# What Archilyzer is

Archilyzer takes a channel's back catalogue, downloads it, transcribes it on your
own hardware, and builds a static website you host yourself. The result is a
permanent, searchable record of what someone said on video — searchable to the
second, and still there after the original comes down.

It is a program you run, not a service you sign up for. There is no account, no
API key, no server of ours in the path, and nothing phones home.

## The shape of it

Three programs share one library and one pile of data.

- **The editor** is a local admin app. You add channels, watch the download and
  transcription queues, and press the button that builds and deploys a site. It
  runs on your machine and is never exposed to the public.
- **The corpus** is a directory on disk — one folder per channel, holding the
  downloaded media, the transcripts, and a search index. It is deliberately kept
  outside the code: a fresh copy of the software has no data, and updating the
  software never touches your archive.
- **The export** is the public artefact: a static site, pre-rendered to plain
  HTML and JSON. No database, no runtime, no server-side code. It can be hosted
  on anything that serves files.

A fourth piece, the **project site** you are reading, and a fifth, the **MCP
server** described in [AI and MCP](/docs/ai-and-mcp/), round out the workspace.

## How a video becomes a searchable page

1. **List.** Ask the platform what a channel has published. Archilyzer keeps
   this list separately from what it has already fetched, so it always knows the
   difference between "new" and "already have it".
2. **Fetch.** Download what is missing. For channels where the platform already
   publishes captions, those are taken directly. For everything else, only the
   audio is fetched.
3. **Transcribe.** Audio is transcribed locally, by whisper.cpp or one of the
   other supported backends. This is the slow step and the one that wants a GPU;
   nothing is sent to a third-party transcription service.
4. **Index.** Transcripts are cut into pages and written as a paginated JSON
   index, so a browser can search a corpus of tens of thousands of videos
   without downloading it.
5. **Compose and publish.** A static site is built from that index and deployed.

Steps 1–3 can run unattended on a schedule. See
[Running an archive](/docs/operate/).

## What you get at the end

- **Search across every transcript**, with results that jump to the exact second
  of the recording they came from.
- **A player** that follows the transcript, and a transcript that follows the
  player.
- **Multiple sites from one corpus.** Channels are grouped into sites, so one
  archive can publish several public faces without duplicating the data.
- **Bulk downloads** of the transcripts, for anyone who wants the raw material.
- **A machine-readable index** at `/corpus.json`, so an LLM or a script can
  navigate the whole archive without scraping it.

## What it is not

- Not a hosted service. You install it, you run it, you pay for your own hosting.
- Not a forge. The source is a [read-only git mirror](/source/) on this site —
  clone and pull, but nothing takes a push or a pull request.
- Not a downloader you point at one video. It is built around back catalogues:
  thousands of recordings, kept current.
- Not automatic transcription in the cloud. The transcribing happens on your
  machine, at your machine's speed.

Next: [Install](/docs/install/).