1. Install and use VideoSubFinder

2. Install rapid_videocr 

  pip install rapid_videocr

3. Python usage 

  from rapid_videocr import RapidVideOCR

# RapidVideOCRInput has two initialization parameters
# is_concat_rec: Use a single image for recognition or not. The default is False, which means that a single image is used for recognition by default.
# concat_batch: The number of images to be used in overlay is 10 by default and can be adjusted
# out_format: Output format selection, [srt, txt, all], the default is all
# is_print_console: Whether to print the result, [0, 1], the default is 0 for not printing
ocr_input_params = RapidVideOCRInput(
is_batch_rec=False, ocr_params={"Global.with_paddle": True}
)
extractor = RapidVideOCR(ocr_input_params)

rgb_dir = "tests/test_files/RGBImages"
save_dir = "outputs"
save_name = "a"

# outputs/a.srt  outputs/a.t
extractor(rgb_dir, save_dir, save_name=save_name)
  
  from rapid_videocr import RapidVideoSubFinderOCR

vsf_exe = r"G:\ProgramFiles\VideoSubFinder_6.10_x64\Release_x64\VideoSubFinderWXW.exe"
extractor = RapidVideoSubFinderOCR(vsf_exe_path=vsf_exe, is_concat_rec=True)

# video_path can be directory path or video full path.
video_path = 'test_files/tiny/2.mp4'
save_dir = 'outputs'
extractor(video_path, save_dir)

4. Command line usage 

  rapid_videocr -i RGBImages
  
  rapid_videocr -vsf G:\ProgramFiles\VideoSubFinder_6.10_x64\Release_x64\VideoSubFinderWXW.exe -video_dir G:\ProgramFiles\RapidVideOCR\test_files\tiny

Parameter details:

  $ rapid_videocr -h
usage: rapid_videocr [-h] [-video_dir VIDEO_DIR] [-i IMG_DIR] [-s SAVE_DIR]
            [-o {srt,txt,all}] [--is_concat_rec] [-b CONCAT_BATCH] [-p]
            [-vsf VSF_EXE_PATH] [-c] [-r] [-ccti] [-ces CREATE_EMPTY_SUB]
            [-cscti CREATE_SUB_FROM_CLEARED_TXT_IMAGES]
            [-cstxt CREATE_SUB_FROM_TXT_RESULTS] [-ovocv] [-ovffmpeg] [-uc]
            [--start_time START_TIME] [--end_time END_TIME]
            [-te TOP_VIDEO_IMAGE_PERCENT_END]
            [-be BOTTOM_VIDEO_IMAGE_PERCENT_END]
            [-le LEFT_VIDEO_IMAGE_PERCENT_END]
            [-re RIGHT_VIDEO_IMAGE_PERCENT_END] [-gs GENERAL_SETTINGS]
            [-nthr NUM_THREADS] [-nocrthr NUM_OCR_THREADS]

optional arguments:
-h, --help            show this help message and exit

VideOCRParameters:
-video_dir VIDEO_DIR, --video_dir VIDEO_DIR
                        The full path of video or the path of video directory.
-i IMG_DIR, --img_dir IMG_DIR
                        The full path of RGBImages or TXTImages.
-s SAVE_DIR, --save_dir SAVE_DIR
                        The path of saving the recognition result. Default is
                        "outputs" under the current directory.
-o {srt,txt,all}, --out_format {srt,txt,all}
                        Output file format. Default is "all".
--is_concat_rec       Which mode to run (concat recognition or single
                        recognition). Default is False.
-b CONCAT_BATCH, --concat_batch CONCAT_BATCH
                        The batch of concating image nums in concat
                        recognition mode. Default is 10.
-p, --print_console   Whether to print the subtitle results to console. -p
                        means to print.

VSFParameters:
-vsf VSF_EXE_PATH, --vsf_exe_path VSF_EXE_PATH
                        The full path of VideoSubFinderWXW.exe.
-c, --clear_dirs      Clear Folders (remove all images), performed before
                        any other steps. Default is True
-r, --run_search      Run Search (find frames with hardcoded text (hardsub)
                        on video) Default is True
-ccti, --create_cleared_text_images
                        Create Cleared Text Images. Default is True
-ces CREATE_EMPTY_SUB, --create_empty_sub CREATE_EMPTY_SUB
                        Create Empty Sub With Provided Output File Name (*.ass
                        or *.srt)
-cscti CREATE_SUB_FROM_CLEARED_TXT_IMAGES, --create_sub_from_cleared_txt_images CREATE_SUB_FROM_CLEARED_TXT_IMAGES
                        Create Sub From Cleared TXT Images With Provided
                        Output File Name (*.ass or *.srt)
-cstxt CREATE_SUB_FROM_TXT_RESULTS, --create_sub_from_txt_results CREATE_SUB_FROM_TXT_RESULTS
                        Create Sub From TXT Results With Provided Output File
                        Name (*.ass or *.srt)
-ovocv, --open_video_opencv
                        open video by OpenCV (default). Default is True
-ovffmpeg, --open_video_ffmpeg
                        open video by FFMPEG
-uc, --use_cuda       use cuda
--start_time START_TIME
                        start time, default = 0:00:00:000 (in format
                        hour:min:sec:milisec)
--end_time END_TIME   end time, default = video length
-te TOP_VIDEO_IMAGE_PERCENT_END, --top_video_image_percent_end TOP_VIDEO_IMAGE_PERCENT_END
                        top video image percent offset from image bottom, can
                        be in range [0.0,1.0], default = 1.0
-be BOTTOM_VIDEO_IMAGE_PERCENT_END, --bottom_video_image_percent_end BOTTOM_VIDEO_IMAGE_PERCENT_END
                        bottom video image percent offset from image bottom,
                        can be in range [0.0,1.0], default = 0.0
-le LEFT_VIDEO_IMAGE_PERCENT_END, --left_video_image_percent_end LEFT_VIDEO_IMAGE_PERCENT_END
                        left video image percent end, can be in range
                        [0.0,1.0], default = 0.0
-re RIGHT_VIDEO_IMAGE_PERCENT_END, --right_video_image_percent_end RIGHT_VIDEO_IMAGE_PERCENT_END
                        right video image percent end, can be in range
                        [0.0,1.0], default = 1.0
-gs GENERAL_SETTINGS, --general_settings GENERAL_SETTINGS
                        general settings (path to general settings *.cfg file,
                        default = settings/general.cfg)
-nthr NUM_THREADS, --num_threads NUM_THREADS
                        number of threads used for Run Search
-nocrthr NUM_OCR_THREADS, --num_ocr_threads NUM_OCR_THREADS
                        number of threads used for Create Cleared TXT Images

5. View results

Go to the save_dir directory to view the results.

Edit this page

Last updated 21 May 2025, 19:10 -0600 . history