Context

Over 70 million people stutter globally, yet most speech technologies fail to serve them, because they’re trained only on fluent speech.

Current speech technologies, such as automated phone menus or computer-generated captions, often fail to work well for people who stutter due to training on fluent speech. To make these speech technologies more responsive and accessible for people who stutter, it is important to collect more diverse stuttered speech data to improve these technologies.

Challenge

Design a mobile app that empowers individuals who stutter to practice speaking confidently in a supportive and low-pressure environment. The app should also encourage voluntary data contribution for more inclusive AI speech technologies, with a focus on user consent, transparency, and clarity around dataset use. You may incorporate elements like gamification, progress tracking, or personalized goals to make the experience engaging and user-driven.

User Research

To set the stage for our work, we interviewed several students and individuals who stutter, refined our objectives based on their feedback, and distilled our findings into these three key insights.

Solution

With the insights in hand, we knew what to do, create a gamified mobile app that makes practice low-pressure and engaging for folks who stutter. Hello, SpeakEasy!

Based on our research, we defined a primary user persona that reflect our target users’ pain points, goals, and requirements.

Once we aligned on the focus, we mapped a clear information architecture, defined the key design and logo elements, and moved straight into execution under a tight deadline.

Iteration

Given the compressed timeline, we moved directly from whiteboard wireframes to mid and high-fidelity prototypes. We iterated a lot, retiring plenty of versions, and arrived at a robust UI for the final push, enhanced with engaging artwork and fluid animations.

The dashboard was one example of the several iterations we did throughout the design sprint. As the primary entry point, the dashboard received the most refinement to achieve the ideal balance of clarity and engagement.

Final Designs

Login Screen with Transparency

We ensured users knew precisely where their data was going and how their voice recordings would be used. Based on interviews, transparency emerged as the key expectation, so we treated it as a non-negotiable top priority.

During interviews, we asked participants about their go-to games, and these four came up again and again. We chose them because (1) they’re universally familiar, so the learning curve is minimal, and (2) they’re word-based without heavy chatter, so players can warm up without feeling overloaded.

Individual Game Designs

Game 1: Wavelength

I love Wavelength, it’s my go-to road-trip game, so getting to design it was a treat. I built the flow with smooth transitions between each guess and recording, and made the UI snap to edits in real time. I also prioritized SpeechPal at the end, our AI that gives players personalized pointers on what to watch for, so we can improve accessibility in our speech tech while supporting players’ speech goals.

Game 2: Scattegories

Next up: Scattergories. We kept the UI deliberately minimal so players could lock in on the timer and their word list. After each round, SpeechPal delivers a quick debrief of a few things: what it noticed, where you nailed it, and a couple of easy wins for next time, if applicable.

Game 3: Contact!

We built Contact for flow: a clean countdown, a mic cue, and instant transitions from guess to result. Sync on the same word to score, miss, and you can adjust and try again without breaking rhythm. Post-round, SpeechPal offers pointers, clarity, pacing, and filler usage, to keep progress steady and accessible.

Game 4: Brainteaser

Brainteaser gives players a fun way to exercise both their brain and their speech through daily riddles. Players can solve each riddle by speaking into the mic, with the option to re-record their answer or flip the card for a hint when they’re stuck. A new riddle unlocks each day, encouraging players to come back and keep practicing. SpeechPal also offers helpful tips and tricks along the way when applicable.

Reflections

Winning second place was definitely a plus, but what really stuck with me from this experience was just how quickly our team clicked. It was the first time I’d ever worked with this group, but from the jump, ideas were always flying around, every single person brought something brilliant to the table. I remember thinking, “Yep, these are my people. I want to do every designathon with them from now on.”

Key Takeaways

• Prioritize fuller speech samples (phrases and sentences, not just single words) to help the models learn faster. Add game modes that prompt longer responses (mini-stories, sentence chains, guided prompts).
• Tune SpeechPal’s feedback for longer utterances: pacing, prosody, and filler-word patterns.
• Update scoring to reward clarity and complexity, not just speed or accuracy.