Turn years of exported WhatsApp conversations into a beautiful, interactive memory experience.
DearChats reads a chat export and rebuilds it as a private 3D universe of curated moments — drifting photos, ambient music, and a quiet narrator that points at the things you'd otherwise forget. The ordinary messages that become precious years later.
The easiest way to experience DearChats is the hosted version — no setup required:
Sign in with Google, upload a WhatsApp export, and start exploring for free.
- 3D constellation — your relationship as a universe of mood-colored stars, each a curated moment.
- Year navigation — drill into any year as a 12-month mini-constellation with per-month photo decks.
- Reader Mode — a memoir-style scroll for those who'd rather read (also the fallback when WebGL is unavailable).
- Private Dictionary — your nicknames, rituals, inside jokes, and recurring phrases, with first/last-used dates.
- Memory Film — a short, shareable 9:16 video cut from the emotional arc of the whole chat.
- A restrained narrator — prose that earns emotion through specifics, never adjectives.
WhatsApp export (.zip / .txt)
│
▼
Parse + segment → AI curation pipeline → 3D universe + reader + film
DearChats runs a multi-stage AI pipeline: it triages messages, maps the relationship into chapters, extracts recurring patterns, curates the most meaningful moments per year, and writes all the on-screen prose in a single consistent voice. Photos are classified to prefer real shared moments over screenshots.
Built with: Next.js 15 (App Router) · React 19 · TypeScript · Tailwind · three / react-three-fiber · OpenAI · SQLite (better-sqlite3) · ffmpeg + @napi-rs/canvas for video.
- Node.js 20+
- ffmpeg on your
PATH(only needed for the Memory Film feature) - An OpenAI API key
- Google OAuth credentials (sign-in is Google-only)
git clone https://github.com/opxiahub/dearchats.git
cd dearchats
npm installcp .env.example .env.localFill in .env.local:
| Variable | Required | Notes |
|---|---|---|
OPENAI_API_KEY |
✅ | Powers the curation pipeline |
GOOGLE_CLIENT_ID |
✅ | From the Google Cloud Console |
GOOGLE_CLIENT_SECRET |
✅ | From the Google Cloud Console |
GOOGLE_CALLBACK_URL |
✅ | http://localhost:3000/api/auth/google/callback for local |
OPENAI_MAIN_MODEL / OPENAI_MINI_MODEL / OPENAI_VISION_MODEL |
optional | Override the default models |
NEXT_PUBLIC_MAX_UPLOAD_MB |
optional | Max upload size in MB (defaults to 250) |
Google OAuth setup: create OAuth credentials in the Google Cloud Console and add
http://localhost:3000/api/auth/google/callbackas an Authorized redirect URI.
npm run devOpen http://localhost:3000.
| Command | Description |
|---|---|
npm run dev |
Start the dev server |
npm run build |
Production build |
npm start |
Run the production build |
npm run typecheck |
Type-check with tsc |
Data is stored locally in .data/ (SQLite database + uploaded media + rendered films). It is gitignored and never leaves your machine.
A Dockerfile is included (it installs ffmpeg and bundles the native dependencies). Build and run it with your environment variables supplied at runtime.
Your conversations are personal. When self-hosted, everything stays on your own infrastructure — chat exports, media, and generated content live in the local .data/ directory and are only sent to the OpenAI API for curation. Nothing is shared publicly.
MIT © Opxia