sumsori

README excerpt

# Sumsori (숨소리)

**A communication bridge that reads between the lines — turning unspoken feelings into shareable art.**

Sumsori analyzes what you say (or write) and uncovers what you _really_ mean. It transforms the gap between spoken words and hidden feelings into expressive illustrations, helping people communicate emotions they struggle to put into words.

> Built for the **Google DeepMind Gemini 3 Seoul Hackathon** (Feb 28, 2026)

---

## What It Does

Sumsori serves as a **communication assistant** — not an emotion analyzer, but a tool that helps people deliver the feelings they can't express directly.

### Voice Mode

Record a short voice message (2–30 seconds). Sumsori listens to _how_ you speak (tone, pace, tremor, pauses) alongside _what_ you say (words, themes, sentiment), then reveals the real feeling underneath.

**Example:**
- You say: _"I'm fine, really. Eating well, doing great."_
- Your voice says: Trembling, slow, fragile
- What's really inside: **Loneliness**
- Result: `#contentment → #loneliness` with an illustration of quiet solitude

### Text Mode (Accessibility)

Designed with **deaf and hard-of-hearing users** in mind. Type what you want to say, choose a voice persona, and Sumsori will:

1. Analyze the surface meaning vs. hidden feeling in your words
2. Generate a **spoken audio version** (TTS) with emotional delivery
3. Create an illustration that captures the true feeling

This gives people who communicate primarily through text a way to convey emot