How to call native OCR APIs in Winstis plugins for screen text recognition?
I'm building a plugin that extracts tabular data from screen regions into CSV files, and want to leverage the built-in OCR service.
The design documentation specifies:
OCR: Built into the platform; plugins invoke it directly without bundling models.
Questions regarding current SDK:
- Does it accept image Base64 or a temp file path?
- Can we specify language hints (e.g., zh, en)?
- What is the callback/promise pattern for long-running recognition?
6 days ago
1.2k
3 Replies
AL
Regarding OCR APIs:
- The SDK provides
winstis.ai.ocr(imageSource), accepting both absolute file paths and Uint8Array buffers. - It uses OS-native hardware-accelerated OCR models with auto-detect for Chinese and English.
- Full docs will be published with Plugin SDK v0.2.0 this week!
6 days agoReplied by:
DA
@Alex Got it. Is there any discrepancy in recognition accuracy between macOS and Windows offline?
6 days agoReplied by: