Extension Center
Find extensions by tag or choose Official releases. Review the source and install an approved version.
Explore extensions
8 extensions
spel
Officialblockether · v0.1.21
Spel browser automation with verified installation, reserved sessions, CDP, JavaScript, snapshots and screenshots.
vis-decisions
Officialblockether · v0.6.1
Train, resume, validate and publish Laya, GLiNER, Decision 2.0 and D3 decision models in Vis.
vis-graphify
blockether · v0.15.3
Code knowledge graphs for Vis with graphify: search, change impact, exact references, safe renames and moves, domain flows, and a knowledge base server with ontology, documents and undoable edits.
vis-lang-clojure
Officialblockether · v1.17.3
Clojure tools for Vis: zprint and cljfmt formatting, clj-kondo lint, reader syntax checks, test runs and an nREPL.
vis-lang-interface
Officialblockether · v2.11.3
The shared contract for Vis language extensions.
vis-lang-python
Officialblockether · v1.14.0
Python tools for Vis: ruff formatting and lint, pytest runs and a managed project REPL.
vis-optmem
Officialblockether · v0.6.0
Permanent memory for Vis agents, based on OptMem by Victor Taelin.
vis-podcaster
Officialblockether · v0.1.0
Two-host podcasts with Vis speech engines, ASR-verified takes and quality gates.
blockether/vis-podcaster
OfficialTwo-host podcasts with Vis speech engines, ASR-verified takes and quality gates.
Install
Only approved releases are listed. Each version keeps its reviewed commit.
Approved releases (1)
- 0.1.0Oct 26, 2026546426a6b045Release notes
This command selects version 0.1.0 at its approved commit. It never follows a moving tag or branch.
Project: run from your project directory to install in .vis/extensions. This command does not add an extensions entry to vis.yml. Add --save to also record the installation. Save an installation in configuration to share the setup.
Review the source first. --trust allows extension code and build backends to run with your permissions.
vis-agent extension install \
'blockether/vis-podcaster' \
--version '0.1.0' \
--project \
--trustThen start Vis or use /reload. Dependencies are prepared automatically.
Already installed? Check, update or roll back
vis-agent extension versions 'blockether/vis-podcaster'
vis-agent extension update 'blockether/vis-podcaster' --trust
vis-agent extension rollback 'blockether/vis-podcaster' --trustUse the identifier shown above, not the Python package name. Add --project for a project installation. Update selects the latest approved stable version; rollback restores the previous pinned source. Choose a specific version and read the lifecycle contract.
Project details
- Version
- 0.1.0
- Extension
- blockether/vis-podcaster
- Project folder
- Repository root
- License
- Apache-2.0
- GitHub stars
- 0
- Downloads
- 0
- Updated
- Oct 26, 2026
- GitHub owner
- Blockether
- Reviewed commit
- 546426a6b045
- Metadata checked
- Oct 26, 2026
Dependencies
Python >=3.12
- vis-agent>=0.3.4
README
vis-podcaster
Turn a written dialogue into a mastered two-host podcast episode with Vis speech engines.
When to use
- You want a short spoken episode on a topic, with two hosts who talk to each other.
- You want each line checked by speech recognition, so that a mispronounced line is recorded again.
- You want to compare your episode with audio from another tool, against the same quality gates.
What you get
Ask Vis for an episode in a chat, for example: "Make a five-minute podcast about the Broad Street
cholera outbreak." The podcast skill then does these steps:
- It researches the topic, writes the script and checks the facts against sources.
- It scores the script against ten review criteria, such as hook, roles, listening and ending.
- It speaks each line, checks it with speech recognition and records it again when the check fails.
- A second recognizer judges the finished file.
The result is episode.mp3, episode.vtt subtitles and a quality report. The report gives the word
error rate, the critical facts that the listener hears, loudness and pace.
Requirements
- A Vis build with the speech API (
vis.speech). - For the fastest and most natural voices: Apple silicon and a Python with
mlx-audioat~/.vis/podcaster/mlx-venv/bin/python, or inVIS_PODCASTER_QWEN_PYTHON. Without it, the same voices work onqwen3-tts-local, but slower. - A five-minute episode takes about five minutes to render on
qwen3-tts-mlx.
Install
Install vis-podcaster from the Vis Extension Center, then select it for your sessions.
Use it from Python
check = await podcast.check("episode.json")
print(check)
r = await podcast.render("episode.json", "out/episode")
print(r)
skills/podcast/SKILL.md gives the script format, the voices and the repair steps.
examples/great-wall.json is a complete script.
Voices
The named voices (stuart, caro, fabien, estelle and others) are cloned from bundled clips
with CC0-1.0 or CC-BY-4.0 licenses. src/vis_podcaster/voices/voices.json gives the source and the
credit of each clip. Clone only a voice whose owner agrees.
Develop
uv sync
uv run pytest
uv run ruff check src tests
License
Apache License 2.0. See LICENSE.
Community feedback
Votes and display names are not verified identities. One vote per network address; shared networks share a vote. Select your current choice again to clear it.
Loading feedback…