FiavaionDictate

What is FiavaionDictate?

Speak your thoughts. Get polished text.

FiavaionDictate is a browser-based dictation tool built for people who want to think out loud and get clean, structured output. Speak naturally — it transcribes in real time, sends your text through an AI correction pass, and gives you a diff of every change so you’re always in control.

Built for developers using Claude Code, but powerful enough for any writing workflow.

How It Works

FiavaionDictate runs as a lightweight local server — one Python file, zero dependencies beyond Python 3. Open your browser, hit the mic, and you’re dictating.

Your voice → Web Speech API → Raw transcript
                                ↓
                    80+ voice commands (punctuation, formatting, editing)
                                ↓
                    AI correction (Ollama / Gemini / Claude / GPT)
                                ↓
                    Clean output + diff sidebar

The two-pane interface shows your raw dictation on the left and the AI-corrected version on the right. Every correction is shown as a diff — accept, reject, or ignore with a voice command.

Key Features

Real-Time Speech-to-Text

Built on the Web Speech API — the same engine powering Chrome’s built-in voice input — FiavaionDictate delivers fast, accurate transcription with confidence feedback. Words are colour-coded by confidence so you can see at a glance where to check.

Supports all languages available in your browser’s speech engine.

80+ Voice Commands

No mouse required. Control everything with your voice:

Category	Examples
Punctuation	”period”, “comma”, “question mark”, “new paragraph”
Editing	”delete that”, “scratch that”, “undo”, “clear all”
Formatting	”bold that”, “heading two”, “bullet point”, “code block”
AI	”correct that”, “simplify that”, “improve that”
Session	”save session”, “new session”, “play macro”

Full voice commands reference →

AI Correction — Four Ways

Choose how much AI involvement you want:

Mode	Privacy	Cost	Best For
Ollama (local)	100% private	Free	Most users — great quality, zero cost
Google Gemini	Cloud	Free tier	Quick setup, generous daily limits
Anthropic Claude	Cloud	Paid	Nuanced writing, long-form content
OpenAI GPT	Cloud	Paid	Versatile, well-known models

All API keys are encrypted with AES-GCM and stored only in your browser’s localStorage — never sent to Fiavaion.

Set up AI providers →

Macro Recorder

Record a sequence of voice commands and replay them instantly. Build macros for your most common workflows — start a meeting note, create a code comment block, format a status update — and trigger them with a single phrase.

Session Management

Every dictation session is automatically saved. Search your history, pick up where you left off, or export sessions for use elsewhere. Sessions are stored locally in your browser.

Analytics Dashboard

Track your dictation habits over time — word count trends, most-used commands, correction rates by AI provider. All data stays on your device.

Quick Start

Three steps:

1. Download — grab the zip, extract it, and you’re ready. Or try the live web version first (local AI not available in the live version).

2. Start — double-click start.bat (Windows) or start.command (Mac/Linux). Your browser opens to http://localhost:8080 automatically.

3. Setup Wizard — on first launch, the built-in wizard guides you through choosing an AI mode and testing your microphone. Takes about 2 minutes.

Full Getting Started guide →

First-Run Wizard walkthrough →

Technical Details

• Zero dependencies — pure Python 3 standard library. No pip install required.
• Runs anywhere — Windows, macOS, Linux. Any machine with Python 3 and Chrome/Edge.
• Browser-first — all UI logic runs client-side as ES6 modules. No build step.
• Encrypted storage — API keys use PBKDF2 key derivation + AES-GCM encryption via the Web Crypto API.
• Local server — server.py serves static files and two API routes: /api/projects (file system project discovery) and /api/ai/proxy (secure cloud API forwarding).
• GitHub Pages mode — accessing from fiavaion.github.io/dictate/ auto-detects no local server and hides project-specific features. STT and cloud AI still work.

Browser Requirements

Browser	STT Support	AI Support
Chrome 80+	✓ Full	✓ All providers
Edge 80+	✓ Full	✓ All providers
Firefox	✗ Web Speech API not supported	✓ Cloud AI only
Safari	Limited	✓ Cloud AI only

For the best experience, use Chrome or Edge.

Privacy & Data

FiavaionDictate is built privacy-first:

• No tracking — no analytics, no usage data sent anywhere
• No cloud processing by default — use Ollama for fully offline AI
• No external servers — your text never touches Fiavaion infrastructure
• Encrypted keys — API keys are locked in your browser with AES-256 equivalent encryption
• Open source — every line of code is on GitHub

When using cloud AI (Gemini, Claude, OpenAI), your text is sent directly to that provider’s API — not routed through Fiavaion. You control which provider you use and can switch or clear keys at any time.

What’s New in 0.1.0 Beta

First Public Release

The core dictation loop, AI correction pipeline, and voice command system are ready for testing. We’d love your feedback.

Onboarding Wizard

A brand-new first-run wizard walks you through AI setup step by step — including automated Ollama detection, cloud provider comparison, and a live mic test.

Macro Recorder

Record any sequence of voice commands and play them back with a single phrase. Macros are saved per session and persist across browser reloads.

AES-GCM Key Encryption

All API keys are now encrypted at rest using PBKDF2 + AES-GCM via the Web Crypto API. No plaintext keys ever stored.

Documentation

• Getting Started — Install and set up in minutes
• First-Run Wizard — Detailed wizard walkthrough
• AI Setup — Configure Ollama, Gemini, Claude, or OpenAI
• Voice Commands Reference — All 80+ commands