Ingestion

The moment a file enters the library is the cheapest moment to do this work. Steps that are skipped at ingestion are rarely done later, which is how libraries fill up with untagged files that nobody can find or preview.

Ways files enter a DAM

Route	Typical use
Upload in the browser	A marketing team adds the photos from a new shoot or the files of a campaign
Bulk import from storage	Moving an archive from a shared drive or cloud storage, or a regular delivery from a photo studio
Import with a product catalogue	Product photos matched to rows of a spreadsheet or catalogue export, which supply product names, SKUs and categories
API	A product information system, a web-to-print tool or an agency's system sends files automatically

Steps of an ingestion pipeline

1. Validation and virus scan. The system checks that the file arrived complete and scans it before anyone can download it.
2. Format identification. The real file type is read from the content, not taken from the extension.
3. Duplicate check. A checksum finds exact copies. A perceptual hash finds the same image resized, recompressed or lightly edited. See duplicate detection.
4. Metadata extraction. Technical properties are read from the file, together with Exif and IPTC fields written by cameras and photographers. Filenames that follow a convention can be parsed into fields.
5. Previews and renditions. Thumbnails, web-size previews, video previews and page images of documents. See rendition.
6. Enrichment. AI tags and captions, text read from images by OCR, transcripts of speech, and the data that semantic search and visual search need.
7. Rules and indexing. Default values, automation rules and access rules are applied, and the asset becomes searchable.

A well-built pipeline runs these steps in the background, so an import of thousands of files does not slow down people who are using the library.

Ingestion and migration

Moving an existing archive into a DAM is ingestion at scale: folder names become metadata, files are matched to catalogue records, and the pipeline enriches a backlog that nobody would tag by hand. Some practical points:

• Migrate in slices, such as one brand, one year or one product line, and check what the pipeline produced before starting the next. Teams that copy everything first and plan to organise it later rarely do.
• Map folder names and filename conventions to fields before importing, so the information they hold is not lost.
• Decide how to handle duplicates before they arrive: skip the copy, keep both, or replace the older file.
• Keep the source read-only until the import has been checked.

The guides on migrating from cloud storage to a DAM and on DAM implementation cover the full process.

Ingestion in Freedam

• Files arrive by browser upload (large files are sent in chunks), through the API, or from import sources that read S3-compatible storage, WebDAV servers and cloud drives, on a schedule if needed. See the asset library.
• Import sources can read a product catalogue spreadsheet, match its rows to files and fill metadata fields and vocabulary terms from its columns.
• Every file is scanned for viruses before it is stored, and an infected file is rejected.
• Exact copies are found by checksum and near-duplicates by perceptual hash. The upload page lets you compare each pair, then delete the upload, keep both, or replace the existing file while keeping its metadata. See duplicate detection.
• Exif and IPTC fields are read into the asset's fields, and filename parsing rules turn naming conventions into field values.
• Previews are generated for images, video, audio, PDFs, Office documents, EPS files, design files and archives, and the text of documents is extracted for search. See rendition.
• Images are analysed by AI for tags, a caption, printed text and the asset class; speech in video and audio is transcribed; Content Credentials are verified. See auto-tagging.
• Automation rules run on each new asset to set metadata, add tags, assign collections, change the workflow status or notify people.
• Each stage runs as a background job, and each upload batch has its own page listing its files, their status and the duplicates found.

Frequently asked questions

What is the difference between upload and ingestion?

Upload is the transfer of the file to the system. Ingestion includes the upload and everything that follows until the file is a finished asset: scanning, metadata extraction, duplicate checks, previews and enrichment.

What is the difference between data ingestion and asset ingestion?

Data ingestion is a data engineering term for loading records from applications and databases into a data warehouse or data lake. Asset ingestion, the meaning used here, is loading files such as images, videos and documents into a DAM, with the processing that makes them findable.

Should files be tagged before or after ingestion?

Let the pipeline do the first pass: embedded metadata, AI tags and values from filenames or a catalogue cost nothing to add. Then add what only people know, such as campaign, approval status and usage rights, ideally in the same session as the upload, often with a bulk edit across the whole batch.