WeChat Style Distiller
Turn your WeChat chat history into a linguistic style profile and a persona prompt that makes any LLM reply in your voice.
What It Analyzes
Message Patterns
Length distribution, bubble bursts, punctuation habits, turn-taking speed
Linguistic Features
Emoji frequency, laughter tokens, catchphrases, opener/closer habits
Code-Switching
English/Chinese ratio, language mixing patterns, register awareness
Privacy-First
All analysis runs locally. Your chat data never leaves your machine.
About This Project
Personal NLP project by Sunchuangyu Huang, 2026. This is a methodology showcase demonstrating how linguistic style profiling and persona generation work.
The original tool extracts patterns from your real WeChat history — how you actually text (message length, emoji usage, code-switching, catchphrases) — and generates a system prompt that makes any LLM reply in your voice.
This demo site uses synthetic personas only to demonstrate the methodology without accessing real user data. Explore the pipeline, feature analysis, and pre-generated personas to understand how the tool works.
Tech Stack: Original tool is Python 3.12+ with pure-stdlib pipeline. This showcase is built with Next.js 16, TypeScript, Tailwind CSS v4, and deployed on Vercel.
Repository: github.com/rNLKJA/wechat-style-distiller