Everyday text, recognized fast
Clear text goes to Apple Vision first for on-device recognition and automatic language detection—ideal for quick selections and instant copying.
Textract OCR
Select it.Text in hand.
Select anything on screen to recognize, copy, or translate it instantly.Photos, PDFs, and complex layouts become editable Markdown, too.
Content stays on your MacOn-device translationComplex layouts supported
Recognition and translation happen on your Mac. The source is never uploaded.
Document · Research Notes.pdf
本地模型会分析版面结构，
再输出清晰的 Markdown。
表格、公式和手写内容全程留在本机。
Recognition complete · Copied
本地模型会分析版面结构，再输出清晰的 Markdown。表格、公式和手写内容全程留在本机。
Translation complete · Chinese → English
The local model analyzes the layout before producing clean Markdown. Tables, equations, and handwriting stay on this Mac.
Deep recognition
Local model · Complex layout
- 1Select
- 2Recognize and copy
- 3Translate
On-device capabilities, each with a role
The right way to handle every kind of content
Apple Vision handles everyday text, GLM-OCR handles complex layouts, and Apple Translation handles content across languages. All three run on your Mac.
Complex layouts, deeply parsed
For tables, equations, handwriting, or layered layouts, the installed GLM-OCR model can continue processing the complex structure on your Mac.
Recognize, then translate
When the source differs from your target language, Apple Translation generates the translation on your Mac. Neither source nor translation needs to be uploaded.
Photos and PDFs to Markdown
Turn photos & PDFs into Markdown
Drop in handwritten notes, scans, or multi-page PDFs. After batch processing, headings, lists, and tables stay clear and ready to edit.
Batch processingStructure preservedEntirely on-device
# Experiment Notes
## Layout analysis
The model extracts blocks first, then rebuilds heading hierarchy and lists.
| Content | Output |
| --- | --- |
| Tables | Markdown tables |
| Headings | Hierarchy preserved |
| --- | --- |
| Tables | Markdown tables |
| Headings | Hierarchy preserved |
Complex layout recognition
Complex layouts, clearly recognized
Clear text is recognized quickly. For tables, equations, handwriting, or layered layouts, Smart mode calls the local model for deeper processing.
Three recognition modesTables · Equations · HandwritingEntirely on-device
FastClear text · Fast results
Everyday
SmartSystem recognition + Local model
Recommended
DeepTables · Equations · Handwriting
Complex content
Translate on selection
Copy the text. See the translation.
After you select, the source is copied as usual. When a different language is detected, your Mac generates the translation locally. Copy only the translation or take both.
Automatic language detectionOn-device translationSource not uploaded
Source
Select any region on screen — text becomes editable in one shortcut.
Translation · Generated on-device
框选屏幕上的任意内容，按一次快捷键，就能得到可编辑的文字。
Done in three steps
Select, copy, and translate in one flow
No tool switching. No interruption.
-
Select what you need
Press the shortcut and drag a selection. Textract reads only the area you choose, never your entire screen.
-
Recognize and copy automatically
OCR runs on your Mac and sends the result straight to the clipboard. Languages and QR codes are recognized in the same pass.
-
Translate when needed
When the source differs from your target language, the translation appears automatically. The entire process stays on your Mac.
Get started
Put the text on your screen to work
One shortcut lets you select, recognize, copy, and translate. Photos and PDFs become editable Markdown, too. Everything is processed on your Mac.