I first tried DeepSeek and switched to OpenAI a day later.
  • Recording and storage: the audio service and the complaint model. Complaints are saved in a local JSON file.
  • Tests for loading complaints and for the connection to the speech and language models.
  • The app’s structure: the welcome screen, separate handling for iOS and Android, and a fixed screen orientation.
Grumble's screen for recording a complaint.
Recording a complaint. The app turns it into text and sorts it into a category.

Research and design

Much of the project was research and user testing: a survey, paper prototypes, a competitor scan and an advice report. We talked the design through together.

My part

We were a duo. I built the speech-to-text and the sorting into categories, and the recording and storage. My teammate did most of the frontend, including the bubbles that show how big each problem is.

What I learned

A lot about UX, and about research methods like surveys and testing paper prototypes with students.

Status

Finished. The code is on GitHub.