TM Capture (beta)

Capture source and target text from webpages or images into a bilingual table, then export as Excel or TMX.

Capture is a tool for professional translators who need to build translation memories from content that exists outside their CAT tool — web pages, images, PDFs, and Office documents. CAPTURE FROM ANY PAGE Click Capture Source, drag a box around the original text, then Capture Target on the translation — each pair becomes a row in a bilingual table. Works on real page text and, via OCR, on images, canvases, and scanned PDFs. Keyboard shortcuts (Alt+Shift+S / Alt+Shift+T) keep the capture loop fast. IMPORT WHOLE DOCUMENTS Import a source document and its translation directly from your computer — Word (.docx), PowerPoint (.pptx), Excel (.xlsx), PDF, or images. TM Capture reads both, drops untranslated boilerplate automatically, aligns the rest into segment pairs using AI, and splits everything to sentence level. Existing .xlsx and .tmx files can also be imported to continue building on them. AI-ASSISTED SEGMENTATION AI Split divides a captured block into correctly aligned sentence pairs based on meaning, not just punctuation — it handles merged, split, reordered, duplicated, and added sentences, and marks anything doubtful with a warning flag instead of guessing silently. BUILT FOR REVIEW Per-row review checkmarks with a progress bar, a local QA check (untranslated rows, numbers missing from the target, spacing issues), search and filtering, duplicate removal, undo, and multiple named projects — one independent table per document you work on. EXPORT One click exports the table as Excel (.xlsx) or as a standard TMX 1.4 translation memory ready for Trados, memoQ, and other CAT tools, named after your project and target language. BRING YOUR OWN KEYS, KEEP YOUR DATA OCR uses Google Cloud Vision and AI features use Google Gemini — both called directly from your browser with API keys you supply in Settings (both have free tiers). There is no developer server: no analytics, no accounts, no data collection. Everything you capture stays in your browser except the content you explicitly send to Google's APIs for OCR or alignment.