PlugWiki

byted-mediakit-audio Skill for Claude: What It Does and How to Install It

Lets Claude find speech, split voice from background sound, convert audio and read metadata through MediaKit.

Installs
12.3K
on skills.sh
Stars
218
on GitHub
Files
5
besides SKILL.md
Updated
Sep 2026
last repo commit

byted-mediakit-audio lets Claude run four audio tools from the mediakit-cli audio group. They find the start and end of speech, read audio metadata, separate voice from background sound and transcode audio. Three of the four are Cloud tools and metadata reading runs in Cloud or Local mode.

Before any tool, Claude reads the shared skill and then the tool's reference file. Trimming, mixing and other edits go to the editing skill, and subtitles go to the video skill. It loads when you work with audio files or the audio track of a video.

Good for

  • Podcasters and editors who need voice split from background sound
  • Anyone converting audio to another codec or container
  • Teams finding speech segments in recordings

Install byted-mediakit-audio

Any client

  1. 1

    Have these values ready, then replace the placeholders in the snippet.

    MEDIAKIT_API_KEYRequiredCloud API key from the Volcengine console. The README shows local mode without a key.
    MEDIAKIT_OUTPUT_PATHOptionalFolder for local-mode output. The README default is ~/.mediakit/temp.
  2. 2

    Run this in a terminal.

    shell
    npx @volcengine/mediakit-cli install -y

Requires Node.js 18 or later.

The installer sets up the mediakit-cli tool and the skills that ship with it. The editing, audio, image and video skills tell Claude to read byted-mediakit-shared first, so install the set and not one skill alone. There is no Claude app upload route because the skills run the local mediakit-cli command.

From the project README. Source.

What's in the skill

When it loads

面向音频文件或视频中的音轨,处理语音边界定位、音频媒资信息探测、音频转码与码流封装适配、人声与背景声分离等目标。若对象和目标族已明确属于音频内容理解、音频转码、音频格式治理或音轨分离,但具体做法不确定,可先加载本 Skill 探索。

Sections in SKILL.md

  1. 使用规则
  2. 澄清与跨域路由
  3. 工具列表

Key facts

License
MIT
Installs on skills.sh
12,288 (Nov 2, 2026)
Last repo commit
Sep 2, 2026
Skill version
0.2.1 (from SKILL.md frontmatter)

What it includes

  • Find the start and end times of speech in audio or video

    Cloud only.

    Network · detect-voice-activity

  • Read standard media info for an audio file

    Runs in Cloud or Local mode.

    Runs code · probe-audio-metadata

  • Split voice and background sound into two audio files

    Cloud only.

    Network · separate-voice

  • Convert audio codec, parameters and container

    Cloud only.

    Network · transcode-audio

From the official docs

Permissions

Medium risk

What it can do

  • Read your data
  • Create and change things
  • Delete things
  • Read-only mode

Watch out for

  • Cloud tools send your media to MediaKit

    Input goes to MediaKit's cloud as a public HTTP/HTTPS URL, a MediaKit, VOD or TOS link, or, for tools that accept it, a local file path that the CLI's media adapter uploads (per byted-mediakit-shared SKILL.md). The README names cloud-first as the default mode, so check the mode before giving Claude private media. github.com

  • Cloud calls use your MediaKit API key and Volcengine billing rules

    Cloud tools run through the MediaKit API key you create in the Volcengine console. The README says using the cloud APIs means following Volcengine's billing rules, which it links, and it lists no prices here. It describes local mode as zero cost. Read the rules before batch jobs. github.com

  • The installer and updater run npm packages

    The install command runs @volcengine/mediakit-cli through npx, and the README says mediakit-cli update uses npm install -g. Install only if you trust the package. github.com

  • Output can be written to your own storage

    Cloud tool references describe an optional media_output_destination (vod://space or tos://bucket). Setting it stores results in your Volcengine VOD space or TOS bucket, and the first use needs you to authorize MediaKit to write there. Claude is told to pass it only if you give it. github.com

Security advisories

No published security advisories as of Nov 2, 2026.

Technical details
Turning on read-onlyNo read-only mode. Review each command before it runs.
Talks toMediaKit cloud API on Volcengine (Cloud mode: your media or its URL, plus your API key)
Sign-in and keysREADME: MEDIAKIT_API_KEY comes from the Volcengine console. mediakit-cli init stores it using the method you choose (config, shell or export), or you can set it as an environment variable.
Scopes
  • The skill's frontmatter asks for the shell permission and requires the mediakit-cli binary

  • Claude sets MEDIAKIT_SURFACE=skill and MEDIAKIT_RUNTIME on each real call so the CLI knows the caller

  • Local mode runs ffmpeg and ffprobe on your machine and writes results to ~/.mediakit/temp by default

Questions

What does byted-mediakit-audio do?

It lets Claude run MediaKit's audio tools through mediakit-cli. They find where speech starts and ends, separate voice from background sound, convert audio formats and read audio metadata.

How do I install byted-mediakit-audio?

Use the install box at the top of this page. It runs the MediaKit installer, which sets up the mediakit-cli command and the skills together. Then follow the set-up and check step in the box.

Do I need an API key?

For Cloud tools, yes: the README says to get a MediaKit API key from the Volcengine console. The README shows local trimming as running without an API key, and local mode needs ffmpeg 5.1.x and ffprobe.

Can I upload it to the Claude app?

No. The skill runs the local mediakit-cli command through the shell, so it needs a client that can run commands on your machine, such as Claude Code. The README does not document a Claude app route.

Can it trim or mix audio?

Not here. SKILL.md routes trimming, joining, speed changes, fades, mixing, extracting a track from video and merging audio with video to the editing skill. Subtitle and speech-to-text jobs go to the video skill.

Does it work offline?

Only for audio metadata. SKILL.md marks probe-audio-metadata as Cloud or Local and the other three tools as Cloud. The README's feature table lists all four audio tools as Cloud, so check the tool's --schema for the mode.

Is it safe?

Installing the skill copies files only: the folder holds SKILL.md, a LICENSE and four reference files, with no scripts. Using it makes Claude run mediakit-cli in your shell, and Cloud tools send your audio or its URL to MediaKit's cloud and use your API key. Review each command and the mode before it runs. Cloud calls follow Volcengine's billing rules, which the README links.

Information

updated Nov 2, 2026

Folder
skills/byted-mediakit-audio

volcengine/mediakit-cliskills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 3
  1. byted-mediakit-shared SKILL.md · github.com
  2. volcengine/mediakit-cli README · github.com
  3. byted-mediakit-audio SKILL.md · github.com

Updated Nov 2, 2026.