All plugins

podcast

Turn any article, PDF, paper, notes or the day's news into a mastered two-host podcast episode: neural voices per host, speaker-aware pauses, intro chime, -16 LUFS loudness, embedded chapters and a timestamped transcript in one MP3. Includes an audio-overview skill and a daily news-briefing blueprint. Script text is sent to Microsoft's online Edge text-to-speech service via edge-tts (no account or key); episodes stay local. No telemetry. Disclosure — episode script text (every spoken line), voice names and rate/pitch are sent to Microsoft's online Edge text-to-speech service via edge-tts (no account or key); the voice list is fetched from the same service and cached for 7 days; audio, transcripts and metadata stay in the plugin data directory; uses ffmpeg from PATH or the user's ffmpeg_path setting.

suhas-svgVoicev1.0.0Updated 2 hours ago

About

Turn any article, PDF, paper, notes or the day's news into a mastered two-host podcast episode: neural voices per host, speaker-aware pauses, intro chime, -16 LUFS loudness, embedded chapters and a timestamped transcript in one MP3. Includes an audio-overview skill and a daily news-briefing blueprint. Script text is sent to Microsoft's online Edge text-to-speech service via edge-tts (no account or key); episodes stay local. No telemetry. Disclosure — episode script text (every spoken line), voice names and rate/pitch are sent to Microsoft's online Edge text-to-speech service via edge-tts (no account or key); the voice list is fetched from the same service and cached for 7 days; audio, transcripts and metadata stay in the plugin data directory; uses ffmpeg from PATH or the user's ffmpeg_path setting.

Podcast Studio for Hermes

Turn anything Hermes can read (an article, a PDF, a paper, your notes, today's
news) into a real two-host podcast episode you can listen to: natural neural
voices, pauses between speakers, an intro chime, loudness mastering, chapter
markers and a timestamped transcript, all in one MP3.

You: Read https://en.wikipedia.org/wiki/Rayleigh_scattering and make me a
     1-minute two-host podcast episode about it.
Hermes: Done: 1-minute episode with Alex and Sam, mastered MP3.
        - Chapters: Why the sky is blue (00:02), Tiny molecules, huge
          effect (00:15), Beyond the sky (00:41)
        MEDIA:.../episodes/20261107-140110-why-the-sky-is-blue/episode.mp3

On Telegram, Discord and other messaging platforms the episode arrives as an
audio file you can play right in the chat.

What it adds

| Surface | What you get |
| --- | --- |
| Tools | podcast_produce, podcast_voices, podcast_episodes |
| Skill | podcast:audio-overview: how to read sources and write a script that sounds like two people talking |
| Automation blueprint | podcast:daily-briefing: a morning news episode on topics you choose, delivered to your chat |
| Slash command | /podcast list, /podcast voices LOCALE, /podcast doctor |
| CLI | hermes podcast list, hermes podcast play [ID], hermes podcast voices [LOCALE], hermes podcast doctor |

| Tool | Does |
| --- | --- |
| podcast_produce | Takes a script (title, 1-3 hosts, lines with optional chapters), synthesizes every line in its host's voice, then masters the episode: speaker-aware pauses (shorter for the same speaker, longer between chapters), an intro chime, EBU R128 loudness normalisation to -16 LUFS (the podcast standard), ID3 tags and embedded chapter markers. Writes episode.mp3, transcript.md and episode.json. |
| podcast_voices | Lists the 300+ available voices by locale and gender, with their personality tags, so episodes can be in Spanish, Hindi, Japanese and more. |
| podcast_episodes | Lists past episodes, or re-sends one by id. |

A prompt alone can't do this. The model writes the script, but the plugin turns
it into produced audio: parallel synthesis with throttle-aware pacing and
retries, sample-accurate assembly, mastering and chapters, and a library of
episodes you can replay.

Requirements

  • Hermes Agent 0.22.5 or newer.
  • An internet connection (voices are synthesized online, see Privacy).
  • ffmpeg for mastering. Hermes ships one; otherwise any ffmpeg on PATH
    works, or set ffmpeg_path. Without ffmpeg the plugin still delivers an
    episode, but joins the clips without chime, pauses, loudness pass or chapters.
  • Python dependency: edge-tts (installed
    by Hermes with the plugin).

Install

hermes plugins install suhas-svg/hermes_podcast --enable

Once the plugin is in the curated catalog:

hermes plugins install podcast --enable

Then check the setup:

hermes podcast doctor

Try it

  • "Make a 3-minute podcast about this article: "
  • "Turn this PDF into an audio overview I can listen to on my walk."
  • "Read my notes in notes.md as a single-narrator episode."
  • "Haz un episodio de dos minutos en español sobre ."
  • With the Zotero Research plugin: "Make a podcast episode about the papers in
    my Zotero collection on diffusion models."
  • Schedule podcast:daily-briefing with /blueprint for a news episode every
    morning.

Settings

Set under plugins.entries.podcast.settings in config.yaml, for example
hermes config set plugins.entries.podcast.settings.pause_ms 450.

| Key | Default | Meaning |
| --- | --- | --- |
| host_a_voice | en-US-AndrewMultilingualNeural | Voice of the first default host (Alex) |
| host_b_voice | en-US-AvaMultilingualNeural | Voice of the second default host (Sam) |
| pause_ms | 350 | Pause between speakers, in milliseconds |
| intro_chime | true | Play a chime at the start of each episode |
| max_lines | 160 | Most lines one episode may have |
| max_chars | 24000 | Most spoken characters (about 25 minutes) |
| concurrency | 3 | Lines synthesized in parallel |
| ffmpeg_path | empty | ffmpeg to use; empty means find it on PATH |

Episodes are stored in the plugin's data folder
(<HERMES_HOME>/plugin-data/podcast/episodes/), one folder per episode.

Privacy

The spoken text of each script line is sent to Microsoft's online neural
text-to-speech service (the one behind Microsoft Edge's Read Aloud) through the
open-source edge-tts library. No account or API key is used. Don't produce
episodes from text you are not willing to send to Microsoft. Nothing else leaves
your machine; episodes, transcripts and the cached voice list stay local.

Development

python -m pip install pytest edge-tts
python -m pytest -q

Tests inject a fake synthesizer, so they run offline. The mastering test needs
ffmpeg and ffprobe on PATH and is skipped without them.

License

MIT

README from the reviewed commit 1c0c4ed; it updates when the author re-pins.

More by suhas-svg

2 plugins by suhas-svg in the catalog.

suhas-svgankiCreate, search, and review Anki flashcards from Hermes through the local AnkiConnect API: adds notes to a deck (creating it if needed), searches the collection, lists due cards, and grades reviews. Writes to the local Anki collection and changes scheduling only when its tools or /anki command are used. Requires Anki desktop with AnkiConnect (add-on 2055492159); no telemetry; optional API key via ANKI_BRIDGE_API_KEY.Communitytools