SpeechForGood

Automatic speech recognition (ASR) engines suffer from a major problem: they do not collect sufficient data from users who have speech impairments. To correct this imbalance, we designed SpeechForGood, a web application that changes data collection into an engaging, interactive game. Users complete short, playful voice challenges to upload clear audio samples directly to an open-source database. Instead of static forms, the app presents vibrant audio targets, progress bars, and reward badges. This playful design turns a boring task into an enjoyable daily activity.

During a fast-paced designathon, our team focused on accessibility, simple visual interfaces, and user motivation. We tested button sizes, high-contrast layouts, and instant visual feedback loops so users with motor or vocal constraints can navigate the application without difficulty. The judges awarded SpeechForGood first place because the product solves a real technical barrier with empathy and smart interaction design. Through contributions from people of different speech patterns, this web app in turn, helps machine-learning models understand every human voice.