Subtitle Edit 5.3.0 released

[niksedk]

Subtitle Edit 5.3.0 is out - 34 days after the 5.2.0 release, collecting the work of 26 betas and 4 release candidates.

What's new since 5.2.0

• Closed captions: read CEA-608/708 from .ts, .mkv, .mp4, .mxf, .vob and .mpg, plus EZTitles binary, PSP UMD and HD-DVD sup
• Speech to text: speaker detection, a choice of VAD, "Isolate speech", and new Parakeet, Phonon-2 and Index-Echo models
• Text to speech: OpenAI-compatible TTS, ElevenLabs v4, KugelAudio, Breeze TTS 2, and Supertonic-3, voice manager
• OCR: much more accurate nOCR, a new Cyrillic database, "Optimize database" and a redesigned trainer
• Auto-translate: LM Studio and API Route engines, translate in place, and batch translation into several languages
• Command palette (Ctrl+Shift+K): find and run any command by name
• ASSA draw: redesigned with layers, a shape library (speech bubbles, arrows, symbols), curve tools, SVG import and live preview
• Improve time codes: re-time a roughly synced subtitle with forced alignment, checked by speech to text
• Remux video: combine video, multi-track audio (with mixing) and soft subtitles - plus Cut video transitions/fades and burn-in of - Import from DVD: open IFO/VOB directly with a title picker, fast rip and exact cell timing
  selected lines
• New FFmpeg video player with hardware decoding, plus an editor-style layout 14 with a track timeline and an experimental mpv Metal player on macOS
• Accessibility: screen reader names across settings and tool windows, and F1 help in 145 more dialogs
• Bug fixes and minor improvements

---

Command palette

ASSA draw with the shape library

Import subtitles from DVD

Remux video

Full details in the change log.

Download

Grab it from the v5.3.0 release page.

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 :)

[santoshsachin93-wq]

Hi guys! I've been using Subtitle Edit for a few months now. I'm not an expert or a professional, so I'd appreciate your help in finding the best settings for converting Japanese movie video into English subtitles.

My ultimate goal is to get the best possible English subtitles for Japanese movies that are around 2 hours and 30 minutes long. To save time while testing different settings, I've cut three short clips of approximately 1 minute and 30 seconds each from different parts of a movie. This allows me to compare the results without having to process the entire movie every time. I've tried a few different settings, but the best results I've achieved so far are only okay. I've tried Large-v3, but it actually gave me poorer speech-to-text recognition than Large-v2.
For now, my main problem is that Subtitle Edit is skipping some Japanese dialogue entirely during speech recognition, rather than producing inaccurate translations. I'd like to improve the Japanese transcription first before focusing on translation quality. Are there any settings, models, or techniques you'd recommend for improving Japanese speech recognition in Subtitle Edit, particularly for preventing spoken dialogue from being missed? I'd appreciate any suggestions or recommendations based on your experience.

My current settings in Subtitle Edit v5.2.0 are:

1. Speech to Text
   Engine: Purfview Faster Whisper XXL
   Input language: Japanese
   Model: Large-v2 (2.9 GB)
   I use this to generate Japanese subtitles first.

2. Auto Translate
   Engine: llama.cpp (local LLM)
   Translation: Japanese to English
   Model: TranslateGemma 12B (8.5 GB)

The results are okay, but I'd like to achieve much better translation accuracy. I'm really grateful that Subtitle Edit exists, and I have a lot of respect for all the work that the developers and contributors put into it. I'm just wondering whether my current setup is already giving me the best possible results or whether there's something I could improve. I also have access to Google Gemini 3.1 Pro for one year through an offer, so I'd like to make the most of that subscription if it can help improve my results.