Landing page
In call with Anger character
In call with Sad character
In call with Anxiety character
In call with Joy character
Summary tab

Inspiration

Inspired by the film Inside Out, we wanted to tackle the chaos of internal monologues. Many people struggle to process complex emotions because they feel like a tangled web. We create Reflexia, where your feelings aren't just abstract thoughts—they are distinct characters you can actually talk to.

What it does

Reflexia is a platform that transforms a person's inner world into a virtual chat room. Each of your core emotions (joy, anger, anxiety, etc.) is represented by a unique AI character in a video-call style interface. Users can host a meeting for the emotions to navigate difficulties, or simply find balance through real-time, multi-dimensional dialogue with their own psyche.

How we built it

We utilized Agora to power the real-time interaction in the meeting rooms. OpenAI drives the distinct personalities of each emotion. We also integrated ElevenLabs for high-quality, character-specific voice synthesis.

Challenges we ran into

The biggest hurdle was managing pipeline latency. Coordinating a flow that goes from Agora (Audio) → STT → LLM (Processing) → ElevenLabs (TTS) → Agora (Playback) in real-time is incredibly demanding. Reducing the "lag" between a user speaking and an "emotion" responding was critical to maintaining the immersion of a natural conversation.

What we accomplished

We successfully transformed the concept of "internal reflection" into a functional system. Seeing a user engage in a conversation between themselves, and their emotions—and actually gaining clarity from it—was a huge win. We are also proud of the modular architecture we built, which allows for stable real-time communication across a complex AI stack.

What we learned

We gained deep insights into the nuances of AI personality prompting—ensuring that "Anxiety" sounds cautious but helpful rather than just stressful. Technically, we learned how to optimize high-latency pipelines and the importance of tight integration between RTC (Real-Time Communication) tools and generative AI.

What's next for Reflexia

The next step is to be more predictive, helping users identify emotional patterns over time. We also plan to optimize the pipeline further to achieve near-zero latency and explore adding more "character slots" so users can customize their own emotional boardrooms.