Skip to content

Latest commit

 

History

History
25 lines (16 loc) · 1.46 KB

ocr-models.md

File metadata and controls

25 lines (16 loc) · 1.46 KB

OCR model setup

简体中文

The English Auto Detect workflow uses CRAFT for word boxes and PP-OCRv6 for recognition. Install requirements-ocr.txt in ComfyUI's Python environment and restart ComfyUI. The workflow's OCR model downloads note links to each file and lists where to save it.

For a remote ComfyUI instance, put the files on the server running ComfyUI. Browser downloads are saved on your own computer.

Manual installation

Download these files and place them in your ComfyUI model directory:

Download Destination under ComfyUI/models/pixtext_ocr/
CRAFT model ZIP — extract craft_mlt_25k.pth craft_mlt_25k.pth
PP-OCRv6 weights PP-OCRv6_medium_rec_onnx/inference.onnx
PP-OCRv6 configuration PP-OCRv6_medium_rec_onnx/inference.yml

From the PixText node folder, verify the installed files with ComfyUI's Python:

python scripts/verify_ocr_models.py --models ../../models/pixtext_ocr

The command prints Verified for all three files when their SHA-256 checksums match. Run the workflow after the files are in place.