TranscribeYoutube

TranscribeYoutube

Fetches the complete transcript of any YouTube video and generates an Obsidian-ready .md note — directly from VS Code, with zero external dependencies.

Uses the InnerTube Player API (iOS client), the same API used internally by the obsidian-yt-transcript plugin. No yt-dlp, no API keys, no installs beyond Python 3.9+.

When to Use This Skill

• User asks to transcribe a YouTube video
• User wants to capture video content into their Second Brain / Obsidian vault
• User mentions a YouTube URL and wants a transcript note generated
• Trigger phrases: TranscribeYoutube, transcript, YouTube transcript, capture video, extraer subtítulos, transcribir vídeo

Requirements

• Python 3.9+ — standard library only, no pip installs needed
• Internet connection
• The video must have subtitles available (automatic or manual) on YouTube
• Obsidian installed with the vault name configured (see Configuration below)

Configuration

Set the OBSIDIAN_VAULT environment variable to your vault's absolute path:

# macOS/Linux
export OBSIDIAN_VAULT="/path/to/your/Obsidian/VaultName"

# Windows (PowerShell)
$env:OBSIDIAN_VAULT = "C:\Users\you\Documents\Obsidian\VaultName"

Also set OBSIDIAN_VAULT_NAME if your vault name differs from the folder name:

export OBSIDIAN_VAULT_NAME="MyVault"

If not set, the script uses the defaults defined at the top of transcribe_youtube.py.

Command

python3 TranscribeYoutube/scripts/transcribe_youtube.py <YOUTUBE_URL>

Optional language preference (default: tries es then en):

python3 TranscribeYoutube/scripts/transcribe_youtube.py <URL> --lang en

Full Workflow

1. Run the script with the YouTube URL
2. The script fetches metadata + transcript via InnerTube API and writes the .md file
3. Obsidian opens automatically on the new note
4. Copy the wikilink printed in the terminal into the main resource note under the video iframe:

   Transcript: [[Transcripciones/VIDEO_ID - Title]]
   

5. Optionally, ask Copilot to generate a summary from the transcript and add it to the resource note

Output Format

File saved at:

<VAULT>/Atlas/Recursos/Transcripciones/VIDEO_ID - Title.md

Structure:

---
tags: [transcripcion, recurso]
video-id: <ID>
url: https://www.youtube.com/watch?v=<ID>
canal: "<Channel>"
titulo: "<Title>"
idioma: es
fecha-guardado: YYYY-MM-DD
---

# Title
**Canal:** ...
**URL:** ...

← Main note: `[[...]]`  *(update this wikilink)*

---

## Transcripción

[00:00](https://...&t=0) First grouped text block...
[00:30](https://...&t=30) Next block...

Lines are grouped every 30 seconds (same grouping as the YTranscript plugin). Timestamps are clickable and open the video at that exact second.

File Naming Convention

VIDEO_ID - Title (max 60 chars).md

Examples:

• 7GqL6xUvDuM - Imagen a VECTOR AUTOMÁTICAMENTE Affinity 3.0.md
• dQw4w9WgXcQ - Never Gonna Give You Up.md

Cross-Platform Support

Notes

• No external dependencies — pure Python 3.9+ standard library
• Uses the same InnerTube API (iOS client) as the YTranscript Obsidian plugin
• If a video has no subtitles, the script exits with a clear error message
• If the note already exists, it is overwritten (useful to refresh transcripts)
• The ← Main note: line must be updated manually after creating the resource note

Reference Files

• scripts/transcribe_youtube.py — main script