LibreAI / 切り抜き動画 / youtube-shorts-pipeline
// clipping · rushindrasinha/youtube-shorts-pipeline

youtube-shorts-pipeline

youtube-shorts-pipelineは、rushindrasinhaが公開しているオープンソースの切り抜き動画プロジェクトです(GitHub: rushindrasinha/youtube-shorts-pipeline)。GitHubスターは2,312、フォークは536で、MITライセンスの下で公開されています。GitHubでの最終更新は2026年6月9日です。

Verticals: one command turns a topic into a researched, voiced, captioned, and uploaded YouTube Short.

★ 2.3k スター⑂ 536 フォークMIT更新 2026年6月
▦ ビジュアル — このプロジェクトのREADMEには埋め込み可能なスクリーンショットがないため、
ここではプレースホルダーを表示しています。下のガイド自体は完全な内容です。
// プロジェクトの基本情報

プロジェクト基本情報

GitHubスター
2,312
フォーク
536
ライセンス
MIT
GitHubでの最終更新
2026年6月9日
カテゴリ
切り抜き動画
掲載タイプ
完全なインストールガイド
公式サイト
Official site
// 仕組み

仕組み

  1. Clarify your goal

    Pick a niche (15 ship built in, from tech and finance to cooking and true crime) and a topic, or let topic discovery suggest one.

  2. Get the project

    Clone the repository and install the Python requirements; make sure ffmpeg is available on your system.

  3. Install and configure

    Configure provider keys for the stages you use, for example an LLM key for scripts and a Gemini key for visuals; free options exist for several stages.

  4. Run your first task

    Run the full pipeline for a topic, or work stage by stage with draft, produce, and upload commands.

// 機能

選ばれる理由

Niche intelligence profiles

A YAML niche profile shapes script tone, hooks, visual style, caption look, music mood, and thumbnail strategy at every stage, and you can author your own.

Research-grounded scripts

Research uses DuckDuckGo plus optional source scraping, and the script stage is instructed to use only researched facts as an anti-hallucination gate.

Multiple LLM and voice providers

Scripts via Claude, Gemini, GPT, Ollama, MiniMax, or LiteLLM; voiceover via Edge TTS (free), ElevenLabs, MiniMax, 60db, or macOS say.

Whisper captions in two formats

Word-level timestamps produce burned-in ASS captions with highlighting and an SRT file uploaded as closed captions.

ffmpeg assembly with music ducking

B-roll gets Ken Burns motion, is cropped to 9:16, and is assembled with voiceover and mood-matched music that ducks under speech.

Private-by-default YouTube upload

Uploads publish with generated title, description, tags, captions, and thumbnail, kept private by default, with resumable stages and a dry-run mode.

Advertisement
// 自分で実行

自分で実行してみる

  1. Clone and install

    Install the pipeline's Python dependencies after cloning.

    git clone https://github.com/rushindrasinha/youtube-shorts-pipeline.git
    cd youtube-shorts-pipeline
    pip install -r requirements.txt
  2. Configure providers

    Add the API keys for the providers you chose to the local config (see README provider tables).

    # set ANTHROPIC_API_KEY / GEMINI_API_KEY etc. per the README configuration table
  3. Run the pipeline

    Produce a complete Short from a topic in one command.

    python -m verticals run --topic "your topic" --niche tech
// 活用例

作れるもの

Advertisement
// よくある質問

よくある質問

Who maintains it?

It is maintained by rushindrasinha, with community-contributed providers noted in the release notes.

Is it free to use?

The code is MIT-licensed and free. Running it can cost small per-video API fees depending on providers; the README documents a zero-cost mode using Ollama and free Edge TTS.

Does it need a paid API key?

Only for the providers you pick. Claude, Gemini, or ElevenLabs need their keys, while the local Ollama plus Edge TTS path needs no paid key at all.

How does it compare to a well-known alternative?

General AI video generators make clips; Verticals is an end-to-end publishing pipeline that researches, scripts, voices, captions, and uploads a finished Short tuned to a niche.

How do I get started?

Clone the repo, pip install -r requirements.txt, set your provider keys, and run python -m verticals run with a topic and niche.

// さらに探索

似ているツール

切り抜き動画

MoneyPrinter

MoneyPrinter — automate YouTube Shorts from a single topic: local Ollama writes the script and metadata, stock footage and voiceover are assembled into a finished short, MIT

★ 14.1k スターMIT
ガイドを見る →
切り抜き動画

clips-studio

Turn Twitch, Kick, and YouTube streams into captioned Shorts locally.

★ 85 スターAGPL-3.0
ガイドを見る →
切り抜き動画

openfang-auto-clip

Automated clipping pipeline that turns long videos into reusable shorts.

★ 38 スターMIT
ガイドを見る →
切り抜き動画

Autoshorts

AI engine that finds viewer-grabbing moments and turns them into shorts.

★ 91 スターCustom
ガイドを見る →
切り抜き動画

ViralMint

Local-first video pipeline: trend scouting, clipping, captioned shorts.

★ 122 スターAGPL-3.0
ガイドを見る →
切り抜き動画

shortcast

Native macOS app: drop a long video → auto-cut viral shorts, captioned for TikTok/Reels/YouTube, reframed to vertical, and scheduled. 100% on-device (Gemma 4…

★ 389 スターApache-2.0
ガイドを見る →
// 代替候補

代替候補: 切り抜き動画

LibreAIカタログに掲載された切り抜き動画プロジェクトの中で最もスターの多いもの — youtube-shorts-pipelineが合わなかった場合に検討したい、現実的な候補リストです。

// さらに見る clipping

さらに見る: 切り抜き動画

カタログの中でyoutube-shorts-pipelineとスター数が近い切り抜き動画プロジェクト — 前後の棚もそのまま見て回れます。

youtube-shorts-pipelineを入手する

無料でオープン — ソースコードを入手して、リポジトリにスターを付け、作り始めましょう。