Drop a script here, or click to browse
Supports PDF, DOCX, PNG, JPG, WebP
Advanced: high-precision transcription (optional)
When on, SayCaps sends your image or scanned PDF to Anthropic's Claude API for a character-accurate transcription. Uncertain words are flagged as [unclear] for review. The default Tesseract path stays free and fully local — this is an opt-in upgrade.
Your key is stored only in this browser (localStorage) and goes directly from your browser to Anthropic — SayCaps has no backend. When high-precision is enabled, image data does leave your machine. Get a key →
How it works
- PDFs: If the PDF has a text layer, we extract it directly. Scanned PDFs go through OCR.
- DOCX: Read directly from the file structure.
- Images: Preprocessed before OCR for best results.
- Paste: Skip file handling entirely and paste script text directly.
- Flagging: Low-confidence OCR words are highlighted in yellow.
- Script parsing: Detects character names and strips noise like page numbers.
- Export: Drag the text box in the preview to position it, then export. The preview matches the export exactly.