Subtitle Edit 4.0.3 and 3.6.13

[mail2tom]

After installing Subtitle Edit (4.0.3 and 3.6.13) I tried to install the Whisper-Faster files. All went well with the required downloads until it came to the language model. With both versions of the software I received an error that I had no internet connection. Is Subtitle Edit looking at the wrong location to download the files?

[davexnet]

Try the latest beta
https://github.com/SubtitleEdit/subtitleedit/releases/download/4.0.3/SubtitleEditBeta.zip

[VoodooFX]

I received an error that I had no internet connection.

Most likely it's because older Windows.

[mail2tom]

Thanks for the replies. Tried the lastest Beta with the same results. Oddly enough the only thing the program won't download is the models. All other files, ffmpeg and Purfview's Faster-whisper executable work from within the program. Anyway, I downloaded the models independent of Subtitle Edit and will see if that works.

Edit: Manual download of the various Purfview's Faster-whisper and CPP models works fine. It would be nice if I knew why Subtitle Edit would not download them.

[robena]

Hi,

I have a pretty standardized for me workflow:

1) Remove HI
2) Run a pre-configured Multiple Replace
3) Running 2 passes of Fix Common Errors
4) Re-run Multiple Replace, that sometimes finds something new after step 3.

It's automatized with an AutoHotkey macro, that sends the appropriate shortcuts, waits for the popups, and sends whatever is necessary to make them work.

But sometimes, I have a lot of subs to process that way. For example 22 subs for a series season, and it takes a lot of time loading the subs one by one and waiting for AutoHotkey to do this work.

Unfortunately, SE does not provide all the command line arguments to do all these steps in batch mode.

I have a long background in C programming, that started on Unix 40 years ago to finally end up on Windows. Because of this, I use VS2022 mainly to debug, but my large projects are built in C and with command line tools (make, cl, lib, etc...).

Learning C# and actually studying most of SE code would have taken me weeks, so I gave it a go with AIs.

That did not work at first, both ChatGPT and Grok failed to do something usable, ending in an endless loop of test and retry.

Then I tried with Github Copilot that came really close, but still not operational. I used Github Copilot project making it better with ChatGPT, then Grok when it was stuck, and finally with Deepseek.

And now it seems that I have a working SE derived program that parses C:\Users\{username}\AppData\Roaming\Subtitle Edit\Settings.xml and performs all these steps.

That took one day to do that using these 4 AIs!

If anybody is interested, let me know the best way to share it with SE users.

[sam12345]

@robena

please share

[robena]

@robena

please share

Here is a link with the whole VS2022 project:

https://e.pcloud.link/publink/show?code=XZ8UoUZGkV4JJbm4S8NK7uz17Bdq0yAPkNX

That's on my personal account, and it won't be there forever. Feel free to repost it on somewhere more permanent, or just tell me where to post it myself!

If you just want to use the batch program, use:

subtitleedit-main\src\BatchProcessor\bin\Release\net48\Subtitle EditBatch.exe

I did not bother to make it accept more than one file at a time, because I use a wrapper script that calls it and deals with multiple files passed as argument.

The script is integrated to the shell, all I have to do is select several .src files, right click and use the script that is on the menu:


[Attachment 89281 - Click to enlarge]

There might be simple tools to do that now, but I used the original Microsoft project described here:

https://github.com/yanxijian/ShellExtContextMenuHandler

[sam12345]

@robena

https://e.pcloud.link/publink/show?code=XZ8UoUZGkV4JJbm4S8NK7uz17Bdq0yAPkNX

Link not working. Kindly upload on google drive or firebase

[robena]

@robena

https://e.pcloud.link/publink/show?code=XZ8UoUZGkV4JJbm4S8NK7uz17Bdq0yAPkNX

Link not working. Kindly upload on google drive or firebase

Weird, I just tried it with Chrome and Firefox, and if you click download on the upper right (not on the big key symbol in the middle), it works.

Here is a Google link:

https://drive.google.com/file/d/1fN4_bIrYs9Vn19ovc8hObRH-_uwRn12a/view?usp=sharing

Let me know if you have any problem, glad to be of assistance.

[robena]

I just got this:

Based on internal checks and an automated review, Google determined that "subtitleedit-main.zip" contains content that may violate Google Drive's Malware and Similar Malicious Content policy. This file can no longer be copied, and can no longer be shared with or viewed by other users, and notifications to users will not be sent. File owners retain access, and files in shared drives can still be accessed by the shared drive's organizers.*If you think this is an error and would like the Trust & Safety team to review this file, request a review below.

Let me try something else.

[robena]

Let's try this link:

https://drive.google.com/file/d/1N98ZcODvCLMPBKZS0B94aYURgpzN5F_y/view?usp=sharing

Rename the file to videohelp.zip, and extract it with the following password:

Not a malware!

I used WinRAR to encode it, not sure if other programs will work. If you don't have WinRAR, download a trial, it's free to test.

I guess that Google does not like EXE files. If that still does not work, I'll post just the sources, and you'll have to build it with VS2022.

You may need to install libraries not standard, I'll help if you don't have them installed.

Feel free to scan the directory with something like Malwarebytes.

[sam12345]

@robena

All links worked. I have to apply VPN in my region to download them

Can you make a project on Whisper {Audio to Text}

[robena]

@robena

All links worked. I have to apply VPN in my region to download them

Can you make a project on Whisper {Audio to Text}

I did that, but the process is a bit complex, I would have to spend a lot of time to document it, I did not do it properly at the time.

And whisper produces bad subs in my opinion.

Instead, I made a project with this:

https://www.assemblyai.com/

When you subscribe, the first 456h are free, which will last me a long time, I find subs 99.9% of the time on https://www.opensubtitles.org/en/search/subs or https://www.addic7ed.com/shows.php.

I built a C project for that, and results are very good, much better than with Whisper.

Here is the project:

https://e.pcloud.link/publink/show?code=XZ0JVIZsuwKzJpfahyfy8mrSBVVYLm5RXI7

I cannot provide an exe, my API key is hardcoded.

You need to get one from their site, and update the Assembly\Assembly\Assembly.c file:

#define API_KEY "YOUR KEY HERE" // Replace with your actual API key

You need git to install some libraries. If you don't have it:


Open Command Prompt/PowerShell.
Run: winget install --id Git.Git -e --source winget
This pulls the latest Git and handles PATH setup automatically.
Verify with git --version.

Clone the vcpkg Repository in an admin command shell:

cd C:\
git clone https://github.com/Microsoft/vcpkg.git

Bootstrap vcpkg:

cd vcpkg
.\bootstrap-vcpkg.bat

Integrate vcpkg with VS2022

.\vcpkg integrate install

Install libcurl via vcpkg

.\vcpkg install curl : x64-windows

Removes the spaces in "curl : x64", I had to put them here otherwise it displays as an emoticon!

I know, it's a bit complicated. I did that with Gork, if you hit a wall, ask it!

It's really worth it, the quality compared to whisper is night and day.

[robena]

Maybe there is simple way for you to use my whisper project (no guaranties):

Download:

https://e.pcloud.link/publink/show?code=XZ8JVIZrPrTKeFtpF4VFpRCv6pjfLS5TMvV

It's a whisper.exe program to be used with:

whisper "file" --model medium --task transcribe --word_timestamps True --device cuda

Not sure at all if it will work as is on your system. Remove --device cuda if you don't have an NVDIA GPU.

Edit: you need at least to install Python311

[sam12345]

@robena

Can it be done in google colaboratory with python libraries installed therein. We have been doing with Whisper in colaboratory

Runtime - Goolge T4 GPU

!pip install git+https://github.com/openai/whisper.git
!sudo apt update && sudo apt install ffmpeg

then upload file (audio/video)

then

!whisper "ENTER FILE NAME HERE" --model medium.en

[robena]

@robena

Can it be done in google colaboratory with python libraries installed therein. We have been doing with Whisper in colaboratory

Runtime - Goolge T4 GPU

!pip install git+https://github.com/openai/whisper.git
!sudo apt update && sudo apt install ffmpeg

then upload file (audio/video)

then

!whisper "ENTER FILE NAME HERE" --model medium.en

Google Colab gives you a free (sometimes paid) NVIDIA T4 GPU runtime. Whisper runs much faster on that than on CPU.

So yes, you should be able to do that if your PC resources are limited. I never tried, I don't now the exact process.

That won't change the result compared to using whisper locally. Compared to AssemblyAI, results are poor.

But if you have a huge number of videos to process, it's free while AssemblyAY, after the first free 456 hours, won't be. 456 more hours cost $50.

[sam12345]

@robena

commands worked but not

cd vcpkg
.\bootstrap-vcpkg.bat

corrected

cd vcpkg
./bootstrap-vcpkg.bat, ./vcpkg integrate install, ./vcpkg install curl : x64-windows

You have to place front slash in all ...not the back slash

Got it ...done but it requires Visual studio too.

[robena]

@robena

commands worked but not

cd vcpkg
.\bootstrap-vcpkg.bat

corrected

cd vcpkg
./bootstrap-vcpkg.bat, ./vcpkg integrate install, ./vcpkg install curl : x64-windows

You have to place front slash in all ...not the back slash

Got it ...done but it requires Visual studio too.

\ worked for me, usually on Windows / and \ are the same. Maybe it depends on the terminal you are using.

You need indeed Visual Studio 2022 to build the project once you'll key your API key, it's free and easy to install.

If something does not work, ask ChatGPT. I'm happy to help, but you'll get faster responses with it, it's really worth using it.

[sam12345]

@robena

Project worked perfectly. Yes, Little bit technical. srt made needs sone bit of improvement with line spacing [The length of the lines is too long]; have to use subtitle edit to make the correction.

Movie files working well but not working that well with music videos. Check the results here and music video here

PS: Kindly download the videos

[robena]

@robena

Project worked perfectly. Yes, Little bit technical. srt made needs sone bit of improvement with line spacing [The length of the lines is too long]; have to use subtitle edit to make the correction.

That's because it parses YOUR settings, then is will do exactly the same as a the GUI does when you customize it.

Edit: does it still make lines too long after setting the length with SE? If yes, it's missing this setting when parsing.

Movie files working well but not working that well with music videos. Check the results here and music video here

I don't know how to download these files, all I get is a preview without subs! I never used filebase.

Let me know how to download them.

That said, it's likely that you need different settings for music rather than with films. Nothing to do by the content being music, just your tastes on how to do things differently.

The only solution is to make your music settings wit the GUI, and save the C:\Users\{username}\AppData\Roaming\Subtitle Edit\Settings.xml some place.

Then make the settings you want with a movie, and save it with a different name.

After that, use 2 scripts with your favorite language (.BAT is fine) to copy into C:\Users\{username}\AppData\Roaming\Subtitle Edit\ the appropriate settings, maybe one script called BatchMusic and the other BatchMovies.

The purpose of this program is to use the default settings. If you need several settings for different content, that's the only way.

If you need to change the settings constantly, use the GUI, it's made for that.

Me, I don't, that's why I wrote this batch processing. It takes me a few seconds to process a whole 22 episodes srt files now compared to 10 minutes with the GUI.

[sam12345]

@robena

You can download them by clicking the 3 dot button on the right side below
[Attachment 89346 - Click to enlarge] then click on download
[Attachment 89347 - Click to enlarge]

[sam12345]

does it still make lines too long after setting the length with SE?

No with SE its fine but making with Assembly

[robena]

does it still make lines too long after setting the length with SE?

No with SE its fine but making with Assembly

Using an SE pass after using Assembly is not too bad.

Maybe there is a way to customize assembly to make shorter lines, but since I'll use SE batch anyway to do that, I don't feel like wasting time.

You could send the code used in my project to their support and ask what to do. Or ask ChatGPT!

[robena]

@sam12345

What's wrong with:


[Attachment 89348 - Click to enlarge]

Edit: I ran whisper, and you're right, for this clip it's much better.

No idea why, Assembly works much better for movies and series. I guess you can use both.

[robena]

Your video clip has an audio track that trips Assembly, making it stop before the end.

I have no idea why, even re-codding the audio to FLAC does not help.

You have other ways to use Assembly here:

https://www.assemblyai.com/docs/pre-recorded-audio/export-srt-or-vtt-caption-files

I tried the C# one after making mine by curiosity, but it did not work. I did not waste time with their support to know why.

The Python version is much less technical, you might want to try it.

Don't hesitate to ask their support for help.

[sam12345]

To see the best results of .srt files in movies with Assembly project. Modify characters lenght from 42 to 60
[Attachment 89356 - Click to enlarge] in Assembly.c and run the SE batch [4.0.13] with these settings.xml

[robena]

Thanks, I'll try that.

I got an answer from Assembly support:

I ran some tests on your file and I was able to recreate this issue. After listening to the file, I believe the issue is related to the fact that this file is a song. Our model's training does not consist heavily of songs and background music can cause hallucinations or missed text. I will file a bug report for you. Because this requires model retraining, which is a very involved process, I do not have an ETA for when transcriptions of songs/music will improve. I would suggest following our Changelog https://www.assemblyai.com/changelog page to learn more about new features and improvements as they're released.

[sam12345]

I got an answer too from Assembly support:

We're always looking to improve our models and we do have plans to release a model update with improved ASR accuracy in late November or early December.
If you aren't already, you can follow our Changelog page to learn more about new features and improvements as they're released.

[sam12345]

@robena

Not able to make the script.

The script is integrated to the shell, all I have to do is select several .src files, right click and use the script that is on the menu:

Image

. Confused with ChatGPT. Can you send an easy method

[robena]

@robena

Not able to make the script.

The script is integrated to the shell, all I have to do is select several .src files, right click and use the script that is on the menu:

Image

. Confused with ChatGPT. Can you send an easy method

Here:

https://e.pcloud.link/publink/show?code=XZb5RIZxjaINsm04qJKc4fK3WReOpDkPkRk

That's a minimum example.

If you look at:

\ShellExtstd\LNtoALL\C++\CppShellExtContextMenuHan dler\LNtoALL.cpp

You'll see these lines:

static ALLCALLS allcalls[] =
        {
           { L"&FileDo1", NULL, NULL, NULL, NULL, FALSE, TRUE, NULL, ".c,.cpp,.h",
             "\"C:\\Program Files\\SEDIT\\wsrexx.exe\" \"D:\\m1\\RexxMacs\\FileDo1\""
           },
           { L"&SEDIT",   NULL, NULL, NULL, NULL, TRUE,  TRUE, NULL, "*",
             "\"C:\\Program Files\\SEDIT\\xed.exe\""
           },
           { L"&FileDo2", NULL, NULL, NULL, NULL, FALSE, TRUE, NULL, "*",
             "\"C:\\Program Files\\SEDIT\\wsrexx.exe\" \"D:\\m1\\RexxMacs\\FileDo2\""
           }
                                                                                                               
        };

That makes a menu for right clicking on files.

The first item will work with those having ".c,.cpp,.h" extensions, and call a program named wsrexx.exe with D:\m1\RexxMacs\FileDo1\ as argument.

The second line calls my text editor SEDIT with any file.

I like the REXX language, so all my scripts use it. I use S/REXX, an interpreter I wrote myself: www.sedit.com

It's a commercial product, but there are free alternatives if you are curious like:

https://regina-rexx.sourceforge.io/

Mine is better of course.

You can use anything here. You can put the Assembly and Batch SE EXEs for example.

There is VS2022 project to build it:

ShellExtSTD\LNtoALL\C++\LNtoALL.sln

To register the DLL, use:

ShellExtSTD\LNtoALL\C++\CppShellExtContextMenuHand ler\RegMenuHandler.bat

Similarly, you have:

ShellExtstd\LNtoALLD\C++\CppShellExtContextMenuHan dler\LNtoALL.cpp

to add an item when you right click a directory.

You could build a script that scans a directory and all sub-directories to run Assembly or Bath SE on all SRT files for example.

About using SEDIT for subtitles, or a free alternative like:

https://hessling-editor.sourceforge.net/

although mine is MUCH better of course.

Both use something called "Selecting Editing".

That allows to display and edit lines that met a certain criteria, like a bunch of regexp.


I have a bunch of helpful macros like this one that shows all the errors the SE missed:

signal on novalue
option mixed setenv
nq = arg(1)
'all'
                                                                                                               
'regt e'
'case r'
'all ~/-->'
'all r/[A-Z]{3,}|[:\[\]\{\}\\/]/'
r1 = rc
'case i'
'show/sync/|/corrected/|/ripped/|/caption/|/subtit/|/sdh/|/subtitle/|/copyright/|/translated/|/-1/|/(/|/)/'
r2 = rc
r0 = (r1 = 2) & (r2 = 2)
if r0 then
  { rr = lookMus()
    if r0 & (alt.2() = 0) & nq = '' then 'exit'
  }
                                                                                                               
'msg'
'top'
return

That's incredibly helpful to flash HI comments that are not in uppercase like "McGee: I said". It also looks for SE OCR errors on music symbols with the LookMus() macro.

Also, it's the only way to remove HI for UK shows that don't use a ":" for comments, which subs like "The Ridge S01-E01.srt".

If I run this macro with SEDIT (or something similar with THE, not sure what it could do exactly) after running SE on it:

I see:


[Attachment 89373 - Click to enlarge]

It's easy in a glance to select and remove the HI line, but not lines like:

"I WILL be staying here"


[Attachment 89374 - Click to enlarge]

The regexp shows all words in caps with more than 3 characters. It also shows lines with a ":" and plenty of stuff that you might want to remove.

That removes lines 679, 787, etc.. but NOT the lines in between, that's the beauty of selective editing.

I configured SEDIT so that hitting F6 saves the files, runs SE batch on it and then reloads the file. Running SE is necessary to clean the file SRT file which will have empty lines. That take .5 seconds!

F7 does the same, but with the GUI.

It takes maybe 10s to remove these kind of UK HI comments. Wit SE, you have to scroll hundreds or thousands of lines and do them one by one.

If you're curious, I can give you a free SEDIT license, as a fellow videohelp member, no strings attached.

You can customize everything, like this menu that I use for subs:


[Attachment 89375 - Click to enlarge]

Or try THE and Regina if you want to stay with the free software.

REXX is a powerful and very easy to lean language. S/REXX is better of course.

I put this script in the shell for procesing Batch SE:

/*
 * BatchSRT
 */
                                                                                                               
signal on novalue
option setenv mixed
address exec
procedure
if 0 = index($PATH, $HOME'/rcmd') then  $PATH = $PATH';'$HOME'/rcmd;'$HOME'/cmd'
                                                                                                               
batch = '"D:\m1\VS Projects\Subtile Edit\subtitleedit-main\src\BatchProcessor\bin\Release\net48\SubtitleEditBatch.exe"'
                                                                                                               
i = 1
do forever
   file = arg(i, 'x')
   if file = '' then return
   if ft(file) = '.srt' then sayx batch '"'file'"'
   i = i+1
end

It's easy to change for Assembly to.