--- title: Core Concepts description: How Preservated thinks about institutions, artifacts, assets, collections, metadata, and generated work. section: getting-started order: 4 updated: 2026-06-14 verified: 2026-06-14 related: [guides/working-with-artifacts, guides/collections-and-permissions, reference/supported-file-formats] features: [collections, artifact-metadata, multi-format-assets, ai-analysis] --- # Core Concepts You'll learn the mental model behind Preservated so the rest of the admin tools feel predictable. Preservated is built around a museum archive pattern: keep the original evidence safe, describe the intellectual object clearly, and treat machine-generated work as reviewable help rather than unquestioned truth. ## Institutions are the boundary An **institution** is your museum, archive, society, or collecting project. It owns the collections, artifacts, assets, members, settings, and public site under its URL slug. One Preservated account can belong to more than one institution. Your role is evaluated per institution, so you might be an admin for your museum and only a viewer for a partner organization. ## Artifacts are the things people describe An **artifact** is the intellectual object in the collection: a photograph, diary, poster, oral history, map, or born-digital file. Artifact metadata answers visitor and cataloging questions: - What is this called? - When was it created? - Who made it, appears in it, or is connected to it? - What subjects and places does it relate to? - What rights statement, license, or credit should travel with it? The artifact record is where curators work. It is also the stable page visitors cite, share, and revisit. ## Assets are the files behind artifacts An **asset** is a digital file attached to an artifact. A single artifact can have one asset or many: a TIFF master, a rendered PDF, page images, a video proxy, an audio file, a transcript, or a supplemental document. Preservated keeps masters unchanged. When it needs a thumbnail, web image, PDF preview, MP3, text layer, or deep-zoom tile, it creates a derivative and tracks how that derivative was made. :::tip When in doubt, describe the artifact and let Preservated manage the files. File-specific details such as dimensions, codec, page count, and extracted text live with assets automatically. ::: ## Collections are access-aware groups Collections group artifacts by donor, subject, exhibit, department, project, or any other structure your institution uses. They are not folders: one artifact can belong to multiple collections, and one membership can be marked primary for breadcrumbs. Collections also carry visibility. A collection can be public, members-only, or admin-only. Artifact visibility is derived from collection memberships and can be narrowed with an artifact-level override. For the full rule, see [Collections and Visibility](/docs/en-US/guides/collections-and-permissions). ## Metadata can be curated, imported, or generated Preservated tracks where information came from: - **Curated metadata** is confirmed by your staff. - **Imported metadata** comes from files, spreadsheets, external providers, or legacy systems. - **Generated metadata** comes from AI analysis, OCR, transcription, computer vision, or derivative processing. Generated work is useful, but it is not silently promoted into the curated record. You review suggestions, accept what helps, and keep the provenance trail for later inspection. ## Text layers make media searchable Text in Preservated can come from several places: a transcript of an oral history, OCR from a scanned letter, embedded text in a PDF, or captions from a video. These are stored as **text layers** on assets. Text layers can be raw AI output, reviewed AI output, or curated text. Because each layer is anchored to the source, search results can point to the right page, paragraph, timestamp, or region instead of only linking to the artifact as a whole. ## A typical record grows over time A common artifact lifecycle looks like this: 1. You import a file or external record. 2. Preservated stores the master and creates derivatives. 3. Technical and source metadata are attached. 4. AI tools suggest descriptions, dates, subjects, OCR, transcripts, or visual features. 5. Staff review the suggestions and curate the public record. 6. The artifact is placed in collections, published, shared, cited, downloaded, or used in tours and exhibits. The important idea is that Preservated does not force all of this to happen at once. You can start with a filename and a scan, then add authority, context, and interpretation as the collection matures.