--- title: Supported File Formats description: Image, document, audio, video, 3D model, and metadata import formats Preservated accepts. section: reference order: 1 updated: 2026-07-16 verified: 2026-07-16 related: [guides/importing-photos, guides/transcription-and-ocr, guides/downloads-and-sharing, guides/browse-views] features: [multi-format-assets, document-rendering, import-sessions, 3d-model-assets, georeferenced-maps] --- # Supported File Formats A reference for what file types you can import and which derivatives Preservated generates for each. ## Metadata Import Files Metadata import files describe collection records, source URLs, or asset relationships rather than serving as preservation masters themselves. They are processed through import sessions, then any referenced media files or remote images are imported using the asset formats below. | Formats | Notes | | ------------------ | --------------------------------------------------------------------- | | CSV | Generic spreadsheets and catalog exports, including PastPerfect-style rows. Comma- and semicolon-delimited files are both detected automatically. | | TSV | Tab-delimited spreadsheet exports | | JSON | Arrays, provider records, and API response envelopes | | JSONL / NDJSON | One record per line; used by sources such as LOC Pictures, Washington State Digital Archives, and Cleveland Museum of Art imports | | XML | OAI-PMH records, PastPerfect PPWE/XML/DPLA exports, and other structured catalog exports | A plain-text `.txt` file is accepted as a metadata file only when it actually contains a delimited table with a header row; otherwise it's imported as a document. Provider mappings preserve source identifiers and unmapped fields as import provenance so records can be audited after migration. ## Uploading files together as a package When you upload files directly rather than pasting a source URL, you can select a metadata file and any number of asset files at once — see [Importing Photos and Documents](/docs/en-US/guides/importing-photos) for how they're matched to each other. A few limits apply to that upload: - **Per-file limits** depend on kind: images up to 200 MB, audio up to 500 MB, video up to 2 GB, documents up to 100 MB, 3D models up to 500 MB, and other file kinds up to 50 MB. An institution admin can raise or lower these limits. - **TIFF and GeoTIFF files get a higher default limit — 2 GB** — instead of the general 200 MB image limit, because archival scans and scanned maps commonly exceed it. This applies to every `.tif`/`.tiff` file regardless of whether it carries embedded geographic metadata; there's no separate "GeoTIFF" upload path. Platform staff can raise an institution's TIFF limit further for large-format map digitization. - **Package limits**: up to 2,000 files and 5 GB total per upload. Split a larger migration into more than one import session. - System files your operating system adds automatically (`.DS_Store`, `Thumbs.db`, `desktop.ini`, and similar) are detected and skipped — you don't need to remove them first. :::note Files larger than about 200 MB upload in parts rather than as one transfer. If your connection drops partway through, retry the upload — completed parts aren't re-sent, so a large file doesn't have to start over from zero. ::: ## Images | Formats | Notes | | -------------------------------- | ----------------------------------------------------------- | | JPEG, PNG, WebP, GIF | Imported as-is | | TIFF | Common for archival scans and scanned maps; large files supported (see the upload limits above) | | HEIC/HEIF | Converted automatically for web display | Generated for every image: thumbnails, web-size derivatives, and deep-zoom (IIIF) tiles so visitors can pan and zoom high-resolution scans without downloading the full master. ### Georeferenced maps When a TIFF carries embedded georeferencing tags (a GeoTIFF), Preservated extracts that geographic information automatically in the background after upload — no special import step is needed, and the image itself works exactly like any other image while extraction runs. A clean result activates automatically; a map whose coordinate system needs confirmation or whose geometry is unusual (crossing the antimeridian, or spanning a large enough area that the precision can't be confirmed automatically) is held as a draft for staff review, and a file with missing or corrupt georeferencing tags gets a clear failed state explaining why. An ordinary TIFF with no georeferencing tags simply imports as an ordinary image. Once a map is active, its artifact page offers an **Image | Map** view — see [Public Discovery](/docs/en-US/guides/public-discovery#viewing-a-georeferenced-map). ## Documents | Formats | Notes | | ---------------------------------------- | --------------------------------------------------------------------------- | | PDF | Page images and embedded text are extracted; scanned PDFs can be OCRed | | Word — .docx, .doc, .rtf, .odt | Rendered to a high-fidelity page preview; text extracted for search | | PowerPoint — .pptx, .ppt, .odp | Rendered to a high-fidelity page preview; text extracted for search | | Excel — .xlsx, .xls, .ods | Rendered to a high-fidelity page preview; text extracted for search | Multi-page documents get per-page navigation in the viewer, and OCR text becomes a searchable, downloadable text layer. In browse and grid views, hovering (or focusing with the keyboard) a multi-page document's thumbnail flips through its first few pages — see [Browse views](/docs/en-US/guides/browse-views#preview-a-documents-pages-without-opening-it). Word, PowerPoint, Excel, and OpenDocument files are converted once to a PDF and shown with the same page-by-page viewer as a scanned PDF, so visitors read the document as it was laid out — not just its extracted text. Your original file is always preserved unchanged and stays downloadable alongside the rendered PDF. Plain-text, Markdown, and HTML files keep a text-only preview. :::note High-fidelity document rendering requires the document-conversion service to be configured for your deployment. When it isn't, these files still import and show their extracted text and a download link — they just don't get the page preview. Spreadsheets that render to an unusually large number of pages fall back to the text preview. ::: ## Audio | Formats | Notes | | -------------------- | ---------------------------------------------- | | MP3, WAV, M4A, FLAC | Waveform preview generated; playable in the browser | Audio can be transcribed with AI; transcripts are timestamped, searchable, and exportable. ## Video | Formats | Notes | | -------------------- | --------------------------------------------------------- | | MP4, MOV, AVI, WebM | A web-friendly proxy and poster frame are generated | Large or unusual codecs are transcoded to a streamable proxy; the original master is preserved unchanged. Videos can be transcribed like audio. ## 3D Models | Formats | Notes | | -------------------- | -------------------------------------------------------------------------------------- | | GLB (Binary glTF) | Self-contained `.glb` files up to 500 MB; a preview image and an interactive in-browser 3D viewer are generated | A GLB file must be **self-contained**: every texture and mesh buffer embedded in the one file. Files that reference other files or web URLs are rejected with a clear reason, as are files that need rendering features the viewer doesn't support. Compressed geometry and textures (Draco, KTX2/Basis, meshopt) are fully supported. After upload, Preservated validates the file, records its technical details (triangle and vertex counts, materials, textures, animations, and model-space dimensions), and renders a poster image that represents the model in grids, search results, and sharing previews. How the interactive viewer behaves depends on the model's size: - **Up to 50 MB** — the 3D view loads automatically, showing the poster while the model streams in. - **50–200 MB** — visitors see the poster with a "Load 3D view" button that states the download size, so nobody pays for a large download without choosing it. - **Over 200 MB or extremely detailed models** — the poster and technical details are shown, with the model available from the Download page. The viewer supports orbit, pan, and zoom with mouse, touch, or keyboard, plus reset-view and fullscreen controls. The model never moves on its own. Visitors whose browsers can't render 3D still see the poster, description, and download option. :::note Model-space dimensions describe the file's own coordinate system — they are not physical measurements of the object. Record real-world dimensions in the artifact's metadata. ::: Multi-file 3D formats (`.gltf` with separate textures, OBJ, STL, PLY) aren't accepted yet — export a self-contained GLB from your 3D software instead. Most tools (Blender, RealityCapture, Metashape) offer this directly. ## Masters are never modified Whatever you import is stored byte-for-byte as the preservation master. All web display uses generated derivatives, and every derivative records how it was made. :::note If you have material in a format not listed here, convert it to the closest listed format before importing — for example, export camera RAW files to TIFF. Unsupported files fail the import with a clear error rather than importing incorrectly. :::