Getting Started with FiavaionDictate

Five steps to your first dictation

Step 1 — Download the app

1. Click the button above to go to the Releases page
2. Download FiavaionDictate-v0.1.0-beta.zip (the file under “Assets”)
3. Extract the zip anywhere you like — your Desktop, Documents, wherever makes sense

That’s it. No installer, no terminal commands, no git needed.

Not ready to install yet? Try the live demo first — dictation works right in the browser. (Project management and some AI features need the local server, but you can test the core experience.)

Developer? You can also git clone https://github.com/Fiavaion/dictate.git if you prefer to pull updates with git pull.

Step 2 — Start the app

• Windows: Double-click start.bat
• Mac / Linux: Double-click start.command (or run bash start.command from a terminal)

A small terminal window will appear — that’s the server running. Leave it open while you use the app.

Step 3 — The app opens in your browser

FiavaionDictate automatically opens at http://localhost:8080. You’ll see the main dictation interface — a large text area, a microphone button, and a status bar along the bottom.

If the browser doesn’t open automatically, open Chrome or Edge yourself and go to http://localhost:8080.

Step 4 — Follow the Setup Wizard

On your first visit, the Setup Wizard runs automatically. It has five short steps:

1. System Check — confirms the server is running and checks your browser
2. Local AI — helps you set up Ollama for free, private AI correction
3. Cloud AI — shows you free options if you’d prefer cloud AI
4. API Keys — lets you paste any keys you have
5. Ready — you’re set up and ready to dictate

Each step is optional except Step 1. You can skip any AI setup step and come back to it later from the Settings panel.

Step 5 — Click the mic and start talking

Click the microphone button (or press Space when the text area is focused). It turns red when listening. Start speaking naturally.

Try saying: “hello world period” — you should see “Hello world.” appear in the text box. The word “period” is a voice command that inserts a full stop.

When you’re done, click the mic again or say “stop dictation”.

Tip: You can also click directly into the text area to type or paste text manually. Dictation and manual editing work side-by-side — you can speak a paragraph, then click in to fix a word, then continue dictating. To use AI selection transforms (simplify, rewrite, fix, etc.) on existing text, highlight a range with your mouse and say the command.

Getting the files

Option A — Download the release zip (recommended)

Go to the Releases page and download the .zip file from Assets. Extract it anywhere.

Option B — Clone with git (for developers who want easy updates)

git clone https://github.com/Fiavaion/dictate.git
cd dictate

Then run git pull any time you want to update.

Running the server manually

The launcher scripts just call python server.py. You can run it directly:

python server.py

To change the port:

python server.py --port 9000

The server then listens at http://localhost:9000. No other flags are required.

Requirements: Python 3 standard library only. No pip installs, no virtual environments, no dependencies. If you have Python 3 installed, it runs.

config.json

Place a config.json in the app root to configure persistent settings:

{
  "projectsRoot": "/path/to/your/projects"
}

Leave config.json absent entirely if you don’t use the project switcher — the app works fine without it.

Server architecture

The Python server has three responsibilities:

1. Static file server — serves the entire public/ directory at /
2. /api/projects — lists subdirectories of projectsRoot for the project switcher panel
3. /api/ai/proxy — forwards requests to cloud AI providers (Gemini, Anthropic, OpenAI). This proxy exists solely to avoid CORS issues; it adds no processing and does not log request bodies.

All speech recognition happens in the browser via the Web Speech API. The server never receives audio.

Security model

API keys are stored using AES-GCM encryption with PBKDF2 key derivation in your browser’s localStorage. The encryption key is derived from a device fingerprint — keys are tied to your browser profile and are not readable by any other site or by the FiavaionDictate server.

Keys are transmitted directly from your browser to the AI provider’s API endpoint. They do not pass through any Fiavaion infrastructure.

To remove all stored keys: Settings → Security → Clear All Keys.

GitHub Pages mode

When you access FiavaionDictate from the GitHub Pages URL (https://fiavaion.github.io/dictate/), the app auto-detects that no local server is present. In this mode:

• The project switcher panel is hidden
• The /api/ai/proxy route is unavailable — cloud AI calls go directly to provider APIs (requires CORS support from the provider, which Gemini supports)
• All dictation and local AI features work normally

Browser requirements

The Web Speech API is a browser feature — FiavaionDictate cannot polyfill it. Chrome and Chromium-based Edge are the reliable choices.