Wingman

Inspiration

Are YOU socially awkward? Do YOU wish you had more rizz texting and calling your friends, crushes, significant others, whomever? Or maybe YOU'RE just lazy and want someone to write your messages for you! We get it. All of the above apply to the devs of Wingman, which is why we were inspired to make this app -- to empower us and take our texting and calling aura to the next level. ;) 😗

In all seriousness - What it does

Wingman is a browser extension that's compatible with Discord. It'll read your chat history with the person (or people) of your choice, and if you're stuck on what to say next, or want it to spruce up the response you're about to send, it'll provide a suggestion for you. More than that, it'll listen in on your calls _ live _ and generate on-the-spot responses to what the other(s) in the call might be saying. You'll never feel lost for words with your trusty Wingman on your side.

How we built it

We built the frontend using React and Tailwind CSS, while the backend is powered by Flask. For web scraping, we analyzed the HTML structure of Discord pages. We also leveraged the Google Gemini API, using custom prompts to parse images and generate responses. To handle audio input from voice calls, we captured and processed tab audio via JavaScript, then sent the resulting WAV files to the Gemini speech-to-text API for transcription.

Challenges we ran into

We hit a few bumps along the way while building Wingman. One of the toughest challenges was getting real-time transcription of live calls working smoothly. Capturing audio from Discord and processing it with the Gemini speech-to-text API was trickier than expected, especially with all the browser permissions and the technical details involved. Another issue was making sure our text suggestions were relevant and didn’t feel overwhelming. We wanted to provide helpful prompts without bombarding the user with too many options. And, of course, designing an interface that felt simple and intuitive, while still delivering all the cool features, took a lot of trial and error.

Accomplishments that we're proud of

We’re super proud of how the real-time audio transcription turned out. Being able to generate responses during live calls, with little to no delay, feels like a big win. Another big achievement was integrating the Google Gemini API to generate smart, context-aware text and image suggestions. But overall, we’re most excited that we built something that can actually help people feel more confident and engaged in their conversations—whether they’re texting or talking on a call.

What we learned

We definitely learned a lot along the way! Working with real-time data, especially audio, was a whole new world for us, and it gave us tons of insight into how to optimize performance. We also got hands-on experience with web scraping and had to get creative to deal with Discord’s constantly changing HTML structure. Most importantly, we learned how to fine-tune AI to offer more personalized, relevant responses without feeling too robotic.

What's next for Wingman

Now that Wingman has proven to be a success for Discord, we'd like to take it to the next level by extending Wingman functionality to other social media platforms such as Instagram, Facebook, Twitter, etc, and maybe someday to mobile. The potential is sky-high for our Wingman to take flight.

Built With

• flask
• gemini
• html
• javascript
• react
• tailwind