timminator/VideOCR v1.3.3-Beta

VideOCR v1.3.3-Beta

on GitHub

Highlights:

• This new beta shows an performance increase up to 5x in Step 1.
• The CLI version can now be run in parallel to process multiple videos at the same time (thanks @Phr33d0m)
• ISO 639 Language codes are now added to the subtitle file names. This allows video players to identify the correct subtitle language.

Fixes:

• Fix usage of wrongly sized image for subtitle position "any" in ssim calculation. This resulted in a performance decrease in previous versions.

Highlight of the v1.3.x release

• The main highlight is the newly added support for PaddleOCR 3.x.x with greatly improved OCR capabilities in more than 40 languages,
  including English, Chinese, Japanese, French, Spanish, Portuguese, Russian, Korean, etc.

Notes:

• Utilizes standalone version of PaddleOCR from https://github.com/timminator/PaddleOCR-Standalone
• Linux version requires glibc 2.35 or newer (can be checked by running ldd --version in a terminal)

Download Tips:

• Windows:
  • CPU version:
    VideOCR-CPU-v1.3.3-Beta-setup-x64.exe
  • GPU version:
    VideOCR-GPU-v1.3.3-CUDA-11.8-Beta-setup-x64.exe (Nvidia 10 Series graphics cards)
    VideOCR-GPU-v1.3.3-CUDA-12.9-Beta-setup-x64.exe (Nvidia 16 - 50 Series graphics cards)
• Linux:
  • CPU version:
    VideOCR-CPU-v1.3.3-Beta-Linux.7z
  • GPU version:
    VideOCR-GPU-v1.3.3-CUDA-11.8-Beta-Linux.7z (Nvidia 10 Series graphics cards)
    VideOCR-GPU-v1.3.3-CUDA-12.9-Beta-Linux.7z.001 (Nvidia 16 - 50 Series graphics cards)
    VideOCR-GPU-v1.3.3-CUDA-12.9-Beta-Linux.7z.002 (Nvidia 16 - 50 Series graphics cards)