Activity for Simeon Kosnitsky
-
Simeon Kosnitsky posted a comment on discussion VideoSubFinder Forum
Hello, regarding "I usually do this in anime, so when the resolution is too low I use Real-ESRGAN to enhance them before uploading them to google with the python script made by Abu3safeer" Do you use Real-ESRGAN on RGBImages or on TXTImages? (it looks that you upload to google "Restored RGBImages" or TXTImages generated on "Restored RGBImages")
-
Hello, agree, will try to add this feature, but currently there is alternative solution, just checked on left aligned video, you can set through GUI "Join Images Split Line Text" to something like (by addding multiple space to the end): "[sub_id] " As result you will get:
-
Also try to check by defining number of threads for run search == 1 (or for clear images == 1). The issue can be related to not enough memory.
-
Hello, not sure but possibly it is blocked by Antivirus which sometimes happen at first run (Avast). Try to use this version (with extended logging) with opening video by both OpenCV and FFMPEG (ffmpeg has more logging in case of video open) https://drive.google.com/file/d/1ytJN986zIM8KrouVzsnoR2MMZ2lV3_5C/view?usp=share_link and provide report.log file and screen short for stuck. Does it stuck at the beginning of "Run Search" or it passed part of video and stuck at some moment or something else...
-
Hello, not sure but possibly it is blocked by Antivirus which sometimes happen at first run (Avast). Try to use this version (with extended logging) https://drive.google.com/file/d/1ytJN986zIM8KrouVzsnoR2MMZ2lV3_5C/view?usp=share_link and provide report.log file and screen short for stuck. Does it stuck at the beginning of "Run Search" or it passed part of video and stuck at some moment or something else?
-
Hello, not sure but possibly it is blocked by Antivirus which sometimes happen at first run (Avast). Btw how stuck is looked? Can you provide report.log + error.log (if it was generated) + screen short?
-
Ability to choose subtitle color like SubRip
-
Already supported.
-
Okay, try to use this package: https://drive.google.com/file/d/1-VgdOQM-CoGaBwqVaa_jkWSQflztDmI4/view?usp=share_link
-
Hi Kobayashi, Can you please run LD_LIBRARY_PATH=. ldd ./VideoSubFinderWXW | grep "not found" >log.txt And provide result: log.txt According your screen some dependencies now are missed after update.
-
New 6.10 version was released. Some Notes: VideoSubFinderWXW_intel.exe mostly faster then VideoSubFinderWXW.exe on i9900KF In case of 6.10: run_search: intel: real 1m9.142s msvs: real 1m22.828s clear_txt_images: intel: real 3m29.468s msvs: real 3m44.479s Also 6.10 working slower then 6.00 due to fix for small subtitles support [5f1730] : run_search: 6.10 Intel: real 1m9.142s 6.00 Intel: real 1m8.261s clear_txt_images: 6.10 Intel: real 3m29.468s 6.00 Intel: real 3m9.085s Main changes: *) Added join...
-
VSF 6.00 won't load positions setting from cfg file
-
Fixed in 6.10 version
-
Made fixes for Linux support
-
Updated -h support on Windows through CLI
-
Made fixes for Linux support
-
Fixed. Try to use this version: https://drive.google.com/file/d/1fTyzIbgbirV2u8Pe0-cPytd8Abs01J1n/view?usp=share_link
-
Reproduced, made minor fix but after it can be a little more false detections, also file names naming format was changed (not compatible with previous version generated), try to use this version. https://drive.google.com/file/d/1fTyzIbgbirV2u8Pe0-cPytd8Abs01J1n/view?usp=share_link
-
Made fix for CLI support
-
Made fix for Load Settings support
-
Made minor fix for support small subtitles like 'I ...' (but after can be more false detections)
-
Updated chn localization provided by Shijie Zhang
-
Fixed. Try to use this version: https://drive.google.com/file/d/1nbif21-PhsfpPKTtMj9a81mRiyQqr_rr/view?usp=share_link
-
Reproduced, made minor fix but after it can be a little more false detections, also file names naming format was changed (not compatible with previous version generated), try to use this version. https://drive.google.com/file/d/1nbif21-PhsfpPKTtMj9a81mRiyQqr_rr/view?usp=share_link
-
Thanks. Reproduced, will fix today.
-
Reproduced, made minor fix but after it can be a little more false detections, also file names naming format was changed (not compatible with previous version generated), try to use this version. https://drive.google.com/file/d/1GyKi6qCXXMjIPptaXPXEWjZEvLbuwHpA/view?usp=share_link
-
Hello, what concrete parameters do you mean? In case of general.cfg it save only positions of (separating lines) boundary box in "Video Box" and only if you save settings (ctrl+s). If you about last opened video with start and end time for "Run Search" it auto save it on exit to "previous_data.inf". So if next time open program you can select "Open Or Continue Previous Video" through menu.
-
Hello, what concrete parameters do you mean? In case of general.cfg it save only positions of (separating lines) boundary box in "Video Box" and only if you save setting (ctrl+s). If you about last opened video with start and end time for "Run Search" it auto save it on exit to "previous_data.inf". So if next time open program you can select "Open Or Continue Previous Video" through menu.
-
Updated 'Y' and 'I' testing Color Filters support (if average luminance in filter is < 50 than it will show pixels related to filters with original color on top of white background instead of black)
-
Made that order is no mater for separate lines in Video Box
-
Added support to modify any setting from general.cfg through command line by using it's name
-
Fixed critical issue with define Color Filters ranges support.
-
Added Help Popup Windows in Video and Image Box to show supported hotkeys and updated documentation.
-
Fix issue if use "Filter Colors" testing buttons in Video Box during Run Search.
-
Added preview scaled TXTImages in Image Box.
-
Added full screen preview in Image Box and left/right/"mouse wheel" to support.
-
Added auto creation "join_txt_results.txt" on Join Images button.
-
Added mouse wheel support during testing Filter Colors in Video Box (press and hold 'T' and etc).
-
Added "Join TXT/RGBImages" to support by CLI ("ji" option)
-
Added to documentation: ####################################################################################### Known Issues ####################################################################################### ... 6) Broken TXTImages "missed some symbols on them" multiple splits and incorrect timing in case if you define "Use Outline Filter Colors" but subtitles have not good outlines: subtitles has not solid outlines from all sides, some symbols are not contained in inside figures produced by...
-
The main reason of broken ILAImages and as result bad Clear TXT Images is that you have defined "Use Outline Filter Colors" when subtitles have not it: use_outline_filter_color = Lab: l:8 a:128 b:128 In you case subtitles have not outlines. According documentation: ... 2) "Use Filter Colors" and/or "Use Outline Filter Colors" ... *) Use not too strong "Use Outline Filter Colors" only if all symbols are contained inside borders: subtitles has solid outlines from all sides, symbols are contained in...
-
The main reason of broken ILAImages and as result bad Clear TXT Images is that you have defined "Use Outline Filter Colors" when subtitles have not it: use_outline_filter_color = Lab: l:8 a:128 b:128 In you case subtitles have not outlines. According documentation: ... 2) "Use Filter Colors" and/or "Use Outline Filter Colors" ... *) Use not too strong "Use Outline Filter Colors" only if all symbols are contained inside borders: subtitles has solid outlines from all sides, symbols are contained in...
-
Added Join RGBImages to support.
-
Thanks, i have gotten it. I have checked, it produce more better results if just set "Use Filter Colors" == "Lab: l:255 a:128 b:128", during Run Search and Clear Text Images. Not clear why so different results, it will require to make more deep and complex investigation.
-
Thanks, i have gotten it. I have checked, it produce more better results if just set "Use Filter Colors" == "Lab: l:255 a:128 b:128", during Run Search and Clear Text Images. Not clear why so different results, it will require to make more deep snd complex investigation.
-
Thanks, i have gotten it.
-
I have checked, if remove ILAImages and set "Use Filter Colors" == "Lab: l:180-255 a:108-148 b:108-148" Results looks better, but there still exist some issues, for example on: 0_04_25_380__0_04_25_659_0056700229001210072001280.jpeg I will try to fix it in future. If it is possible it will be very helpful if you will share original video.
-
Hi THXTEX, it looks you have not read "Quick Start Guide" from documentation "Docs\readme_eng.txt" which in 6.00 version also available in Help menu (directly through GUI). Your issue is related to this step (ILA Images are broken): 6) Check ILA Images in "ILAImages" folder: subtitles symbols by default will be searched inside white pixels in ILA images, if white pixels in ILA images will not contain some symbols or they are broken, which is possible if you use too strong Color Filters or subtitles...
-
Hello, i will add "Join Cleared Text Images" to support. Regarding "Create Empty Sub From Cleared Text Images", it should be already supported: /cscti, --create_sub_from_cleared_txt_images=<str> Create Sub From Cleared TXT Images With Provided Output File Name (*.ass or *.srt)
-
Hello, i will add them.
-
Strip BOM when joining text files in TXTResults
-
Fixed in 6.00 version.
-
Hi THXTEX, im'not sure what are you trying to do. Provided by you link is on archive with images with outdated naming format and last modified aug 4, 2014. If issue is during run search, i will need video example to reproduce. If issue during clear txt images, then i will need RGBImages, ILAImages, ISAImages, general.cfg (more simple to provide whole archive with program and arctifacts)
-
Crashes when using CUDA
-
You GPU currently is not supported by latest CUDA Toolkit 12.0 Update 1 More details you can find here: https://forums.developer.nvidia.com/t/cuda-11-2-w-gtx-770/164506 According comment for "CUDA 11.2 w/ GTX 770?": Support for Kepler sm_30 and sm_32 architecture based products is dropped. Now supported only: "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.0\bin\nvcc.exe" --list-gpu-code sm_50 sm_52 sm_53 sm_60 sm_61 sm_62 sm_70 sm_72 sm_75 sm_80 sm_86 sm_87 sm_89 sm_90
-
There are Windows x64 and Arch Linux builds which now also support join RGBImages (also with scaling support in range x1-x4), see settings in image. This version doesn't support old images - it have redesigned saved info in file names but still compatible with SubtitleEdit. But forgot to update "Create Sub From Text Results (TXTResults->Sub)" So now "Create Sub From Text Results (TXTResults->Sub)" should work only if: *) TXTImages was generated *) turned on "Use TXTImages Info For Join RGBImages"...
-
There are Windows x64 and Arch Linux builds which now also support join RGBImages (also with scaling support in range x1-x4), see settings in image. This version doesn't support old images - it have redesigned saved info in file names but still compatible with SubtitleEdit. But forgot to update "Create Sub From Text Results (TXTResults->Sub)" So now "Create Sub From Text Results (TXTResults->Sub)" should work only if TXTImages was generated and with turned off "Save Each Substring Separately" (which...
-
There are Windows x64 and Arch Linux builds which now also support join RGBImages (also with scaling support in range x1-x4), see settings in image. This version doesn't support old images - it have redesigned saved info in file names but still compatible with SubtitleEdit. Windows x64: https://drive.google.com/file/d/1NxVC7f_pitphEq6MiVMxofZ_xB8gpKCe/view?usp=share_link Arch Linux: https://drive.google.com/file/d/1eCZRx-r7P8-_R0XKRQgy6M3TUGxNKWym/view?usp=share_link
-
In case of TXT Images you can define scale from x1 to x4 so 480p in x2-x3 should be good enough. In case of RGB images it is not fully clear how better to support it. But i will try to add its support too.
-
Hi Kobayashi, I have released 6.00 version, it include fix for *.ass generation, more details by the link: https://sourceforge.net/p/videosubfinder/discussion/684990/thread/db19783116/?page=1#39b5
-
New 6.00 version was released. Main changes: *) Added "Google Drive" to support, see documentation: Docs\readme_eng.txt *) Added support for 3 localizations: eng, rus, chn (translation by Shijie Zhang) *) Added fonts selection *) Added support for dynamic changes through GUI for: localizations, loaded settings, changing fonts and their sizes *) Extended menu *) And many others, see whole list below List of all changes: [fc3a5d] Made minor fixes [ae01ef] Updated JoinTXTImages to support joined txt...
-
Made minor fixes
-
Big thanks, i will fix it today.
-
Hi Kobayashi, thanks for detection of issue, but for reproduce i need not only results.txt but also TXTImages and settings file general.cfg Also it will very help if you will show which concrete lines are missed. ... According *.ass file i see as min issue with: Dialogue: 0,0:10:39.13,0:10:40.26,Default,,0,0,0,,Hå? ________________ Dialogue: 0,0:10:42.51,0:10:44.43,Default,,0,0,0,,Bộ đồ đó hợp đấy. ... Dialogue: 0,0:14:18.39,0:14:24.65,Default,,0,0,0,,Nếu tạo làm thoả mãn mày như cách tạo làm với...
-
Hi Kobayashi, thanks for detection of issue, but for reproduce i need not only results.txt but also TXTImages and settings file general.cfg Also it will very help if you will show which concrete lines are missed. ... According ass file i see as min issue with: Dialogue: 0,0:10:39.13,0:10:40.26,Default,,0,0,0,,Hå? ________________ Dialogue: 0,0:10:42.51,0:10:44.43,Default,,0,0,0,,Bộ đồ đó hợp đấy. ... Dialogue: 0,0:14:18.39,0:14:24.65,Default,,0,0,0,,Nếu tạo làm thoả mãn mày như cách tạo làm với...
-
Moved JoinTXTImages to separate thread and extended logging
-
Updated JoinTXTImages to support joined txt results and multiple parameters
-
Made fix for "Save Each Substring Separately"
-
Current name format is supported by Subtitle Edit (which is used by multiple users), changing it will break compatibility. https://www.videohelp.com/software/Subtitle-Edit For generate srt or ass file with timing by using TXTImages or RGBImages names you can use buttons in OCR tab: Create Empty Sub From Cleared TXTImages # (TXTImages->Sub) Create Empty Sub From RGBImages # (RGBImages->Sub)
-
After OCR on google drive you need to place all text results to single file \TXTResults\results.txt then press button Create Subs From TXT Results It use TXTImages to get sub_id's and time positions and map them with sub_id's data found in \TXTResults\results.txt
-
There is also option for Join Unscaled TXT Images (which by default now is on), according your note i can replace it by another option like Use Scale For Join TXT Images: 1/2/3/4 etc
-
After OCR on google drive you need to place all text results to single file \TXTResults\results.txt then press button Create Subs From TXT Results
-
I think the reason possibly is that i have zipped archive on Windows, not on Linux, as result permission flags wasn't correctly stored. I will be able to update archive only at Sunday (now not at home). The solution can be do: 1) sudo chmod -R 777 in VideoSubFinderWXW folder (it will recursively add all access flags for all binaries) 2) or simply replace all library files (.so.x) and VideoSubFinderWXW.run (by binaries from 5.80 archive, they should be same) but there still possibly will be required...
-
I think the reason possibly is that i have zipped archive on Windows, not on Linux, as result permission flags wasn't correctly stored. I will be able to update archive only at Sunday (now not at home). The solution can be do: 1) sudo chmod -R 777 in VideoSubFinderWXW (it will recursively add all access flags for all binaries) 2) or simply replace all dll files (.so.x) and VideoSubFinderWXW.run (by binaries from 5.80 archive, they should be same) but there still possibly will be required to chmod...
-
I think the reason possibly is that i have zipped archive on Windows, not on Linux, as result permission flags wasn't correctly stored. I will be able to update archive only at Sunday (now not at home). The solution can be do: 1) sudo chmod -R 777 in VideoSubFinderWXW (it will recursively add all access flags for all binaries) 2) or simply replace all dll files (.so.x) and VideoSubFinderWXW.run but there still possibly will be required to chmod VideoSubFinderWXW (executable binary) to add it permissions...
-
Here are the latest builds for Windows and Arch Linux. Arch Linux: https://drive.google.com/file/d/1dvn7W3wu6vjRYVL5dY3ghfFIvmm4oq0C/view?usp=share_link Windows x64: https://drive.google.com/file/d/1plHVpG-FxR9WqXGKbzeAbv0nZFBsEaKL/view?usp=share_link Official build it can take few weeks.
-
I'm working on finalizing changes, so it will be available on all OS.
-
It will has multiple related new settings: ocr_join_txt_images_split_line = [sub_id] ocr_join_txt_images_split_line_font_size = -1 ocr_join_txt_images_split_line_font_bold = 1 ocr_sub_id_format = [SUB%d] ocr_sub_search_by_id_format = \[SUB%d\](.*?)\[SUB ocr_join_unscaled_txt_images = 1 ocr_max_join_txt_images = 150
-
You can try to use this version, now it support joined txt results data with sub_id translated by google drive. See \TXTImages\ \TXTImagesJoined\ \TXTResults\results.txt \sub.srt https://drive.google.com/file/d/18LPlfV6wjzEuDUvlyWpoqwzZMziiNFhf/view?usp=share_link
-
Some investigation notes. In case of google drive manual usage: upload image to google drive, right mouse click, open with "Google Docs", it will automatically OCR (but this not always working) I have found that in your video example is the best way to: 1) turn on "Save Each Substring Separately" 2) turn off "Save Scaled Images" 3) set "Join TXTImages Split Line Text" to: [begin_time] It successfully recognize first 160 Joined TXTImages from 291 in total (on 180+ it doesn't work)
-
Some investigation notes. In case of google drive manual usage: upload image to google drive, right mouse click, open with "Google Docs", it will automatically OCR (but this not always working) I have found that in your video example is the best way to: 1) turn on "Save Each Substring Separately" 2) turn off "Save Scaled Images" 3) set "Join TXTImages Split Line Text" to: [begin_time] It successfully recognize first 160 Joined TXTImages (on 180+ it doesn't work)
-
Also there exist another python script from looks other user with batch whole process with translation: https://sourceforge.net/p/videosubfinder/discussion/684990/thread/af99aafebf/#f2ed https://gist.github.com/kumorikuma/e4deca4e9384c0391acc784de3a8f015 # Install Python Module: pip install google-cloud-translate # Setup Google Cloud account and billing information: https://cloud.google.com/ # Make a new project and enable "Cloud Translation API": https://console.cloud.google.com/apis/dashboard #...
-
Also there exist another python script from looks other user with batch whole process with translation: https://gist.github.com/kumorikuma/e4deca4e9384c0391acc784de3a8f015 # Install Python Module: pip install google-cloud-translate # Setup Google Cloud account and billing information: https://cloud.google.com/ # Make a new project and enable "Cloud Translation API": https://console.cloud.google.com/apis/dashboard # Install GCloud CLI: https://cloud.google.com/sdk/docs/install # Setup authentication:...
-
I have checked Screen Translator and unfortunately for OCR images it use only Tesseract (as i understand) but for translation it support multiple web API services including google API. BTW i have found that according discussion: https://sourceforge.net/p/videosubfinder/discussion/684990/thread/352389f0e9/#c00e someone already implement batch processing python script for using Cloud Google Vision API https://cloud.google.com/vision https://github.com/Abu3safeer/image-ocr-google-docs-srt https://youtu.be/JXbL-PEoT4I...
-
Updated Open/Load/Save Dialogs to support restore previous locations and file names
-
Updated load settings support
-
Added help dialogs auto scaling.
-
Added font selection support and updated localizations
-
Added saving previous windows locations.
-
Updated text scaling support in Video and Image Box.
-
Moved saved images on Test button to separate folder TestImages
-
Not sure, but possibly yes in future. There is a tool screen-translator which allow to use google_api.js and others for translation. So possibly i will add this feature in future. https://sourceforge.net/projects/screen-translator/
-
Removed outdated duplicated files and moved them to single common places.
-
Added Chinese translation from Shijie Zhang
-
Made default output subtitles name and directory in consistent with the last opened video.
-
Renamed TestImages to DebugImages
-
Fixed issue with separating line overlapping in video box on Windows
-
Made fixes for localization switch support through GUI
-
Made fix for first and last video frame detection in Run Search.
-
Added dynamic controls scaling dependently from text sizes and updated load settings support.
1 >
