Auto Sub ReTimer GUI (GPU-CPU) V1.8.2

What's New in v1.8.2

Auto Sub ReTimer GUI.bat

• The cmd window now opens minimized to keep the GUI in the foreground, and closes automatically when the GUI is exited.

Whisper – Fase0

• Fixed an issue that could cause a stall during Fase0.
  Previously, to proceed it was necessary to manually delete leftover files from previous processes, close and reopen the GUI. This will no longer occur.

Whisper ReTimer, GUI

• Fixed an issue where, if starting the Whisper ReTimer process multiple times without restarting the GUI, files were moved to the desktop before the user responded to the "Do you want subtitles to respect scene changes?" prompt.
• Whisper ReTimer now disables along with Whisper if the Faster-Whisper-XXL folder is missing.
• Minor fixes.

Fase5

• Fixed a bug where selecting "Yes" to respect scene changes moved both Final.srt and whisper_adjusted.srt instead of only the required file.
  Now:
  • If "Yes" is selected → only Final.srt and the related MKV are moved.
  • If "No" is selected → only whisper_adjusted.srt and the related MKV are moved.

Credits

• Special thanks to @Adrikss (@Adriks11 on Telegram) for testing.

If you want to use Whisper and Whisper ReTimer in Auto Sub ReTimer GUI, download and place the Faster-Whisper-XXL folder inside the Auto Sub ReTimer folder.

Whisper taken from Purfview/whisper-standalone-win

Full Changelog: V1.8.1...V1.8.2

Auto Sub ReTimer GUI (GPU-CPU) V1.8.1

Auto Sub ReTimer:

"Fase1":

• Added two parameters to maximize quality and reduce artifacts in voice separations.

• Added a parameter to speed up the process.
  In CPU mode, there will be a significant speed increase.
  In GPU mode, only processes running on the CPU will be accelerated, so the performance gain will be limited.

(This means that with fewer artifacts and noise, the timing adjustment precision may be higher in some situations)

• Demucs can now handle both multi‑channel and mono audio.
  Since Demucs works best with 2‑channel stereo audio, "Fase1" will convert the audio (in 2–4 seconds) into a temporary WAV file that Demucs can process.
  The conversion will only occur if the track is not 2‑channel stereo, otherwise, it will use the video's audio directly (as before).

(Previously, Demucs used multi‑channel audio to extract vocals, but ended up muting them completely, making it useless for timing adjustments which would either get worse or remain unchanged. Now it works with any type of audio.)

I RECOMMEND USING VIDEOS WITH STEREO AUDIO FOR BETTER VOCAL SEPARATION.

"Fase2":

• Reset some hardcoded values ​​used for testing (these could lengthen leads in certain situations).

Whisper ReTimer

"Fase0":

• Updated Demucs as in "Fase1" of Auto Sub ReTimer.
• Removed some parameters that were slowing down the audio extraction and conversion process. Added a parameter to speed it up.

"whispertiming":

• Updated as in "Fase2" of Auto Sub ReTimer.
• Now uses the audio directly without performing additional conversions.

"Fase5":

• Removed the deletion of "Temp.wav" as it is no longer created.

(If you want to use "Whisper" in "Auto Sub ReTimer GUI" you need to download it and place the "Faster-Whisper-XXL" folder inside the "Auto Sub ReTimer" folder.)

Whisper taken from https://github.com/Purfview/whisper-standalone-win

Full Changelog: V1.8...V1.8.1

Auto Sub ReTimer GUI (GPU-CPU) V1.8

Auto Sub Retimer

General Fixes for all scripts:

• Now there is no need to rename the video to "ep.mkv", the ".mkv" extension will be read directly, regardless of the name.
• OnTop lines will now be processed from Auto Sub Retimer (but not all).

Fase0:

• Improved tag detection for .srt and .ass files.

• Previously, OnTop lines (positioned at the top of the screen) that did NOT overlap would still go to "OnTop.ass/srt" without being processed to improve timing. Now, they will remain in "Dialogues" for processing. Lines with temporal overlap will remain in "OnTop" as before. (If by chance the ontop line is a sign and is not placed in "Signs" because it has few tags, its timing may be adjusted).

• Dialog lines that overlap in time with another dialog line (which are not OnTop, tag or styles) are now also handled. If there is an overlap, the line that creates the overlap will be put on OnTop.ass/srt (temporally or chronologically, depending on the situation) because it cannot be handled by Auto Sub ReTimer.

"Fase3":

• Improved calculations based on the video frame rate automatically detected by Pyscenedetect.
  Now processes subtitles directly with the detected video frame rate without intermediate conversions and without creating additional files (more precise).

• Removed some functions that are no longer needed

"Fase4":

• Fixed an issue in a function that could cause lines to become 0ms long, it now has protections to prevent this from happening.

• Fixed an issue where the "Max range to detect scene change before the initial timestamp (ms)" and "Max range to detect scene change after the initial timestamp (ms)" values in the GUI were not applying correctly, remaining fixed at the hardcoded base value (fallback).

• Changed some hardcoded values to make them slightly more aggressive. This function looks for lines where the initial timestamp is too close to a speech peak (which could cause them to appear delayed visually on video) and inserts a lead-in to compensate for the delay.
  Consequently, the "leftovers" given by "Fase2" will be corrected where possible by "Fase4".

Whisper:

• Changed some parameters to improve transcription and basic timing.

Whisper ReTiming:

• Added two new functions to adapt to the new "Whisper" parameters to improve timing.

Thanks to @Adrikss for the tests (@Adriks11 on Telegram).

Test:

• Added a "Test Snapping like Aegisub" folder.
  It allows you to automatically snap lines (timestamps snap to the closest frame, just like aegi does).
  It's not perfect. If anyone has any suggestions or finds any errors, message me on Telegram @DaiBakaAndry.

(If you want to test, copy the two files "Fase3" and "Fase4" and replace them with the existing ones in "Auto Sub ReTimer\Scripts\Migliora il Timing Dei Sub")

(If you want to use "Whisper" in "Auto Sub ReTimer GUI" you need to download it and place the "Faster-Whisper-XXL" folder inside the "Auto Sub ReTimer" folder.)

Whisper taken from https://github.com/Purfview/whisper-standalone-win

Full Changelog: V1.7...V1.8

Auto Sub ReTimer GUI (GPU-CPU) v1.7

GUI.py:

• 4 new GUI settings

1: "Max gap 'empty' between two lines to attach (ms)".
This setting allows you to configure the threshold for deciding how much "empty" space there must be to merge two subtitle lines. Merging the lines improves fluidity between them (this was already possible before, with a default value of 300, but it was not configurable via the GUI).
If you leave it at 230, the lines won’t stretch too much to attach to the next line, but there will be a natural "break" based on the fluidity of the speech. (This is obviously a personal preference, but you can now adjust the threshold to your liking).

2: "Max range to detect scene change before the initial timestamp (ms)".
Allows you to set the maximum distance to detect a scene change and link it before the initial timestamp of the line.

3: "Max range to detect scene change after the initial timestamp (ms)".
Allows you to set the maximum distance to detect a scene change and link it after the initial timestamp of the line.

4: (Config Whisper)
Allows you to change the language and select your preferred Whisper model.

Whisper ReTimer:

• Updated "Fase0", "Fase3", and "Fase4" with the latest changes in "Auto Sub ReTimer".

Fase1.py:

• Removed a useless step in Demucs (Slightly faster process).

Fase2:

• Changed the Db to detect audio peaks in speech.
  It's a small change, but it significantly alters how peaks are detected.
  The improvement mainly applies to low and nuanced speech note.
  Obviously, it’s not perfect, but it should be far better than before.
  This means you will have to find your preferred values ​​again if you had customized them in the GUI.

Fase3:

• Removed some deprecated parameters (Slightly faster process).

Fase4:

• Removed the modified function from v1.6 and reinstated the one from v1.5, which caused fewer errors.
  The new modified function, based on the v1.5 version, is now "smarter". It should make far fewer errors than before. (Errors still depend on the crappiness of the loaded subtitle timing and false positives/undetected scenes from pyscenedetect).

• More...

(If you want to use "Whisper" in "Auto Sub ReTimer GUI" you need to download it and place the "Faster-Whisper-XXL" folder inside the "Auto Sub ReTimer" folder.)

Whisper taken from https://github.com/Purfview/whisper-standalone-win

Full Changelog: V1.6.2...V1.7

Auto Sub ReTimer GUI (GPU-CPU) v1.6.2

Fix and improvements Auto Sub ReTimer:

Fase2:

• Removed the unnecessary ".wav" conversion from the very first versions of the script, now it reads the necessary file directly, speeding up the process.

Fase4:

• "min_peaks_count" set to 3 instead of 2. This results in fewer errors in linking the final time stamps of lines to scene changes.

Fase6:

• Removed the deletion of "temp.wav" since "Fase2" no longer creates these files.

(If you want to use "Whisper" in "Auto Sub ReTimer GUI" you need to download it and place the "Faster-Whisper-XXL" folder inside the "Auto Sub ReTimer" folder.)

Whisper taken from https://github.com/Purfview/whisper-standalone-win

Full Changelog: V1.6.1...V1.6.2

Auto Sub ReTimer GUI (GPU-CPU) v1.6.1

Fix and improvements

GUI:

• Added a timer to keep track of how long "Auto Sub ReTimer" will take to complete all the "Fase."
  (Not implemented in "Whisper" and "Whisper ReTimer")

Auto Sub ReTimer:

Fase3:

• Adjusted the progress bar. It will no longer "go crazy" during scene change detection. (Slightly faster because it no longer has to manage multiple progress bars for each video segment).

Whisper ReTimer:

• Updated "Fase3" as in the latest version 1.6 of "Auto Sub ReTimer."

• Updated "Fase4" as in the latest version 1.6 of "Auto Sub ReTimer."

(If you want to use "Whisper" in "Auto Sub ReTimer GUI" you need to download it and place the "Faster-Whisper-XXL" folder inside the "Auto Sub ReTimer" folder.)

Whisper taken from https://github.com/Purfview/whisper-standalone-win

Full Changelog: V1.6...V1.6.1

Auto Sub ReTimer GUI (GPU-CPU) v1.6

Fase3:

• The progress bar in "Fase3" is now visible in the GUI.

• Improved scene change detection and reduced false positives (removed ContentDetector which only caused false positives with almost no benefit).

• Implemented video parallelization in pyscenedetect, making it significantly faster.
  In short, based on the loaded subtitle lines, it will detect which parts of the video to discard and which parts to feed to pyscenedetect (without cutting the video, it relies on the timing of the lines).
  Thus, not only does it reduce scene change detection time by not processing the entire video, but the process is also speed up by parallelizing only the relevant video segments.

• Implemented logic to prevent memory leaks during the process.

• The progress bar may now "go crazy" when "Fase3" starts, as multiple video segments are processed simultaneously at high speed, resulting in multiple active progress bars at the same time (this is not an issue if you see this behavior, it's normal).

Fase4:

• Improved the logic of a function that caused issues in previous versions. Now, with the new "Fase3" and the new "Fase4," you'll have fewer lines to adjust in the final file (of course, it’s not perfect and can still make mistakes. Remember that there are configurable values in the GUI adjust them according to your timing preferences).

• Swapped "Fase0" with "Fase1".
  Now, when you press "Start" you’ll immediately be prompted to load the subtitles for processing. This reduces user interaction (previously, you had to wait for the original "Fase0" to finish before loading the subs, meaning you had to wait for it to complete, now it’s more immediate).

Fase6:

• Now, you’ll be asked whether you want to merge "On top.ass/.srt," "Comments.ass" and "Signs.ass" into the final file. If you choose not to merge them, the separate files will also be moved to the desktop.

The "On Top" lines will be inserted into "Final.ass/.srt" in timestamp order.

The "Comments" lines will be placed at the very top.

The "Signs" lines will be placed at the bottom, after the dialogue.

• If "On Top.ass/.srt," "Comments," and "Signs" are empty, they will no longer be moved to the desktop. (Previously, they were moved even if empty. The user would have deleted them afterward anyway.)

Fase0:

• Fixed the duplicate "[Events]" issue that occurred when separating "On Top.ass," "Comments.ass" and "Signs.ass," which also appeared in the "Final.ass" file.

(If you want to use "Whisper" in "Auto Sub ReTimer GUI" you need to download it and place the "Faster-Whisper-XXL" folder inside the "Auto Sub ReTimer" folder.)

Whisper taken from https://github.com/Purfview/whisper-standalone-win

Full Changelog: V1.5...V1.6

Auto Sub ReTimer GUI (GPU-CPU) v1.5

Fase1:

Added separation of "Signs" and "Comments" from the .ass subtitles loaded in "Fase1".

(No need to manually sort everything anymore, the program does it all for you)

If the program detects lines that it can't determine as "Signs" or "Dialogue," it will open a window and let you decide where to place them (it also shows the style name of the lines for better clarity).

Now, your final files will be:

• "Comments": Contains commented lines and empty lines (necessary separation to adjust dialogue timing).

• "Signs": Contains all signs, opening and ending subtitles (necessary separation to adjust dialogue timing).

• "On Top": Contains all dialogues positioned at the top of the screen (necessary separation to adjust dialogue timing).

• "Final": Subtitles with adjusted timing.

Fase6:

• Added new files to be moved to the desktop (if an .ass subtitle file was loaded).

GUI:

• Added "Fase1" information to the GUI.

(If you want to use "Whisper" in "Auto Sub ReTimer GUI" you need to download it and place the "Faster-Whisper-XXL" folder inside the "Auto Sub ReTimer" folder.)

Whisper taken from https://github.com/Purfview/whisper-standalone-win

Full Changelog: v1.1...V1.5

Auto Sub ReTimer GUI (GPU-CPU) v1.1

• Improved "Fase3" to better detect scene changes in dark videos, minimizing false positives as much as possible (even in generic scenarios).

• The cache is no longer deleted, consequently, after the first use of the program, subsequent uses will be faster.

• Refined some values.

• Updated whisper package to latest version.

(If you want to use "Whisper" in "Auto Sub ReTimer GUI" you need to download it and place the "Faster-Whisper-XXL" folder inside the "Auto Sub ReTimer" folder.)

Full Changelog: v1.0...v1.1

Auto Sub ReTimer GUI (GPU-CPU)

• Added a GUI that simplifies everything.

• Added the settings for "Fase2" and "Fase4" in the GUI, which previously had to be manually modified by opening the corresponding files.

• Now everything is automatic; there's no need to use the site to extract "vocals" with Demucs, it does everything automatically and locally.

• Added GPU support as an optional choice for the user. Check the documentation on the main GitHub page.

• Files are better organized.

(If you want to use "Whisper" in "Auto Sub ReTimer GUI" you need to download it and place the "Faster-Whisper-XXL" folder inside the "Auto Sub ReTimer" folder.)