| Control | Purpose |
|---|---|
A- / A+ |
Decrease/increase translated text size. |
مطابق / مقروء / موحد |
Cycle layout mode. |
اصل |
Copy captured original text. |
ترجمة |
Copy translated text. |
ترجم |
Capture, OCR/read text, translate, and render overlay. |
↑ / ↓ or mouse wheel |
Scroll translated overflow; use Ctrl + wheel to resize while translated. |
👁 |
Toggle live preview. |
▣ خلفية |
Cycle translated background color. |
◌ % |
Cycle translated background opacity. 100% becomes reader-page mode. |
حبر |
Cycle translated text color. |
القارئ / فصل القارئ |
Open a detachable reader window with its own text size, wrapping, background, opacity, ink, copy, and dictionary controls. |
معجم المحدد |
Send the selected native-reader/editor text to the dictionary workflow and mark its source OCR block. |
- Translate a document.
- Select text in the transparent mirror editor or detached reader, or click a visible translated/source hit region.
- The selected term is highlighted and the corresponding OCR/source block is marked on the original capture when measured provenance is available.
- Right-click and choose معجم المحدد, or use the visible dictionary button.
- Use the المعجم tab for the linguistic answer and التوثيق التقني for OCR rectangles, hashes, local Merkle root/path, and translation-source proof.
- Drag the dictionary card if it covers the source text.
The dictionary prompt asks the model to classify the document domain, provide at least five alternatives, explain fit/non-fit contexts, and end with a decisive summary. It is instructed to act as a provenance auditor: do not treat MT or original fallback text as a verified Sarmad glossary path. If the AI gateway is unavailable, the panel reports the exact failure instead of inventing alternatives.
For publisher-style or confidential manuscripts, keep the Sarmad gateway
configured and decline the MT prompt unless you explicitly accept sending text
to Google gtx / MyMemory. MT fallback is not executed by default, is labelled
in the overlay when used, applies to the whole capture after confirmation, and
is not equivalent to the domain-aware academic AI translation path.