will downloading Youtube video metadata through Google Takeout be a huge archive file? What does the metadata consist of, exactly?
will downloading Youtube video metadata through Google Takeout be a huge archive file? What does the metadata consist of, exactly?
No, downloading only your YouTube video metadata will not be a huge file. Because metadata is strictly text-based, the archive will be a lightweight file. It will only become a massive archive if you choose to download the actual, original video files alongside the metadata.
What the Metadata Consists Of
When you use Google Takeout to export your YouTube data, the metadata is typically provided as CSV or JSON files. It includes a comprehensive set of information about your videos, such as:
Identification: Video ID, channel ID, upload date, and time.
Content: Video title, full description (including timestamps and hashtags), and tags.
Technical details: Video duration, dimension (2D/3D), definition (HD/SD), and content ratings.
Links & Media: URLs of the video thumbnails in various sizes (default, high, maxres).
Note: YouTube currently limits standalone CSV metadata exports to approximately 200 videos per file.
FollowUp
If you'd like, let me know:
-
How many videos are in your channel?
-
Are you planning to download the actual video files too?
I can help you estimate the exact total file size and time required for your backup.
I only have one video in my own channel. I selected Youtube History, so I imagine that's all the videos I've watched - is that only the list of titles + creator + URL, or does selecting metadata create a condition to pull the metadata of my Youtube history?
No, selecting "YouTube History" will not pull the deep metadata of all the videos you have watched. Google Takeout does not trigger a script to fetch the descriptions, tags, and definitions of other creators' videos for your archive.
Instead, it strictly exports your own account's activity log.
What Your YouTube Watch History Actually Contains
If you choose the JSON format (which is highly recommended over HTML for data readability), each video entry in your watch-history.json file contains only the following basic data points:
Title: The name of the video at the time you watched it (prefixed with "Watched...").
Title URL: The direct link to the video (containing the unique video ID).
Time: The exact timestamp (date and time) of when you played the video.
Subtitles (Creator Info): The name of the channel that uploaded the video, along with that channel's URL.
Products: Listed simply as [YouTube].
What is Omitted
You will not receive:
Video descriptions or hashtags.
Video categories, tags, or dimensions.
Video duration (the JSON file will not tell you if you watched a 5-minute clip or a 3-hour livestream).
View counts, likes, or subscriber numbers of those channels.
Because it is purely a text-based list of your personal history, the file size will remain very small (usually just a few megabytes, even if you have watched tens of thousands of videos over several years).
FollowUp
If you are trying to parse this history data, let me know:
What is your main goal for extracting your watch history? (e.g., finding an old deleted video, calculating your total watch time, or importing it into a visualization tool?).
I can guide you on the best tools or methods to extract deeper data from those URLs.
Feeding a curated record of my media consumption history into this project as a reference: https://gitingest.com/NateBJones-Projects/OB1
Uploaded files:
https://gitingest.com/NateBJones-Projects/OB1
Using your YouTube watch history as a reference for NateBJones-Projects/OB1 is a smart way to bootstrap your personal knowledge base. NateBJones-Projects/OB1 is specifically designed to store your raw context, generate vector embeddings, and abstract metadata separately in a Postgres/Supabase database.
However, because the standard Google Takeout watch-history.json only contains bare-bones text data (just titles, dates, and basic URLs), feeding it directly into OB1 via Gitingest or direct ingestion will only allow your LLM or autonomous agents to query what you watched and when. The AI will not implicitly know the underlying topics, code examples, or tutorials discussed in those videos.
How to Maximize Your Watch History for OB1
To turn your watch history into a rich external brain reference, you need to enrich the bare URLs into full context. Here are the best ways to prepare that data before feeding it into your system:
Use a Multi-Step Ingestion Pipeline:
Instead of dumping the raw JSON file, write a lightweight script or use an OB1 integration to parse the URLs.
Scrap Video Metadata Externally:
Pass the video IDs from your Takeout history through tools like yt-dlp or the YouTube Data API to automatically extract the full video descriptions, tags, and categories.
Pull the Subtitles/Transcripts:
This is the ultimate level of utility. By fetching the video transcripts (using tools like youtube-transcript-api), you can feed the literal spoken content of your entire media history into OB1. Your AI agents will then be able to semantically search deep code concepts or workflow advice you heard months ago.
Leveraging OB1's Architecture
Because OB1 decouples your raw data from its vector embeddings, any watch history records you stream into your Supabase database can be cleanly re-indexed over time. If you pipe the data in via an a href="https://github.com/NateBJones-Projects/OB1" Model Context Protocol (MCP) server, any connected LLM (like Claude or ChatGPT) can query your history dynamically to recall past context while you work.
FollowUp
If you want to move forward with this, let me know:
- Do you want help writing a quick Python script to clean your Takeout file and fetch the video descriptions/transcripts?
- Are you deploying OB1 using the standard Supabase cloud stack or a self-hosted Docker/Kubernetes setup?
I can provide the exact steps to format your history so it passes OB1's structured memory rules flawlessly.