sase-listen

Turn Markdown into chaptered, loudness-normalized MP3 audio editions — narrated by Gemini TTS, built for listening to SASE research on a commute or walk.

uv tool install sase-listen
sase-listen doctor
sase-listen render report_narration.md -o episode.mp3

Each episode is a mono 24 kHz, 64 kb/s MP3 at −16 LUFS with one ID3 chapter per ## heading, an embedded cover, and a spoken AI-disclosure intro. Episodes reach your phone through Telegram's music player or a private podcast feed for AntennaPod. Full docs: https://sase-org.github.io/sase-listen/.

How it works

1. Write a narration script — by hand for research reports (sase-listen guide prints the authoring rules, brief by default, sase-listen lint checks them), deterministically from any Markdown (sase-listen script), or from an article or PDF URL or local PDF file (sase-listen script https://example.com/article, sase-listen render paper.pdf -e full).
2. Render it — sase-listen render synthesizes each chapter, gates quality chunk by chunk, masters to a loudness-normalized MP3, and commits the episode to a local library with a manifest.
3. Listen — publish to your private feed (sase-listen publish) or get the MP3 delivered over Telegram.

See Getting started and How it works.

Status notes

These commands are not implemented yet and exit 1 with a stub message:

• sase-listen audition, sase-listen ls, sase-listen cache
• sase-listen doctor --online (offline checks work; the one-word live synth check is still a stub)

Everything else in sase-listen --help works. The CLI reference marks each stub.

Contributing

See CONTRIBUTING.md. Bug reports and PRs welcome; keep Conventional Commit titles so release-please can cut releases.

Background

sase-listen implements the commute-audio research via epic sase-1e3. Details and provenance links: Background.