# Changelog of Google Lens OCR API: Image to Text with Standby (`getascraper/google-lens-ocr`) Actor

- **URL**: https://apify.com/getascraper/google-lens-ocr/changelog.md
- **Full Actor documentation**: https://apify.com/getascraper/google-lens-ocr.md

## Changelog

### \[1.3] - 2026-09-09

#### Added

- ⚡ A live Standby REST endpoint (`GET /ocr?imageUrl=...`) for single-image requests with no run start-up wait, alongside normal batch runs.
- 🔤 Real word-level bounding boxes and rotation angle, opt-in via `outputDetail: "words"`. Every box comes directly from Google's own response, never estimated.
- 📚 Paragraph grouping and writing direction (left-to-right, right-to-left, or top-to-bottom), opt-in via `outputDetail: "paragraphs"` or `"words"`.
- 🎯 Detected regions and objects in the image, at no extra cost, shown alongside paragraphs and words.
- 🗂️ A new dataset view for paragraphs, words, and regions.

#### Fixed

- 💳 Failed images are now correctly billed as near-free audit rows instead of the full per-image price.
- 📝 The Store listing and FAQ no longer contradict the Actor's own capabilities.

#### Changed

- 🖼️ The input form now leads with a single image URL, the way most runs actually use this Actor. Bulk URLs, Base64 data, and key-value store records moved to an "advanced" section, still fully supported.

### \[1.2] - 2026-08-01

#### Added

- 🧬 Added Base64 image inputs with optional correlation IDs.
- 🗄️ Added Apify key-value store image inputs.
- 🏷️ Added stable `inputId` correlation across all source types.
- 📐 Added source image dimensions, word counts, source-provided line geometry, and explicit geometry availability flags.
- 🌐 Added optional translated text with explicit partial-result handling.
- ⏱️ Added per-input duration, status, error code, and error message fields.
- 📊 Added a `RUN_SUMMARY` record with submitted, successful, partial, and failed counts.
- ⚡ Added bounded batch concurrency while preserving input order.

#### Changed

- ✅ Every submitted image now produces one honest dataset row, including failed inputs.
- 🛡️ Removed invented zero-coordinate fallbacks and the upstream full-image geometry fallback.
- 🧾 Reduced output detail choices to capabilities supported by the current OCR source.
- 💳 Aligned PPE billing with the tiered Store price. Successful and partial OCR results emit the processed-image event, while failed audit rows are not charged.
- 📴 Disabled Standby mode and removed the web server schema.
- 🗂️ Updated dataset views for batch overview, text, line geometry, and errors.

#### Removed

- 🧩 Removed unsupported language and region hints.
- 📦 Removed duplicated paragraph output and unsupported word-box claims.
- 🌐 Removed the Standby HTTP server and its `/ocr` endpoint.

### \[1.1] - 2026-07-02

#### Changed

- 🧹 Cleaned up the Store listing page for a clearer, faster read.

### \[1.0] - 2026-06-29

#### Added

- ✨ Added image-to-text extraction.