Subtitle Edit 5.4.0 Released: New Audio Sync, Translation and OCR Tools

Where to get it and what platforms it covers​

The GitHub release lists these packages:

• A Windows x64 installer (Inno Setup) and a Windows x64 ZIP.
• A Windows ARM64 ZIP.
• macOS builds for Apple Silicon and Intel.
• Linux x64 and ARM64 tarballs, plus an x64 Flatpak.
• SeConv command-line converter builds for the same platforms.

Subtitle Edit 5 is the cross-platform generation. It is built with Avalonia UI and .NET 10. Windows users get the same codebase as everyone else, so this is a modern UI toolkit app and not the old WinForms-era Subtitle Edit 4.

Download sizes vary by source. Neowin lists roughly 100 MB. The third-party site subtitleedit.org says the installer is 58.8 MB and the portable ZIP is 111 MB. The GitHub release page I reviewed gives file sizes for the SeConv assets but not the Windows installer, so treat any single figure with caution and check the file you actually download.

What's new in 5.4.0​

The changelog highlights, as relayed by Neowin, cover a lot of ground. Here is the version grouped by task.

Look and feel​

• A more compact, consistent interface with live theme switching and new SVG icons.
• New themes: Nord, Midnight Cinema and Sakura.
• A visual theme picker with preview cards, plus a shortcut to cycle through themes.

The release-candidate notes say the SVG icons are meant to stay sharp at any display scale. That should help on mixed-DPI Windows setups.

Syncing and timing​

• Subtitle synchronization with audio, including automatic offset and frame-rate detection. Beta notes describe this as ffsubsync-style detection, found under Video > More.
• Audio-track synchronization with live preview.
• More keyboard shortcuts: splitting lines, frame-by-frame movement, adjusting neighboring subtitle timings, and copying plain text.
• Video player upgrades: playback speed controls, previous/next subtitle buttons and subtitle toggling.
• Waveform editing can now move and resize ASSA lines together when they share identical timings.

Translation and AI​

• Source and translated text are shown side by side.
• Selected rows can be translated together with context.
• Find-and-replace now works in the translation column.
• A new Opper translation engine and AI-generated synopsis support. The synopsis feature feeds llama.cpp advanced translation.
• Updates to the existing AI and speech-recognition engines.
• Microphone speech recording straight into the subtitle text box.

OCR​

• An optional video preview in the OCR window.
• OCR limited to chosen time ranges.
• New llama.cpp OCR models, including LightOnOCR-3, LFM3-VL 2B and QARI-OCR for Arabic.
• Better Latin and Cyrillic recognition.

Formats and export​

• Support for YouTube timed-text (.ytt/srv3), with pen styles, positioning, previews and a styles window.
• Export of D-Cinema SMPTE 2014 MXF tracks and Excel dialogue lists (.xlsx).
• Better image-based exports, with more ASSA tag support. The RC4 notes list \r, \rStyle, \fsp, \blur, \be, \fscx, \fscy and \frz, plus underline and strikeout.
• A rewritten SubRip reader.

ASSA editing​

• Text and shape drawing, groups, in/out and loop animations, and new clip art.
• Six CapCut-style caption effects.
• Redesigned style and property editors, with previews of fonts attached under [Fonts].

Smaller workflow changes​

• Custom error rules using regular expressions.
• Automatic line wrapping while typing.
• Prompts when a subtitle file is modified externally.
• Multiple Replace and spell-check improvements.
• Batch Convert and plug-in discovery windows with search.
• New Romanian and Croatian interface translations. The Croatian one was flagged in RC2 as a machine draft that needs review.

The SubRip claim, put in context​

The headline performance claim is that the rewritten SubRip reader is 2.6–12 times faster and more tolerant of damaged subtitle files. That is the project's own figure, and I haven't seen an independent benchmark. The range is wide, so the speedup probably depends on the file. It matters most if you batch-process large libraries of .srt files. The project also says it fixed parser crashes found through fuzzing. That matters for anyone opening subtitles from untrusted sources.

Auto-update: one claim to treat carefully​

Neowin lists "automatic update installation with restart support" on Windows, Linux and macOS. The project changelog I reviewed did not confirm that line. The third-party download site does say the installer includes Start menu shortcuts and auto-update support. That is a vendor-style page and not a primary source. If you use the portable ZIP, expect to update manually until you've confirmed the behavior yourself.

Upgrade advice​

1. Pick the right package. Use the x64 installer for most PCs, the ZIP for portable use, or the ARM64 ZIP for Windows on Arm.
2. Back up your settings if you've tuned word lists, Multiple Replace rules or custom error rules. They are among the features that were reworked.
3. Test an existing project first, especially one that uses ASSA styling or image-based export. The RC4 notes say tags that can't be drawn no longer strip italic, bold, color or font styling from the exported line.
4. Try the new sync tools on a copy of your subtitle file. Automatic offset and frame-rate detection is convenient, but you should still check the result against the video.
5. Treat the AI and OCR features as optional. Several rely on downloadable models or engines, so check each one's requirements in the app before you depend on it.

Verdict​

This is a feature-heavy release for an editor that was already deep. The biggest practical gains are the audio-based sync tools, the translation workflow with context, the OCR range selection, and a SubRip reader that copes better with broken files. The new themes are cosmetic, but they show the Avalonia-based rewrite is maturing. The speed numbers and the auto-update behavior are the two claims I'd verify on your own machine before relying on them.