What Drive makes from your files

An uploaded file goes through the same sequence every time: it is registered, scanned for malware, and then processed. Processing extracts the file's text and metadata, and for an image it also builds two smaller WebP copies. Nothing is produced for a file the scan blocked.

Where you see the progress

The file's Info panel lists the processing lifecycle, and each stage shows as a chip:

Chip	What it tracks
Registry	Whether the item exists in Drive.
Scan	The malware scan. Clean is the only state that lets processing continue.
Extraction	The parse of the file contents.
Metadata	The facts Drive read out of the file.
Preview	Whether a preview and extracted text exist.
Search	Whether the file is indexed for search.
Readiness	The rollup of the above.
Delivery	Whether the item is published, and its script and content-security policy.

Images

For an image Drive builds exactly two renditions, both WebP:

Rendition	Longest side	Purpose
Thumbnail	320 pixels	List and card thumbnails.
Preview	1600 pixels	Inline viewing.

Each rendition is capped at 8 MB. Renditions are built from a source under 25 MB, no larger than 16,384 pixels on a side and 40 million pixels in total, in GIF, JPEG, PNG or WebP. Anything outside those bounds keeps its original and no rendition.

Some image formats cannot be shown in a browser at all without a rendition. For those the viewer says so in place, for example TIFF originals require a derived browser image before inline rendering. or Camera RAW originals require a derived browser image before inline rendering. The extracted facts stay available under Image facts.

Renditions were activated in production on 2 September 2026.

Documents

Drive does not build page images for documents. It extracts their text and metadata, and the browser renders the document itself: a PDF in a browser viewer, a Word document paginated on the page, a spreadsheet as a table.

A scanned PDF has no text to extract. Drive says Extracted text is unavailable for this item. This PDF may require OCR. and offers Run OCR on the file's Metadata. OCR reports its own state, from OCR queued and OCR running through OCR complete, OCR partial or OCR failed, and its output appears as a related file labelled OCR text.

Run OCR is refused with a stated reason: File type not supported for OCR., Waiting for malware scan pass., Item readiness is blocked for processing. or You do not have permission to run OCR.

Video and audio

A video gets a poster frame, and speech is transcribed on request. See Open and preview a file for the transcript controls.

Archives

A zip, tar or tgz upload is unpacked so its contents become items. The chip reports Extraction queued, Extracting, Extracted, or Extracted with errors, and the item then offers View extracted files.

An oversized archive reports Too large to extract, and Drive explains: The archive is over the extraction size limit, so its contents were not unpacked. The archive itself uploaded successfully and can be downloaded.

Extracted text

The text Drive pulls out of a file is what powers search and retrieval, and what the Text view shows. Its state is reported on the file, and the messages are exact:

• Extracted text is still being prepared. Try again in a moment.
• Raw text isn't available for this file type.
• Text extraction failed for this item.
• Raw text is not available for this item yet.

A failed extraction leaves the file itself intact and downloadable. Use Refresh metadata on the file to ask for another pass.

Related

• Open and preview a file
• File will not open
• Accepted file types