voice-studio | Open Source AI Voice Studio, in Hermes

What it does

Six tools. One job each.

Each tool is mirrored as a slash command, and the bundled voice-studio skill ships usage guidance. Say what you want in chat. Hermes does the rest.

01

Mirror a voice from three seconds.

Clone a voice from a short wav or m4a sample. Saved for reuse, with a preview take before you commit.

/vs-design --sample

02

Describe a voice. Get it.

Write the voice you hear in your head in plain words. The studio designs it, you keep it.

/vs-design

03

Any video, any language, same timing.

Dub a video into other languages with one of your voices. Finished languages are kept if one fails.

/vs-dub

ESDEFRJA+

04

Cast a script. Voice every role.

Turn markdown into chaptered audio. Headings split chapters. Switch narrator mid-script. Interrupted renders resume.

/vs-longform

05

Say it like you mean it.

Read short text aloud in a chosen voice and save it as an audio file. Repeatable seeds for consistent takes.

/vs-speak

06

Every voice you own, one list.

Browse saved voices, the shared gallery, and presets, then check the studio is awake. Read-only and cheap.

/vs-voices/vs-status

Why this way

Voice AI as owned infrastructure,
not rented SaaS.

The plugin inherits the posture of the app it drives: your machine does the work, your voices stay yours.

	VOICESTUDIO PLUGIN	CLOUD TTS APIS
Runs on	Your machine	Their cloud
Voice data	Never leaves your machine	Uploaded to vendor servers
Pricing	Free · open source (MIT)	Per-character billing
Voice cloning	From a sample you provide	Often restricted or paywalled
Source	Open	Closed
Internet	Not needed after setup	Required

Install

Two commands in.

The catalog route pins the exact commit a maintainer reviewed. Manual install points at the voicestudio/ subdirectory, which carries the manifest and the package together.

hermes plugins install voicestudio
hermes plugins enable voicestudio

git clone https://github.com/aska-digital/protean-voice-studio
hermes plugins install ./protean-voice-studio/voicestudio

Prerequisite: the free, open-source VoiceStudio app, running locally (default http://127.0.0.1:3900). Get it at voicestudio.sh, start it, then /vs-status tells you when it is back.

In practice: an audiobook, start to finish:

/vs-voices --search narrator
/vs-longform --file episode.md --voice c79141c4 --format m4b \
  --title "Episode 12" --author "Aska Digital"

The plugin plans the chapters first and renders with progress. If the render is interrupted, rerun with --resume <job_id>. Nothing is lost.

Scripting

Long-form markup.

vs_longform accepts a closed markup set. Anything outside it is read aloud literally, so strip it first.

Markup	Effect
[pause], [pause 500ms]	Silence, default or timed
[slow] / [fast] / [emphasis] / [spell]	Delivery spans
[[word\|say-this]]	Pronunciation pin
[voice:NAME]	Mid-script voice switch
# Heading	Starts a new chapter

Four settings, all optional.

Setting	What it does
base_url	VoiceStudio server base URL (default `http://127.0.0.1:3900`)
default_voice	Default narrator voice id when the user names none
default_engine	Preferred engine when several are loaded
timeout_s	Read timeout in seconds for generation calls (default 600)

Quick answers

Little questions, short answers.

Is it really free?

Yes. The plugin is MIT licensed, and the VoiceStudio desktop app it drives is free and open source.

Does it phone home?

No account, no cloud. The plugin only talks to your local VoiceStudio server. Dubbing uploads your video and cloning uploads your sample to that same local server. Nothing reaches a third party.

Do I need a fancy GPU?

No. The app runs on CPU; a GPU just makes everything faster.

Can I clone any voice?

Only with the owner's say-so. Impersonation and fraud are off the table.

Which Hermes version do I need?

0.21 or newer. The manifest declares it.

Is it in the plugin catalog?

The submission is in review. Until it lands, install manually from the repo URL.

Disclosure

What it does on your machine.

For the record.

All network traffic stays between Hermes and your VoiceStudio server. Nothing is sent to third-party services.
vs_dub uploads your video file to the VoiceStudio server. vs_design_voice --sample uploads your sample recording. Both stay on your local server.
The plugin reads the script, video, and sample files you point it at, and writes generated audio under the plugin output directory (or the --out_path you give).
No credentials are read or stored. No telemetry. No self-updates: catalog installs stay pinned to the reviewed commit.

Source

Open, pinned, tested.

MIT licensed. Manifest plus package live together in voicestudio/; the catalog entry pins subdir: voicestudio. Unit suite: 32 passed, 1 skipped (mocked HTTP, no server needed). Run it with bash run_tests.sh.

aska-digital/protean-voice-studio Release v1.0.0

v1.0.0 · pinned 4cf4ab2 stdlib only · zero dependencies no telemetry · no stored credentials

Open Source AI Voice Studio,
in Hermes.

Six tools. One job each.

Voice AI as owned infrastructure,
not rented SaaS.

Two commands in.

Long-form markup.

Little questions, short answers.

What it does on your machine.

Open, pinned, tested.

Give Hermes a voice.

Open Source AI Voice Studio,in Hermes.

Six tools. One job each.

Voice AI as owned infrastructure,not rented SaaS.

Two commands in.

Long-form markup.

Little questions, short answers.

What it does on your machine.

Open, pinned, tested.

Give Hermes a voice.

Open Source AI Voice Studio,
in Hermes.

Voice AI as owned infrastructure,
not rented SaaS.