When you find a deleted YouTube video through Tube Search, you typically get metadata: a title, description, upload date, and sometimes subtitles. That is already useful. But reading through raw subtitles to understand what a video was about takes time, especially for longer recordings.

We have added AI-powered video summaries to Tube Search. If a video has archived subtitles, the system can now generate a structured summary of its content in seconds.

What You Get

Each summary includes four components:

TL;DR. A single paragraph that captures the essence of the video. Enough to decide whether the full content is worth digging into.

Key moments with timestamps. The most important points extracted from the video, each linked to a specific timecode. This is particularly helpful for long recordings where you need to locate a specific segment.

Detailed breakdown. A longer narrative summary that walks through the video's content from start to finish. For a 30-minute lecture or interview, this can save you the trouble of reading the entire subtitle file.

Automatic tags. Topic tags generated from the content, making it easier to find related videos later.

How It Works

The process is straightforward. Find a video with available subtitles through any search method (channel lookup, video ID, or full-text search), then click the "Summarize" link in the results. The system feeds the subtitle text to a language model, which produces a structured summary. The whole process takes between 30 seconds and two minutes depending on the video length.

Summaries are generated in the language of the subtitles. If a video has English subtitles, you get an English summary. Russian subtitles produce a Russian summary. This happens automatically without any configuration.

Why This Matters for Deleted Videos

The real value of AI summaries becomes clear when dealing with deleted content. A deleted video has no player, no comments, no chapter markers. If Archive.org preserved the subtitles, you can download them, but reading a raw SRT file of a two-hour podcast is not a productive use of anyone's time.

A structured summary turns those raw subtitles into something immediately useful. Researchers can quickly determine the relevance of a deleted video. Journalists can identify which segments of a removed recording contain the statements they are looking for. Anyone doing fact-checking can get the gist of a video that no longer exists on YouTube without manually parsing subtitle files.

Summaries Are Persistent

Once generated, a summary is saved permanently. Other users searching for the same video will see the summary instantly without waiting for regeneration. Over time, this builds a growing library of summarized content, including videos that may be deleted in the future.

MCP Server for AI Assistants

We are also working on an MCP (Model Context Protocol) server that will allow AI assistants like Claude, ChatGPT, and others to interact with Tube Search directly. Through the MCP server, an AI assistant will be able to search for videos, retrieve metadata, read subtitles, and access summaries as part of a conversation, without the user needing to switch between the chat interface and the browser.

The MCP server will be available both as a remote endpoint and as a local npm package for developers who want to integrate Tube Search capabilities into their own tools and workflows.

Access

AI summaries are available to registered users with a Plus subscription or higher. Reading existing summaries is free for everyone. Detailed information about available tiers is on the pricing page.

Try it: tube.archivarix.net

The use of article materials is allowed only if the link to the source is posted: https://archivarix.com/en/blog/ai-video-summaries/