mycast

Text-to-speech using KittenTTS.

Setup

Requires uv. No global Python packages needed.

uv sync

This installs Python 3.12 (if needed) and all dependencies into a local .venv/.

Usage

Generate speech from text

uv run python main.py tts input.txt
uv run python main.py tts input.txt -o episode1
uv run python main.py tts input.txt -v Luna -o episode1
uv run python main.py tts input.txt -s 1.2              # speak faster
uv run python main.py tts input.txt -s 0.8              # speak slower

Output is MP3

Create a new podcast feed

uv run python main.py new-podcast "My Podcast" -d "A podcast about things"
uv run python main.py new-podcast "My Podcast" -o podcasts/feed.xml
uv run python main.py new-podcast "My Podcast" --force   # overwrite existing

Creates an RSS 2.0 feed template (feed.xml by default). Edit it to fill in your podcast details (link, image, category, etc.).

Generate speech and add to a podcast feed

Creates an mp3 from the input text file and adds the results to the podcast feed.

uv run python main.py tts 2026-04-06.txt -f feed.xml -t "April 6 News"

The -f flag appends the generated audio as a new episode to the feed file. -t sets the episode title (defaults to the output filename).

When using -f:

• MP3 and transcript files are placed in the same directory as the feed
• The episode date is extracted from the input filename (YYYY-MM-DD.txt)
• The episode description is the text before the first --- in the input file
• A WebVTT transcript (.vtt) is generated next to the mp3 with timestamps distributed proportionally (by sentence length) across the audio duration
• A <podcast:transcript> tag (Podcasting 2.0) links to the .vtt file with type="text/vtt"
• Running again for the same date replaces the existing episode

Add an existing mp3 + transcript as an episode

If you already have an mp3 and transcript and just want to add it to the feed without re-running TTS:

uv run python main.py add-episode feed.xml episode.mp3 2026-04-07.txt
uv run python main.py add-episode feed.xml episode.mp3 2026-04-07.txt -t "April 7 News"

The mp3 and transcript files are copied into the feed's directory (overwriting if they already exist), and the episode is added/replaced in the feed.

Manage Files in R2 Bucket

Assumes the rclone tool is installed.

List bucket contents for a bucket called mycast (assumes a rclone remote configured named r2:

rclone ls r2:mycast

Upload entire output directory contents:

rclone copy ./output r2:mycast

Voices

Voice	Gender
Bella	Female (default)
Jasper	Male
Luna	Female
Bruno	Male
Rosie	Female
Hugo	Male
Kiki	Female
Leo	Male

Notes

• The TTS model (~25MB) is downloaded from HuggingFace on first run and cached at ~/.cache/huggingface/hub/. Subsequent runs use the cache offline — no network requests.
• MP3 output is encoded at 320kbps CBR via ffmpeg.