Briefed batch-convert-jpg-to-editable-word-with-ocr with Word-document IG, not a TXT-ZIP reskin; coverage gate 0b passed.

Date: 2026-09-08 20:25 Summary: Briefed batch-convert-jpg-to-editable-word-with-ocr with Word-document IG, not a TXT-ZIP reskin; coverage gate 0b passed. Visibility: people

[question] 立项：batch convert jpg to editable word with ocr，需要额外的info Gain增益，区别于其他ocr工具

[try to solve]

Context

The site already has printed-photo OCR that ends in text files: N1 (one JPG → one TXT) and N1-batch (many JPGs → ZIP of TXT + combined.txt). It also has images-to-word, which puts photos into a DOCX as pictures, so the letters are not editable. Planner N4 was a scanned-PDF → Word job and had parked jpg to word ocr as a chip on that unbuilt page.

The request is a new URL whose job is a stack of JPGs → one Word file you can edit. That is a different result file from TXT (planner §0) and a different object from PDF.

Process

1. Read OCR planner §0 / §3.2 / §6.4, N1-batch brief IG table, and images-to-word catalog copy.
2. Locked slug/H1 to the user task sentence: batch-convert-jpg-to-editable-word-with-ocr / Batch convert JPG to editable Word with OCR (Chinese direction: 用 OCR 批量把 JPG 转成可编辑的 Word). Forbidden H1s: OCR to Word, jpg to word, Batch OCR.
3. Wrote IG as controls, not adjectives: queue + skip (shared with N1-batch) plus one sectioned DOCX (Heading 1 = filename, page break, OCR paragraphs), default photo above text for proofing, primary button Download Word (not Download ZIP of txt).
4. Absorbed Planner jpg to word ocr here; left ocr pdf to word / convert scanned pdf to word for unbuilt N4. FAQ must split N1-batch (TXT ZIP) and images-to-word (no OCR).
5. Filled work-tasks/…/00–03 + notes.md; 02=ready, 03=briefs-ready. npm run coverage:gate --phase=0b green. Updated planner, keyword pool, tracker.

Root cause / analysis

A second OCR URL is doorway if the only change is MIME (.txt → .docx inside the same ZIP). Searchers who want Word want to open and edit in Word: sections, headings, optional proof images. Searchers who want a zip of notes stay on N1-batch. Searchers who only want a photo album in Word stay on images-to-word. PDF scanners stay on future N4.

Solution

• Brief folder: work-tasks/batch-convert-jpg-to-editable-word-with-ocr/
• Related (when implemented): N1-batch, N1, images-to-word
• Engine reuse: Tesseract vendor + existing /vendor/docx/
• No page code this session. No long_gap KPI (no human SERP).

Notes / boundaries

• Do not accept PDF. Do not promise column/table layout. Handwriting FAQ-fail.
• Implementation session must keep Convert all / Download Word labels aligned with How, skip auto-OCR on landing (Tier 2 WASM).

[actions]

• Added work-tasks brief; planner/pool/tracker/ocr-scan README
• npm run coverage:gate -- --slug=batch-convert-jpg-to-editable-word-with-ocr --phase=0b