Opened a ready brief for convert-a-jpg-to-text-with-ocr (OCR N1): scene-sentence H1, Planner longs absorbed as chips/FAQ, no page code yet.

Date: 2026-09-07 23:50 Summary: Opened a ready brief for convert-a-jpg-to-text-with-ocr (OCR N1): scene-sentence H1, Planner longs absorbed as chips/FAQ, no page code yet. Visibility: people

[question] 立项：convert-a-jpg-to-text-with-ocr

[try to solve]

Context

The OCR/Scan Keyword Planner analysis (same day) listed four implementable slugs plus a weak table slug. Catalog neighbors that extract PDF text or put photos into PDF/Word all refuse OCR. The user named N1 only: photo/JPG → copyable plain text. Direction A.11 (Tier 2 WASM) is in scope for printed/photo text; handwriting, reflow Word, searchable PDF, and table CSV are other jobs or walls.

Process

Copied work-tasks/_template/ to work-tasks/convert-a-jpg-to-text-with-ocr/. Wrote 00 from the user’s 立项 line plus planner constraints. 01 picked primary direction A (A.11), not an industry invoice tool. 02 filled the §3.3 H scene table, IG preview, converter SEO (snippet-sized meta, How = button labels), Ads/Planner absorb table from N1 §6.1, and intent review (satisfy image→TXT; intentionally fail PDF/Word/table/batch/handwriting product pages). 03 has ten locale briefs with local search phrases (not English calques) and a 0b round row. Ran npm run coverage:gate -- --slug=convert-a-jpg-to-text-with-ocr --phase=0b (green). Updated the keyword pool N1 row defer→build with work_tasks_opened|await_serp, tracker snapshot/decision log, and the ocr-scan planner/README status line. No src/ / catalog / i18n / vendor changes.

Root cause / analysis

A category H1 (Online OCR, OCR to text) would doorway against i2OCR/OnlineOCR and would not match the conversion-pair job. Absorbing OCR queries into existing no-OCR extract/Word pages would be the same class of error as treating generate as validate. Splitting Arabic/Chinese/handwriting/Free into extra URLs would be scaled near-duplicates. The real product axis is the result file: this slug is TXT only. Click-to-load is required so first paint does not pull the OCR WASM (CWV). Auto loadSample() on landing is the documented exception, same family as not auto-running html2canvas on screenshot tools.

Solution

Brief status 02=ready, 03=briefs-ready. English H1 Convert a JPG to text with OCR. Chinese H1 direction: 用 OCR 把 JPG 转成文字. Primary input is one image dropzone (JPEG/PNG/WebP). Buttons: Convert, Copy text, Download TXT, Load sample, Clear. Default languages Chinese+English; Arabic as a chip, not arabic-ocr. Implementation session must POC printed EN/ZH before shipping; Japanese uses a fallback pack if tiny drops it. Related tools already live: extract-text-from-pdf, images-to-pdf. Neighbor FAQ backlinks wait until this page exists (CROSS_TOOL_UPDATE=1).

Notes / boundaries

No weekly long_gap claim (SERP not collected). N2–N5 remain unbriefed. Do not start *Page.ts or ten-locale i18n in this session. Next user command should be 实现, which is session B: vendor WASM, Page opts, English i18n, coverage:gate --phase=2, lint:tool-page.

[actions]

• Added work-tasks/convert-a-jpg-to-text-with-ocr/ (00–03, notes.md); coverage:gate phase 0b green
• Updated docs/seo/keyword-daily-pool.tsv N1 to build; tracker + ocr-scan README/planner status