一键在 Manus 中运行任何 Skill

media-ingest

星标4

分支3

更新时间2026年7月10日 06:35

Ingest media from a URL (YouTube + most sites) with `yt-dlp`/`ffmpeg` — transcript, audio, video, subtitles, formats. NOT for converting a local file to Markdown (use markdown-converter) or writing video scripts (use create).

安装

用 Codex 或 Claude 帮你安装复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。

在 Manus 中运行

来源

kronael

kronael/tools

打开 GitHub 仓库查看创作者相关仓库

下载

在 Manus 中运行

Media Ingest

Pull media from a URL — YouTube and nearly every site yt-dlp supports. Tools: yt-dlp (+ ffmpeg for merge/extract), both local CLIs; the fetch itself needs network.

Transcript — yt-dlp --write-auto-subs --sub-lang en --skip-download --sub-format vtt. ALWAYS flatten the .vtt yourself: auto-caption VTT is ROLLING — each cue repeats the tail of the previous one, so a naive concatenation duplicates most lines. Drop the cue-timestamp lines and [Music]/[Applause] bracket cues, collapse consecutive repeats, join to a paragraph. Emit timestamps ONLY when asked.
Human-authored subs are cleaner than auto — ALWAYS try --write-subs first, fall back to --write-auto-subs only when none exist.
Audio — yt-dlp -x --audio-format opus (or mp3). Video — -F to list formats, then -f <id> and --remux-video mp4 to repackage WITHOUT re-encoding.
NEVER reach for a scraping library or the raw YouTube Data API — yt-dlp is the maintained path and already handles formats, playlists, and sign-in walls.

同仓库更多 Skills

同仓库

global

kronael/tools

Development wisdom and workflow rules. NOT for project-specific conventions (those live in CLAUDE.md, edit via wisdom).

2026-07-184

bugs

kronael/tools

The `BUGS.md` open-issues queue — entry format, lifecycle, pruning to diary. NOT for the record-don't-fix policy (that's CLAUDE.md Bug Triage Protocol), NOT for resolved-bug history (use /diary), NOT for feature backlog (use TODO.md/specs).

2026-07-184

demo

kronael/tools

Terminal demo GIF + MP4 recordings for READMEs and social (asciinema + agg + ffmpeg). NOT for general Makefile targets (use software) or GUI screenshots (use visual).

2026-07-184

kronael/tools

Go development. NOT for non-Go code (use rs, py, ts, tsx, or sh).

2026-07-184

humanize

kronael/tools

Humanize text: strip AI-isms and add real voice. NOT for drafting new copy (use writing).

2026-07-184

kronael/tools

Python development. NOT for shell scripting (use sh) or non-Python code.

2026-07-184

name	media-ingest
description	Ingest media from a URL (YouTube + most sites) with `yt-dlp`/`ffmpeg` — transcript, audio, video, subtitles, formats. NOT for converting a local file to Markdown (use markdown-converter) or writing video scripts (use create).
when_to_use	download a youtube video, download/rip audio from a video, mp3/opus from youtube, get the transcript of a video, youtube transcript, extract subtitles, list video formats, download a playlist, save a podcast episode, yt-dlp
user-invocable	false

Media Ingest

Pull media from a URL — YouTube and nearly every site yt-dlp supports. Tools: yt-dlp (+ ffmpeg for merge/extract), both local CLIs; the fetch itself needs network.

Transcript — yt-dlp --write-auto-subs --sub-lang en --skip-download --sub-format vtt. ALWAYS flatten the .vtt yourself: auto-caption VTT is ROLLING — each cue repeats the tail of the previous one, so a naive concatenation duplicates most lines. Drop the cue-timestamp lines and [Music]/[Applause] bracket cues, collapse consecutive repeats, join to a paragraph. Emit timestamps ONLY when asked.
Human-authored subs are cleaner than auto — ALWAYS try --write-subs first, fall back to --write-auto-subs only when none exist.
Audio — yt-dlp -x --audio-format opus (or mp3). Video — -F to list formats, then -f <id> and --remux-video mp4 to repackage WITHOUT re-encoding.
NEVER reach for a scraping library or the raw YouTube Data API — yt-dlp is the maintained path and already handles formats, playlists, and sign-in walls.