Collecting material
Document Import Studio
Import a source file, review its extracted chunks, handle duplicates and choose what to keep.
Document Import Studio turns a source file into smaller pieces you can read, search and use in Glob. A piece is called a chunk. Review those chunks before making notes or tasks from them, especially when the source contains scanned pages or tables.
Open the list panel's conversation/tools view and choose Document Import Studio. Use Import in its header to select one file.
Supported files
| Format | Files you can select | What to check after extraction |
|---|---|---|
| Plain text | .txt, .text |
Text encoding and paragraph boundaries |
| Markdown | .md, .markdown |
Headings, tables and code blocks |
.pdf |
Page order, columns and any OCR text | |
| Tabular text | .csv, .tsv |
Headers, values and grouped rows |
| HTML | .html, .htm |
The readable text retained from the export |
| Word document | .docx |
Extracted text and structure |
| Images | .png, .jpg, .jpeg, .heic, .heif, .tif, .tiff, .bmp, .gif |
Recognised text and legibility |
PDFs use selectable text where available and can use optical character recognition, or OCR, for pages without a text layer. Image imports also use OCR. The Mac's file-import workflow does not provide the mobile camera scanner.
Import Studio accepts files, not a webpage address to fetch. Save an HTML or text export first. Audio and video are not supported document formats in this build.
Large imports
Some files show Large Import or Very Large Import before extraction. The warning shows the size and, for PDFs where readable, the page count. Choose Import to continue or Cancel to stop before processing.
The PDF warnings begin at 25 MB or 100 pages, and the stronger warning at 50 MB or 250 pages. Other files can show a size warning at 50 MB. These are warnings, not advertised maximum supported sizes or completion-time guarantees.
During extraction, a progress area shows the current phase. Its cancel button requests cancellation; the import stops when processing reaches a cancellation checkpoint. Scanned pages can take longer than selectable text.
Review before saving
When extraction succeeds, Review Import opens. Expand Import Options if needed, then inspect the chunk previews.
The included-count indicator tells you how many chunks will be kept. Possible duplicates are initially left out. Use All to include everything or Skip Duplicates to leave out chunks flagged as duplicates. You can also change individual selections.
A duplicate flag is a similarity warning against previously imported material. Repeated headings and boilerplate can produce a warning even when you need the chunk. Read it before excluding it.
Adjust chunk boundaries
Use Chunk size when the initial pieces are too short or too broad. The control allows 500–5,000 characters, in steps of 250. Choose Apply to rebuild the proposed chunks.
The initial review size is 1,200 characters for PDFs and 2,800 for other formats. Boundaries also depend on the extracted content, so these are targets rather than a promise that every chunk has exactly that length.
Applying a new size reruns duplicate detection and resets chunk selection. Review the new included count before saving. Rechunking uses the extracted text, so it can simplify some original structure. There is no saved-document rechunk control here; make the decision before Save Import.
Save or discard
Choose Save Import, or Save in a narrow panel, to keep the selected chunks. The document moves to the top of Library and becomes selected. Its detail view lists the kept chunks under Usable Items.
Save is disabled if no chunks are selected. If the review shows chunks but nothing is included, choose All or select individual pieces. If it shows no chunks at all, cancel and inspect the source file.
Cancel discards the pending review rather than adding it to the document library. The app also discards the retained copy made for that pending import.
A saved import retains a copy of the original file alongside the extracted material. This is part of the import on this Mac; it does not establish that the original has transferred to another device. Read Mac sync settings.
Read the result
Expand Library and select a document. In a wider panel the library and detail can sit side by side; in a narrower one the library folds above the detail.
Chunks can show page references, suggested tags, summaries, duplicate information and OCR quality. Check names, numbers and columns against the original when OCR quality is low or asks you to verify it. A summary is a condensed reading aid, not a corrected transcription.
Initial summaries and topic groups can be prepared locally. An eligible Apple Intelligence path can improve a selection of summaries afterwards. The file does not need a successful hosted answer merely to exist in the library.
Next, search the document or turn chunks into notes.
When extraction fails
Real import failures include “Could not read text from …”, “Could not open … as a PDF.”, “Could not open … as a DOCX file.”, and “No selectable text was found in …”. Check that the file opens in its usual reader and contains readable text or a legible scan. Export a clean copy and try that if the original is malformed.
“Unsupported document type:” means there is no importer for the file type. Renaming its extension does not convert the file. “Import cancelled.” means processing stopped rather than completing a saved import.