FrameShift 1.21.0
Extract Text from images and PDFs
Right-click a supported image or PDF and choose Extract text, or launch the action from the FrameShift hub. Choose PP-OCRv6 Small for higher accuracy or Tiny for faster processing, then save UTF-8 text, structured JSON or a searchable PDF that retains the original document appearance.
- Supports PNG, JPG/JPEG, WebP, BMP and PDF, including scanned and mixed native/scanned documents.
- PDF options include page ranges, native-text reuse, 200/300 DPI rendering, recognition rotation, line breaks or paragraph grouping, and optional text page markers.
- Reads detected columns in sequence. Oversized PDF pages automatically use a safe rendering resolution and report the quality adjustment instead of failing the document.
- Runs locally with CPU or DirectML. Small (approximately 31.7 MB) and Tiny (approximately 6.7 MB) download on demand with dictionaries, Apache-2.0 notices and SHA-256 verification; subsequent processing works offline.
- Creates a uniquely named output beside each source and preserves original files. Searchable PDF output retains every source page, including pages outside a selected OCR range.
Enable Extract text (images and PDF) in the installer to add the Explorer menus.
Download FrameShift_1.21.0_Setup.exe from release 1.21.0: Windows 10/11 x64, self-contained, no separate .NET installation required. The release also includes a SHA-256 checksum file.
Validation: 754 tests passed, 0 failed, 5 skipped because their Whisper test assets are unavailable. OCR tests executed on both CPU and DirectML. The owner accepted the image/PDF workflow, column correction and oversized-page correction. Complex or skewed layouts can still require reviewing the extracted text. See the feature guide and qualification report for details and remaining installed-qualification limits.