transcribe.dev Documentation

Welcome to the transcribe.dev documentation. This directory contains all technical documentation for the transcribe.dev voice-to-text platform.

Product Overview

transcribe.dev is a system-wide voice-to-text dictation tool with AI enhancement:

Desktop Dictation App

Wispr Flow Competitor - System-wide voice-to-text dictation with AI cleanup

Direction: Speech → Text (STT + AI enhancement)
Platform: Desktop (macOS first, Windows/Linux planned)
Tech: Electron, Deepgram, Claude Haiku, Convex
Status: Active Development

PRD: transcribe-dev-bootstrap-PRD.md

Phase 2: Voice Synthesis API (FUTURE)

ElevenLabs Competitor - Developer-first text-to-speech API

Direction: Text → Speech (TTS)
Platform: Cloud API
Tech: Neural TTS, FastAPI, GPU infrastructure
Status: Planning (begins after Phase 1)

Future docs: future/voice-synthesis/

Documentation Structure

docs/
├── README.md                         # This file
├── transcribe-dev-bootstrap-PRD.md   # PRD (PRIORITY)
├── TENETS.md                         # Core development principles
├── architecture/
│   ├── PROJECT_STRUCTURE.md      # Monorepo structure
│   └── PHASE1_IMPLEMENTATION.md  # Phase 1 technical implementation
├── authentication/
│   └── README.md                 # Auth overview (WorkOS AuthKit)
├── development/
│   ├── GIT_PREFERENCES.md        # Git workflow
│   ├── TYPESCRIPT_BEST_PRACTICES.md
│   ├── SHADCN.md                 # UI component guide
│   └── UI_STANDARDS.md           # UI/UX guidelines
└── future/
    └── voice-synthesis/          # Phase 2 documentation

Quick Start

For Development

Read the PRD - Full product specification
Review PHASE1_IMPLEMENTATION.md - Technical implementation details
Review TENETS.md - Development principles
Check PROJECT_STRUCTURE.md - Monorepo layout
Follow GIT_PREFERENCES.md - Git workflow

Key Technologies (Phase 1)

Component	Technology	Purpose
Desktop App	Electron + React	Cross-platform desktop UI
STT Engine	Deepgram (cloud) / whisper.cpp (local)	Speech-to-text
AI Cleanup	Claude Haiku	Grammar, punctuation, tone
Text Injection	nut.js	Type into active application
Backend	Convex	Real-time sync, user data
Auth	WorkOS AuthKit	User authentication

Development Guidelines

Must-Read Documents

TENETS.md - Core principles guiding all decisions
GIT_PREFERENCES.md - Commit workflow (never push without permission)
TYPESCRIPT_BEST_PRACTICES.md - TypeScript guidelines

Code Quality

TypeScript strict mode always
No any types without justification
Test compilation locally before deployment
Follow UI standards for consistent UX

Monorepo Structure

transcribe-dev/
├── apps/
│   ├── transcribe/     # Web app / Dashboard
│   ├── desktop/        # Electron dictation app
│   └── api/            # Hono API (Cloudflare Workers)
├── convex/             # Shared Convex backend
├── packages/
│   ├── ui/             # ShadCN UI components
│   ├── shared/         # Shared types and constants
│   └── ...
└── docs/               # This documentation

Contributing

When adding documentation:

Phase 1 docs go in the appropriate existing folder
Phase 2 docs go in future/voice-synthesis/
Use clear, descriptive filenames (CAPS_WITH_UNDERSCORES.md)
Link between related documents
Keep docs close to the code they describe

Current Focus: Building the desktop dictation app. All development effort should prioritize Phase 1 features.

On this page