podcast_generator

Converts markdown podcast scripts into spoken-word audio files, using voxport's KokoroDirectTTS for local, offline-capable text-to-speech. No API key, no per-use cost, and your scripts never leave your machine.

• podcast-generator: converts scripts from the command line.
• Drop-folder mode: a Windows scheduled task converts whatever lands in an inputs folder (cloud-synced, if you like), controlled from the small Podcast Watch app (podcast-watch).
• podcast-generator-mcp: an MCP server, so Claude Desktop or another MCP client can write, queue and convert podcasts for you.

This README is a quick start. See MANUAL.md for the full guide: script-writing conventions, complete CLI reference, the voice catalog, and troubleshooting.

Installation

Requires Python 3.10+. The command line works anywhere Kokoro does; the Podcast Watch app and scheduled drop-folder mode are Windows-only.

pip install podcast-generator             # command line and Podcast Watch
pip install "podcast-generator[mcp]"      # plus the MCP server

A virtual environment is recommended. This pulls in Kokoro and torch (about 2 GB) through voxport[kokoro]. Check the install:

podcast-generator --version

Not installed for you:

• ffmpeg: needed for MP3 output unless your soundfile/libsndfile build has native MP3 support. Put it on PATH. WAV output needs neither.
• Kokoro model weights and spaCy's en_core_web_sm: downloaded automatically on the first run. On a firewalled machine, see "Offline use" in the manual.

From a source checkout

git clone https://github.com/brienalkire/podcast_generator.git
cd podcast_generator
python -m venv .venv
.venv\Scripts\activate          # Windows (bash: source .venv/bin/activate)
pip install -e ".[mcp]" pytest

Working on voxport at the same time? Clone it next to this repo and run pip install -e "../voxport[kokoro]" afterwards.

Where things live

Settings and the default folders live in one home folder:

It holds .env (settings), inputs/ (scripts waiting to be converted), outputs/ (finished audio), script_archive/ (converted scripts, and failed/ for ones that didn't convert), and generator.log.

The three folders can point anywhere. Use the Podcast Watch app, or set PODCAST_INPUTS_DIR, PODCAST_OUTPUTS_DIR and PODCAST_ARCHIVE_DIR in .env (see .env.example). Pointing them at a cloud-synced folder (Sync.com, Dropbox, OneDrive) lets you drop a script in from any device and get the audio back the same way.

Usage

Writing a script

A script is a markdown file. Plain paragraphs are spoken. Headers, **Key:** value metadata lines, horizontal rules, and italic-only lines are stripped, so you can label sections and add notes without them being read aloud:

# My Episode Title

**Voice:** af_heart

## OPENING

Welcome back to the show. Today we're talking about...

## CLOSING

That's all for this episode. Thanks for listening.

---

*Source: some-article.com, September 2026.*

Only the two paragraphs are spoken. An italic-only line after the prose starts ends the script, and everything after it is dropped. Keep those lines for trailing notes. See the manual for the full rules.

Running it

podcast-generator                                  # batch: every script in inputs/
podcast-generator --voice af_heart --speed 1.1     # batch, overriding voice and speed
podcast-generator my-script.md --format wav --output episode01.wav   # one file

Batch mode (no file argument) converts every .md script in the inputs folder and writes <script-name>.mp3 to the outputs folder. The first run creates the inputs folder if it doesn't exist. Each script that converts successfully moves into the archive folder, so re-running doesn't reprocess it. A script that fails moves to <archive folder>/failed/ with a .error.txt note, so it isn't retried forever.

Single-file mode converts just the named file. --output only applies here.

Voice precedence: --voice, then the script's **Voice:** line, then af_heart.

Drop-folder mode and the Podcast Watch app (Windows)

Running podcast-generator with no arguments on a schedule is watch mode: each run converts whatever is new in the inputs folder and archives it. A run with nothing to do exits almost immediately, and only one conversion runs at a time.

The Podcast Watch app sets this up and controls it:

podcast-watch               # open the app
podcast-watch --shortcut    # add a "Podcast Watch" shortcut to your desktop

• Set up watch registers a Windows scheduled task, PodcastGeneratorWatch, for your user (no admin rights). It runs the converter with pythonw.exe, so no window flashes up; output goes to generator.log in the home folder.
• Turn ON / Turn OFF enables or disables it; Check every sets the interval (5 to 60 minutes).
• When this app opens can always turn the watch on or off when you open the window.
• Folders picks the inputs, outputs and archive folders.

To remove the task: Unregister-ScheduledTask -TaskName PodcastGeneratorWatch in PowerShell. On macOS or Linux, schedule podcast-generator with cron or a similar tool instead.

Using it from Claude Desktop (MCP)

podcast-generator-mcp is an MCP server, so an AI app such as Claude Desktop can make podcasts directly: write a script, queue it, convert it, and check on it, all from the chat. It runs locally and uses the same folders and .env as everything else.

1. Install it somewhere that stays put. A uv or pipx tool install works well:

   uv tool install "podcast-generator[mcp]"
   

   (pipx install "podcast-generator[mcp]" is the same. A virtual environment works too, as long as you don't delete it.)

2. Print the Claude Desktop config for that install:

   podcast-generator-mcp --config
   

   It gives the command's full path, which Claude Desktop needs because it doesn't see your shell's PATH.

3. In Claude Desktop, open Settings → Developer → Edit Config, merge that "podcast-generator" entry into mcpServers, and fully quit and restart Claude Desktop (on Windows, quit from the system tray). It then appears in the chat box's connectors menu.

See the manual for the tools and an example conversation.

License

MIT. See LICENSE.