Voice Synth

Turn your emails into training data for a GPT that writes like you.

Quick Start

Open Terminal and paste:

curl -fsSL https://raw.githubusercontent.com/s-b-e-n-s-o-n/voice-synth/main/install.sh | bash

That's it. The TUI guides you through everything.

Run again later:

~/voice-synth/voice-synth-tui

What You'll Need

Your emails from Google Takeout:

1. Go to takeout.google.com
2. Click "Deselect all", then select only Mail
3. Download (any size - we handle split exports)

What It Does

1. Imports your mbox (filters spam/trash/drafts, emails older than 5 years)
2. Converts to processing format
3. Cleans & Anonymizes - filters to emails you sent, replaces PII
4. Curates the best examples across topics, removes duplicates

Output: ~/Desktop/style_shortlist.csv

Features

• Visual TUI - Progress bars, stage tracking, live output
• Auto-resume - Quit anytime, pick up where you left off
• Auto-detect - Finds your email address from the mbox
• Drag & drop - Drop files from Finder into the terminal

Privacy

All processing happens locally. Nothing leaves your machine.

PII is replaced using Microsoft Presidio (NLP-based detection):

• Names → [PERSON]
• Emails → [EMAIL]
• Phones → [PHONE]
• Addresses → [LOCATION]
• Credit cards, SSNs, IPs, URLs → anonymized

Command Line

Skip the TUI and run directly:

cd ~/voice-synth

# Full pipeline
~/.cache/voice-synth/venv/bin/python pipeline.py run takeout.mbox --sender you@gmail.com

# Individual stages
~/.cache/voice-synth/venv/bin/python pipeline.py import mail.mbox --out emails.json
~/.cache/voice-synth/venv/bin/python pipeline.py convert emails.json --out emails.jsonl
~/.cache/voice-synth/venv/bin/python pipeline.py clean emails.jsonl --out cleaned.json --sender you@gmail.com
~/.cache/voice-synth/venv/bin/python pipeline.py curate cleaned.json --out shortlist.csv

Uninstall

From the TUI menu, select "Uninstall" and type uninstall to confirm.

Or manually:

rm -rf ~/voice-synth ~/.cache/voice-synth

Requirements

• macOS (ARM64) - Intel/Linux users can build from source
• Python 3.9+

Building from Source

# Install Go 1.21+
go build -o voice-synth-tui .

# Dependencies installed automatically on first run