docs/third-party-components.md
Subtitle Edit uses several third-party tools for features like video playback, audio extraction, and OCR. While Subtitle Edit includes built-in downloaders for these components, you might want to use a specific version or a custom build.
Subtitle Edit 5 also includes more downloadable AI components for speech-to-text, text-to-speech, and OCR. Prefer the in-app download prompts unless you need to install a specific build manually.
⚠️ Warning Subtitle Edit is tested with specific versions of these components. Using other versions is not officially supported and may cause instability.
Subtitle Edit stores these components in its Data Folder.
SubtitleEdit.exe (Windows) or the executable.%APPDATA%\Subtitle Edit
Win+R, type %APPDATA%\Subtitle Edit, and hit Enter)~/.config/Subtitle Edit (or $XDG_CONFIG_HOME/Subtitle Edit).~/.config/Subtitle Edit (or $XDG_CONFIG_HOME/Subtitle Edit).Tip: You can open the Data Folder directly from Subtitle Edit by pressing
Ctrl+Alt+Shift+D(Windows/Linux) orCmd+Alt+Shift+D(macOS).
| Component | File(s) | Destination Path |
|---|---|---|
| FFmpeg | ffmpeg.exe, ffprobe.exe (optional) | [Data Folder]/ffmpeg |
| MPV | libmpv-2.dll | [Data Folder] (root) |
| yt-dlp | yt-dlp.exe | [Data Folder] (root) |
| Tesseract | tesseract.exe, tessdata/ folder | [Data Folder]/Tesseract550 |
| Whisper CPP | whisper-cli.exe, Models/ folder | [Data Folder]/SpeechToText/Cpp |
| Purfview Faster-Whisper XXL | faster-whisper-xxl.exe, _models/ folder | [Data Folder]/SpeechToText/Purfview-Faster-Whisper-XXL |
| Crisp ASR | crispasr.exe, models/ folder | [Data Folder]/CrispASR |
| Qwen3 ASR CPP | qwen3-asr-cli.exe, models/ folder | [Data Folder]/Qwen3ASR |
| Parakeet.cpp | parakeet.exe, model folders | [Data Folder]/parakeet.cpp |
| PaddleOCR | paddleocr.exe, models/ folder | [Data Folder]/OCR/PaddleOCR3-1 |
| Qwen3 TTS (CrispASR) | shares crispasr.exe + models/ from [Data Folder]/CrispASR; reference voices in voices/ | [Data Folder]/TextToSpeech/Qwen3TtsCrispAsr (voices only) |
| Chatterbox TTS (CrispASR) | shares crispasr.exe + models/ from [Data Folder]/CrispASR; reference voices in voices/ | [Data Folder]/TextToSpeech/Chatterbox (voices only) |
| OmniVoice TTS | omnivoice-tts.exe, omnivoice-codec.exe, models/, voices/ | [Data Folder]/TextToSpeech/OmniVoice |
| Kokoro TTS | kokoro-tts-server.exe, models/ | [Data Folder]/TextToSpeech/KokoroTtsCpp |
Used for reading media info, extracting audio, and generating waveforms.
gyan.dev or BtbN. Use "release-essentials" or "release-full".[Data Folder]/ffmpegffmpeg.exe from the download (usually found in a bin subfolder).ffmpeg.exe directly in [Data Folder]/ffmpeg — do not include the bin subfolder itself.ffprobe.exe can also be placed in the same folder.Settings.json (located in the Data Folder):"FfmpegPath": "C:\\path\\to\\your\\ffmpeg.exe"
\\) for Windows paths in JSON.Used as a video player engine.
mpv-dev-... (e.g., mpv-dev-x86_64-20260226-git-d54bad5.7z).mpv-dev-x86_64-v3-...) may offer better performance but require a newer CPU with AVX2 support. Use the standard builds (without "v3") for broader compatibility.[Data Folder] (The root data folder)libmpv-2.dll to the root of the Data Folder.Used to enable mpv to stream online videos (e.g., YouTube, Vimeo, and many other sites) via Video > Open from URL.
[Data Folder] (The root data folder)yt-dlp.exe and place it directly in the root of the Data Folder.Tip: Subtitle Edit can download yt-dlp automatically. When you use Video > Open from URL for the first time, you will be prompted to download it.
Used for converting image-based subtitles (Sup/VobSub) to text.
[Data Folder]/Tesseract550tesseract.exe and tessdata folder) should be placed here.Used for AI-based speech recognition.
[Data Folder]/SpeechToText/Cppwhisper-cli.exe..bin files) go into a Models subfolder: [Data Folder]/SpeechToText/Cpp/Models.Note: It is generally recommended to use the internal downloader for Whisper due to the complexity of model and library dependencies.
Used for GPU-accelerated AI-based speech recognition.
[Data Folder]/SpeechToText/Purfview-Faster-Whisper-XXLfaster-whisper-xxl.exe is in the folder root.faster-whisper-medium) inside the _models folder.Subtitle Edit 5 can download additional ASR engines directly from the Speech to text window.
[Data Folder]/CrispASR. Models go into its models folder. Crisp ASR backends include Parakeet, Canary, Cohere, Fire Red, GLM, Granite, Qwen3, Mega, Omni, and Kyutai.
en, de, fr, es, it, ja, zh, nl, pt, ar, uk, cs. The default is the built-in aligner when the backend supports it, otherwise Qwen3 or Canary CTC depending on the backend; pick a wav2vec2 entry manually to use one of those.[Data Folder]/Qwen3ASR. Models go into [Data Folder]/Qwen3ASR/models.[Data Folder]/parakeet.cpp. Each model has its own folder because the model weights and vocab.txt must stay together.Use Speech to Text for the current engine list and workflow.
Used for OCR of image-based subtitles.
[Data Folder]/OCR/PaddleOCR3-1[Data Folder]/OCR/PaddleOCR3-1/modelsSubtitle Edit 5 can download local TTS servers and models from the Text to speech window.
[Data Folder]/TextToSpeech/Qwen3TtsCrispAsr/voices. The talker GGUFs (VoiceDesign 1.7B or CustomVoice 1.7B) and the 12 Hz codec are downloaded into the shared [Data Folder]/CrispASR/models cache alongside the Crisp ASR speech-to-text models, not under TextToSpeech/Qwen3TtsCrispAsr/models — installing Crisp ASR first is therefore recommended. Older installs that still have model files under the legacy TextToSpeech/Qwen3TtsCrispAsr/models folder are migrated automatically the first time the engine is used.[Data Folder]/TextToSpeech/Chatterbox/voices. The Base / Turbo model GGUFs (T3 + S3Gen) are downloaded into the shared [Data Folder]/CrispASR/models cache alongside the Crisp ASR speech-to-text models, not under TextToSpeech/Chatterbox/models — installing Crisp ASR first is therefore recommended. Older installs that still have model files under the legacy TextToSpeech/Chatterbox/models folder are migrated automatically the first time the engine is used.[Data Folder]/TextToSpeech/OmniVoice. Brings its own omnivoice-tts and omnivoice-codec binaries. Supports 646 languages and voice cloning on CPU. models/ and voices/ subfolders.[Data Folder]/TextToSpeech/KokoroTtsCpp. Models go into the models folder.Use Text to Speech for engine-specific options.
Used for reading media info, extracting audio, and generating waveforms.
sudo apt install ffmpeg) or download static builds from ffmpeg.org.ffmpeg in system paths (e.g., /usr/bin/ffmpeg).ffmpeg binary in [Data Folder]/ffmpeg.Settings.json (located in the Data Folder):
"FfmpegPath": "/path/to/your/ffmpeg"
Used as a video player engine.
libmpv (e.g., sudo apt install libmpv2 or libmpv-dev).libmpv.so.2 or libmpv.so in standard library paths (/usr/lib, /usr/local/lib, etc.).Used to enable mpv to stream online videos via Video > Open from URL.
[Data Folder] (The root data folder)yt-dlp_linux and place it directly in the root of the Data Folder.chmod +x yt-dlp_linuxTip: Subtitle Edit can download yt-dlp automatically when you use Video > Open from URL for the first time.
Used for converting image-based subtitles (Sup/VobSub) to text.
sudo apt install tesseract-ocr).tessdata) is also installed (often separate packages).Used for AI-based speech recognition.
[Data Folder]/SpeechToText/Cppwhisper-cli..bin files) go into a Models subfolder: [Data Folder]/SpeechToText/Cpp/Models.Used for GPU-accelerated AI-based speech recognition.
[Data Folder]/SpeechToText/Purfview-Faster-Whisper-XXLfaster-whisper-xxl binary is present.faster-whisper-medium) inside the _models folder.The same data-folder layout is used on Linux. Prefer the in-app downloaders for Crisp ASR, Qwen3 ASR, Parakeet.cpp, PaddleOCR, Qwen3 TTS (CrispASR), Chatterbox TTS (CrispASR), OmniVoice TTS, and Kokoro TTS because the required files differ by build and model.
Used for reading media info, extracting audio, and generating waveforms.
brew install ffmpeg) or download static builds from ffmpeg.org.ffmpeg in system paths (e.g., /opt/homebrew/bin/ffmpeg).ffmpeg binary in [Data Folder]/ffmpeg.Settings.json (located in the Data Folder):
"FfmpegPath": "/path/to/your/ffmpeg"
Used as a video player engine.
brew install mpv).libmpv.dylib or libmpv.2.dylib in standard library paths (/opt/homebrew/lib, /usr/local/lib, etc.).Used to enable mpv to stream online videos via Video > Open from URL.
[Data Folder] (The root data folder)yt-dlp_macos and place it directly in the root of the Data Folder.chmod +x yt-dlp_macosTip: Subtitle Edit can download yt-dlp automatically when you use Video > Open from URL for the first time.
Used for converting image-based subtitles (Sup/VobSub) to text.
brew install tesseract).tessdata) is also installed.Used for AI-based speech recognition.
[Data Folder]/SpeechToText/Cppwhisper-cli..bin files) go into a Models subfolder: [Data Folder]/SpeechToText/Cpp/Models.Some newer local engines are platform-specific or model-specific. Use the in-app downloaders where available, and check Speech to Text, Text to Speech, and OCR for current engine notes.