pdjdev · Oct 17, 2026 · Oct 14, 2026 · Oct 14, 2026 · Oct 14, 2026 · Oct 14, 2026
diff --git a/Changelog.txt b/Changelog.txt
@@ -1,5 +1,77 @@
 Subtitle Edit Changelog
 
+5.3.0 (14th of October 2026)
+
+  Subtitle Edit 5.3 collects the work of 26 betas and 4 release candidates.
+  A summary of the changes since 5.2.0:
+
+New features:
+* Command palette (Ctrl+Shift+K)
+* Import subtitles directly from DVD (IFO/VOB) with title picker, fast rip and exact cell timing
+* Improve time codes (forced alignment) - re-time a roughly synced subtitle, with a speech-to-text check and waveform preview
+* Remux video - video, multi-track audio (with mixing) and soft subtitles, also embedded CEA-608 captions
+* Cut video transitions and fades, burn-in for selected lines, and 3D (half side-by-side / top-bottom) burn-in and image export
+* Generate background music locally (ACE-Step)
+* Compare: merge view, sync points, time tolerance and taking only text or timing per row
+* Custom shortcut slots built from steps, AltGr modifier, per-plugin shortcuts and many new shortcuts
+* Actor picker, "Set style..." quick pick, and "Shot in"/"Shot out" grid columns
+* Reload the subtitle when it is changed outside Subtitle Edit, and "Frame numbers" time code mode
+* Taskbar/Dock progress and prevent sleep for long jobs (speech to text, text to speech, burn-in, batch convert)
+
+Video player and waveform:
+* New FFmpeg video player with hardware decoding, audio track picker, real frame stepping and Linux sound
+* Experimental mpv Metal video player on macOS, deinterlace setting, and customizable player controls
+* Editor-style layout 14 with a track timeline, and layouts 12 and 13 (text box below the video)
+* Experimental fast waveform renderer, "Show speech only" waveform, grid color and remembered zoom
+* Waveform toolbar: initial text box, move-lines buttons, line breaks and wrapping
+
+Speech to text:
+* Crisp ASR: speaker detection (Sortformer), VAD choice (Silero, FireRedVAD, WebRTC, Whisper-VAD) and "Isolate speech"
+* New models: Parakeet Ultra, Phonon-2, ReazonSpeech/Japanese Parakeet, Voxtral F16 and Index-Echo
+* OpenAI-compatible API works with vLLM; retry of failed clips on selected lines
+
+Text to speech:
+* New engines: OpenAI-compatible TTS, ElevenLabs v4/v4 Turbo, KugelAudio and Supertonic-3
+* Voice manager with transcripts, voice packs and copying voices between engines
+* "Remove original speech", "Speak from current line" and "Play with speech"
+* Optional Sidon cleaning of voice-clone reference audio
+
+Auto-translate and AI:
+* New engines: LM Studio advanced and API Route, model browsing, and new llama.cpp models (Index-Translate, TranslatePsy-AfriSLM)
+* Translate in place, translate one file into several languages in batch convert, and a no-prompt shortcut
+* 22 more languages for the AI engines, MyMemory without an API key, and line breaks kept for ChatGPT/Anthropic
+* AI review: prompt templates and reference context
+
+OCR:
+* nOCR: much better accuracy, new Cyrillic database, "Optimize database", and a redesigned trainer
+* Find and Replace in the OCR window, remembered pre-processing settings, and a per-language spell check dictionary
+* Romanian and Croatian OCR fix lists; better word splitting
+
+Subtitle formats:
+* Read CEA-608/708 closed captions from .ts, .mkv, .mp4, .mxf, .vob, .mpg and .m2v
+* New: EZTitles binary (.ezt), PSP UMD Video, HD-DVD sup, Cheetah Caption ASC, ELR Studio and more SE 4 formats opened again
+* Transport stream, MP4 caption, CEA-708 and Blu-ray sup timing fixes; PAC code page detection and secondary code page
+* ASSA draw: redesigned with layers, shape library, curve tools, SVG import and live preview
+* ASSA styles: overwrite/keep both on import, reorder, font collector; Source view syntax highlighting
+
+Batch convert and seconv:
+* Batch convert: named presets, new functions and filters, multi-language translate, OCR language per file, and many format settings
+* seconv: DVD sup and IMSC image outputs, openai-compatible and DeepL translate, nOCR optimize, and more SE 4 formats
+
+Accessibility and UI:
+* Screen reader names across settings and tool windows, F1 help in 145 more dialogs, and remembered dialog sizes/positions
+* Settings show one category at a time with search and back/forward navigation; settings auto-backup
+* macOS-standard default shortcuts and a new rounded app icon
+
+Translations and engines:
+* All language files synced with English
+* Updated engines: Crisp ASR v0.8.43, whisper.cpp v1.9.5, llama.cpp b11553, CrispEmbed v0.17.12, audio.cpp v0.9.1, WhisperX build 103, FFmpeg 9.0.2 (Windows), libVLC 3.0.24 (Windows), Avalonia 12.1.4 and Flatpak runtime 26.08
+
+  A big thank you to everyone who tested the pre-releases, reported issues and
+  contributed code and translations - and a special thanks to Anthropic for
+  sponsoring a Claude subscription used in the development of this release :)
+
+
 5.2.0 (10th of September 2026)
 
   Subtitle Edit 5.2 collects the work of 32 betas and 7 release candidates.

diff --git a/Dictionaries/ru_NoBreakAfterList.xml b/Dictionaries/ru_NoBreakAfterList.xml
@@ -15,4 +15,92 @@
   <Item>г-жа</Item>
   <Item>г-жа.</Item>
   <Item>г-ца.</Item>
-</NoBreakAfterList>
+  <Item>в</Item>
+  <Item>В</Item>
+  <Item>во</Item>
+  <Item>Во</Item>
+  <Item>на</Item>
+  <Item>На</Item>
+  <Item>с</Item>
+  <Item>С</Item>
+  <Item>со</Item>
+  <Item>Со</Item>
+  <Item>к</Item>
+  <Item>К</Item>
+  <Item>ко</Item>
+  <Item>Ко</Item>
+  <Item>о</Item>
+  <Item>О</Item>
+  <Item>об</Item>
+  <Item>Об</Item>
+  <Item>обо</Item>
+  <Item>Обо</Item>
+  <Item>от</Item>
+  <Item>От</Item>
+  <Item>ото</Item>
+  <Item>Ото</Item>
+  <Item>по</Item>
+  <Item>По</Item>
+  <Item>у</Item>
+  <Item>У</Item>
+  <Item>за</Item>
+  <Item>За</Item>
+  <Item>из</Item>
+  <Item>Из</Item>
+  <Item>изо</Item>
+  <Item>Изо</Item>
+  <Item>до</Item>
+  <Item>До</Item>
+  <Item>для</Item>
+  <Item>Для</Item>
+  <Item>без</Item>
+  <Item>Без</Item>
+  <Item>под</Item>
+  <Item>Под</Item>
+  <Item>подо</Item>
+  <Item>Подо</Item>
+  <Item>над</Item>
+  <Item>Над</Item>
+  <Item>при</Item>
+  <Item>При</Item>
+  <Item>про</Item>
+  <Item>Про</Item>
+  <Item>через</Item>
+  <Item>Через</Item>
+  <Item>перед</Item>
+  <Item>Перед</Item>
+  <Item>из-за</Item>
+  <Item>Из-за</Item>
+  <Item>из-под</Item>
+  <Item>Из-под</Item>
+  <Item>и</Item>
+  <Item>И</Item>
+  <Item>а</Item>
+  <Item>А</Item>
+  <Item>но</Item>
+  <Item>Но</Item>
+  <Item>да</Item>
+  <Item>Да</Item>
+  <Item>или</Item>
+  <Item>Или</Item>
+  <Item>либо</Item>
+  <Item>Либо</Item>
+  <Item>ни</Item>
+  <Item>Ни</Item>
+  <Item>что</Item>
+  <Item>Что</Item>
+  <Item>чтобы</Item>
+  <Item>Чтобы</Item>
+  <Item>как</Item>
+  <Item>Как</Item>
+  <Item>если</Item>
+  <Item>Если</Item>
+  <Item>когда</Item>
+  <Item>Когда</Item>
+  <Item>то</Item>
+  <Item>То</Item>
+  <Item>не</Item>
+  <Item>Не</Item>
+  <Item>я</Item>
+  <Item>Я</Item>
+</NoBreakAfterList>
diff --git a/Dictionaries/rus_OCRFixReplaceList.xml b/Dictionaries/rus_OCRFixReplaceList.xml
@@ -243,6 +243,8 @@
   <PartialWords>
     <WordPart from="Є" to="Е" />
     <WordPart from="ЬІ" to="Ы" />
+    <WordPart from="ьІ" to="ы" />
+    <WordPart from="ьі" to="ы" />
     <WordPart from="КЗ" to="КА" />
     <WordPart from="ЛЗ" to="ЛА" />
     <WordPart from="НЗ" to="НА" />
@@ -282,5 +284,8 @@
     <RegEx find="(?&lt;=[а-яё])3(?![0-9])" replaceWith="з" />
     <RegEx find="(?&lt;=[а-яё])6(?![0-9])" replaceWith="б" />
     <RegEx find="(?&lt;=[а-яё])0(?![0-9])" replaceWith="о" />
+    <!-- 6 вместо б в начале слова перед строчной кириллицей: 6ыл - был, со6аки
+         уже исправлено выше. Число перед словом (6 дней) отделено пробелом. -->
+    <RegEx find="(?&lt;![0-9А-Яа-яЁё])6(?=[а-яё])" replaceWith="б" />
   </RegularExpressions>
 </OCRFixReplaceList>
diff --git a/Ocr/Cyrillic.nocr b/Ocr/Cyrillic.nocr
diff --git a/Ocr/Latin.nocr b/Ocr/Latin.nocr
diff --git a/change-log.txt b/change-log.txt
@@ -1,6 +1,161 @@
 
 Subtitle Edit Changelog
 
+-----------------------------------------------------------------------------------------------------
+
+v5.3.0 (14th of October 2026)
+
+Subtitle Edit 5.3 collects the work of 26 betas and 4 release candidates.
+A summary of the changes since v5.2.0 (full details in the pre-release entries below):
+
+New features:
+* Command palette (Ctrl+Shift+K)
+* Import subtitles directly from DVD (IFO/VOB) with title picker, fast rip and exact cell timing
+* Improve time codes (forced alignment) - re-time a roughly synced subtitle, with a speech-to-text check and waveform preview
+* Remux video - video, multi-track audio (with mixing) and soft subtitles, also embedded CEA-608 captions
+* Cut video transitions and fades, burn-in for selected lines, and 3D (half side-by-side / top-bottom) burn-in and image export
+* Generate background music locally (ACE-Step)
+* Compare: merge view, sync points, time tolerance and taking only text or timing per row
+* Custom shortcut slots built from steps, AltGr modifier, per-plugin shortcuts and many new shortcuts
+* Actor picker, "Set style..." quick pick, and "Shot in"/"Shot out" grid columns
+* Reload the subtitle when it is changed outside Subtitle Edit, and "Frame numbers" time code mode
+* Taskbar/Dock progress and prevent sleep for long jobs (speech to text, text to speech, burn-in, batch convert)
+
+Video player and waveform:
+* New FFmpeg video player with hardware decoding, audio track picker, real frame stepping and Linux sound
+* Experimental mpv Metal video player on macOS, deinterlace setting, and customizable player controls
+* Editor-style layout 14 with a track timeline, and layouts 12 and 13 (text box below the video)
+* Experimental fast waveform renderer, "Show speech only" waveform, grid color and remembered zoom
+* Waveform toolbar: initial text box, move-lines buttons, line breaks and wrapping
+
+Speech to text:
+* Crisp ASR: speaker detection (Sortformer), VAD choice (Silero, FireRedVAD, WebRTC, Whisper-VAD) and "Isolate speech"
+* New models: Parakeet Ultra, Phonon-2, ReazonSpeech/Japanese Parakeet, Voxtral F16 and Index-Echo
+* OpenAI-compatible API works with vLLM; retry of failed clips on selected lines
+
+Text to speech:
+* New engines: OpenAI-compatible TTS, ElevenLabs v4/v4 Turbo, KugelAudio and Supertonic-3
+* Voice manager with transcripts, voice packs and copying voices between engines
+* "Remove original speech", "Speak from current line" and "Play with speech"
+* Optional Sidon cleaning of voice-clone reference audio
+
+Auto-translate and AI:
+* New engines: LM Studio advanced and API Route, model browsing, and new llama.cpp models (Index-Translate, TranslatePsy-AfriSLM)
+* Translate in place, translate one file into several languages in batch convert, and a no-prompt shortcut
+* 22 more languages for the AI engines, MyMemory without an API key, and line breaks kept for ChatGPT/Anthropic
+* AI review: prompt templates and reference context
+
+OCR:
+* nOCR: much better accuracy, new Cyrillic database, "Optimize database", and a redesigned trainer
+* Find and Replace in the OCR window, remembered pre-processing settings, and a per-language spell check dictionary
+* Romanian and Croatian OCR fix lists; better word splitting
+
+Subtitle formats:
+* Read CEA-608/708 closed captions from .ts, .mkv, .mp4, .mxf, .vob, .mpg and .m2v
+* New: EZTitles binary (.ezt), PSP UMD Video, HD-DVD sup, Cheetah Caption ASC, ELR Studio and more SE 4 formats opened again
+* Transport stream, MP4 caption, CEA-708 and Blu-ray sup timing fixes; PAC code page detection and secondary code page
+* ASSA draw: redesigned with layers, shape library, curve tools, SVG import and live preview
+* ASSA styles: overwrite/keep both on import, reorder, font collector; Source view syntax highlighting
+
+Batch convert and seconv:
+* Batch convert: named presets, new functions and filters, multi-language translate, OCR language per file, and many format settings
+* seconv: DVD sup and IMSC image outputs, openai-compatible and DeepL translate, nOCR optimize, and more SE 4 formats
+
+Accessibility and UI:
+* Screen reader names across settings and tool windows, F1 help in 145 more dialogs, and remembered dialog sizes/positions
+* Settings show one category at a time with search and back/forward navigation; settings auto-backup
+* macOS-standard default shortcuts and a new rounded app icon
+
+Translations and engines:
+* All language files synced with English
+* Updated engines: Crisp ASR v0.8.43, whisper.cpp v1.9.5, llama.cpp b11553, CrispEmbed v0.17.12, audio.cpp v0.9.1, WhisperX build 103, FFmpeg 9.0.2 (Windows), libVLC 3.0.24 (Windows), Avalonia 12.1.4 and Flatpak runtime 26.08
+
+A big thank you to everyone who tested the pre-releases, reported issues and
+contributed code and translations - and a special thanks to Anthropic for
+sponsoring a Claude subscription used in the development of this release :)
+
+-----------------------------------------------------------------------------------------------------
+
+v5.3.0-rc4 (13th of October 2026)
+
+* Add Whisper-VAD to the Crisp ASR VAD choices - thx jodilife
+* macOS: show "Search help topics" field with translated menus - thx GitGianluc
+* Fix jitter/crackling in layout 14 (filmstrip thumbnails decode keyframes only) - thx ruonghieu
+* Fix context menu submenu icons not showing on first open - thx GitGianluc
+* Fix Linux/OpenGL fullscreen leaving an mpv instance alive (VRAM leak) - thx vitonique
+* Fix Whisper OpenAI model download "path not found" - thx Shaima-Almarzooqi
+* Fix Auto-translate advanced settings cut off at high UI scaling - thx mmmsh-125
+* Update whisper.cpp to v1.9.5
+* Update Crisp ASR to v0.8.43
+* Update Czech translation - thx Matěj
+* Update French translation - thx Need74
+* Update German translation - thx FunkyKnilch
+* Update Italian translation - thx bovirus
+* Update Japanese translation - thx hisui3393
+* Update Korean translation - thx 12si27
+
+
+-----------------------------------------------------------------------------------------------------
+
+v5.3.0-rc3 (12th of October 2026)
+
+* Add nOCR Cyrillic database (Cyrillic.nocr) with case and "ы" handling
+* Add nOCR "Optimize database" (OCR window + seconv optimize-nocr)
+* Add "Fast (HTDemucs)" model option for TTS "Remove original speech" - thx invio-a11y
+* Improve time codes: fix mistimed overlaps kept, skip lyrics/positioned signs/comments, snap to frames in frame mode
+* Import plain text: hold ends for max CPS, use the Gap setting, allow cancel of alignment
+* Improve Latin.nocr (drop lines that don't tell glyphs apart from lookalikes)
+* Improve help: F1 help in 35 more dialogs, docs updates
+* Batch convert: VobSub uses the profile's font and outline color - thx mehdittttt
+* Batch convert: keep fades in Blu-ray sup and alignment in D-Cinema PNG
+* Blu-ray sup: bridge gaps shorter than half a frame
+* Fix settings export/import/reset missing sections - thx FunkyKnilch
+* Fix list scroll bars covering item content (e.g. command palette shortcuts) - thx GrampaWildWilly
+* Fix OCR upgrade replacing user-trained/optimized nOCR databases
+* Fix high memory use with image subtitles in Point sync via other
+* Fix crash risk when clicking OK while the Matroska track picker preview is loading
+* Fix "New line format" placeholder in Custom text format - thx GitGianluc
+* Update llama.cpp to b11553
+* Update French translation - thx Need74
+* Update Italian translation - thx bovirus
+* Update Japanese translation - thx hisui3393
+* Update Polish translation - thx potplayer-fanpack
+* Update Turkish translation - thx bilimiyorum
+
+
+-----------------------------------------------------------------------------------------------------
+
+v5.3.0-rc2 (11th of October 2026)
+
+* Add command palette (Ctrl+Shift+K)
+* Add experimental mpv Metal video player on macOS
+* ASSA draw: new layout with menus, undo/redo, scale/rotate, line/curve tools, backgrounds and SVG import
+* ASSA draw: add shape tool with shape library (speech bubbles, arrows, symbols) and eyedropper
+* Train nOCR window: new layout with preview, start from an existing database, several font sizes and progress
+* Improve nOCR accuracy (scale-aware matching, auto spaces, shadow threshold, better Latin.nocr)
+* Point sync via other subtitle: load subtitles embedded in mkv/mp4 and .sup files - thx mmmsh-125
+* Find/Replace: delete individual history entries - thx FunkyKnilch
+* OCR: remember image pre-processing settings
+* Format combo: show full format names in the drop-down
+* Improve layout for long translated texts in Spell check and Batch convert - thx Slava46
+* Hide "Close video file" menu item when no video is loaded
+* Improve performance of Netflix check, spell check, shot change extend, Video OCR, VobSub writing and more
+* Fix "Could not convert to double" when typing a frame rate in Change frame rate - thx jojorne
+* Fix Video OCR ignoring the llama.cpp external server setting - thx GitGerby
+* Fix SSA losing [Events] header when [Fonts]/[Graphics] precede it
+* Fix TTML italic/bold/underline inheritance
+* Fix MS Office Workbook, D-Cinema Interop, Projection Subtitle List, FCP XML and IMSC/TTML frame-rate issues
+* Fix Apply duration limits ignoring unticked rows and Modify selection duration defaults
+* Fix OCR fix stripping accents from English loanwords/names
+* Fix shot changes not sorted when assigned or loaded
+* Fix nOCR trainer Stop button and closing saving a partial database
+* Fix image export "Box per line" boxes shifted left
+* Update SharpCompress to 0.50.5
+* Update German translation - thx FunkyKnilch
+* Update Polish translation - thx potplayer-fanpack
+* Update Russian translation - thx LazySwift
+
+
 -----------------------------------------------------------------------------------------------------
 
 v5.3.0-rc1 (10th of October 2026)