Skip to content

Privacy Policy & Disclaimer

Established: October 5, 2026
Last updated: October 28, 2026 (added looking up title information on TMDb)

1. Data the app handles

To make subtitles for videos, KotobaSync (the “App”) handles the following information on your Mac.

  • Video, audio and subtitle files: The App reads and writes only the files you choose in Open and Save dialogs, or files you drop onto it. Transcription is done on your Mac with the Whisper engine (whisper.cpp) and Silero VAD bundled with the App, or, if you choose it, with macOS speech recognition (the Speech framework). To read subtitles burned into the picture or image subtitles, the App uses macOS text recognition (the Vision framework) on your Mac. The original video file is not changed.
  • Subtitle text: The App handles the transcribed source text, the Japanese translation, suggestions, review marks and so on. SRT files you save and their companion files (〜.translation-review.json, 〜.speaker-review.json) are saved where you choose.
  • Autosave: Subtitles not yet saved as SRT are saved for recovery in ~/Library/Application Support/SubtitleStudio/Autosave (up to the 20 most recent). When you save an SRT, the autosave for that work is removed.
  • Saved translation progress, glossaries and work notes: These are saved in ~/Library/Application Support/SubtitleStudio/TranslationCheckpoints, so that translation and review can be continued and the spelling of names and other terms kept consistent.
  • Converted and exported videos: MP4 files converted from MKV and similar files, and exported M4V files, are saved by default in the same folder as the original video (this can be changed in Settings). Only if the destination cannot be written to are they saved in ~/Library/Caches/SubtitleStudio/ConvertedVideos.
  • Transcription and speaker analysis models: Downloaded Whisper models are saved in ~/Library/Application Support/SubtitleStudio/Models, and speaker analysis models you prepare in ~/Library/Application Support/SubtitleStudio/SpeakerModels.
  • Results of the extra analyses: If you turn on the extra analyses, anonymous speaker segments and reference information about scenes obtained from images are saved in the same place as the saved translation progress. No database storing voice characteristics themselves is created.
  • Settings: The translation engine and model, connection destinations, style, save locations, subtitle appearance, update check settings and so on are saved in ~/Library/Preferences/com.subtitle.studio.plist.
  • LM Studio token: If you use authentication with LM Studio, the API token you enter is saved in the macOS keychain.
  • TMDb API key: The API key you enter to search for titles on TMDb is saved in the macOS keychain. You can delete it at any time under “Title information from the web” (Web の作品情報) in Settings.

The name “SubtitleStudio” in the storage locations is kept for compatibility with earlier versions. The App has no account feature, no advertising and no function for collecting usage statistics, crash reports or the like. None of this data is sent to the operator (K386).

2. macOS permissions

In normal use, the App does not ask for any special macOS permissions. It reads and writes only the video and subtitle files you choose.

Only if you choose Apple Speech (macOS speech recognition) as the transcription method does the App use the macOS “Speech Recognition” permission. The App uses Apple Speech only when recognition can be done entirely on your Mac for the chosen language, and does not switch to methods that send audio to Apple’s servers. You can revoke the permission at any time in “Privacy & Security” in System Settings.

3. Network connections

The App connects to the network for the following four purposes only.

  1. Model downloads: When you press “Download” (取得する) for a transcription model in Settings or on the welcome screen, or press “Prepare model…” (モデルを準備…) for speaker analysis under “Characters” (人物設定) in the work notes, the App downloads the model files from Hugging Face (huggingface.co). Only the request for the files is sent; no video, audio or subtitles are sent. After the download, transcription and speaker analysis do not use an internet connection. Access to Hugging Face is subject to that company’s privacy policy.
  2. Communication with the translation engine:
    • Ollama and LM Studio: The App connects only to a translation engine running on the same Mac (127.0.0.1) and sends nothing outside. Connections through proxies, redirects or to cloud models are refused.
    • Codex: Only if you choose Codex, the App sends the subtitle text (source and translation) and translation instructions (which, when title information is used, include the gender and family relationships of characters whose names appear in the subtitles) to OpenAI for translation, creating glossaries and work notes, and review, using your own ChatGPT / Codex usage allowance. A confirmation is shown when you choose it in Settings and before sending begins. Video and audio files are not sent. Only if you turn on “Codex image analysis of the video” (Codexで映像の画像を補助解析する) under “Characters” (人物設定) in the work notes, the App sends still images taken from the video (up to 12 sections per analysis, 3 frames each, no more than 640 pixels on the long side) together with the subtitle text, to obtain reference information about scenes. A confirmation is shown when you turn it on. The handling of information sent to OpenAI is subject to OpenAI’s terms of use and privacy policy. The operator is not responsible for how OpenAI handles it.
  3. Looking up title information: When you open a video, a confirmation is shown first, and only if you choose “Look it up” (調べる) does the App send the title and year that can be read from the file name (and the season for series) to Wikipedia and Wikidata (Wikimedia Foundation, wikipedia.org and wikidata.org) to obtain information about the title’s characters and cast. Subtitle text, audio, video and file locations are not sent. Nothing is sent for videos with names that have no title or year, such as “IMG_1234”. The information obtained is saved on this Mac and is not fetched again for the same title. You can turn this off at any time under “How translation proceeds” (翻訳の進め方) in Settings and delete the saved information. Access to the Wikimedia Foundation is subject to that foundation’s privacy policy. Only if you enter a TMDb (The Movie Database, api.themoviedb.org) API key, when you choose “TMDb” on the search screen or when a title is not found on Wikipedia, the App sends the search words, the title’s ID number and your API key to TMDb to obtain information about the title’s characters and cast. Subtitle text, audio, video and file locations are not sent. Access to TMDb is subject to TMDb’s terms of use and privacy policy.
  4. Update checks: To check whether there is a new version, the App fetches version information from this site (https://k386.sub.jp/KotobaSync/api/version.json) after launch if 24 hours or more have passed since the last check, and every 24 hours after that. If there is a new version, it downloads the distribution file (DMG) from this site, checks the SHA-256, the Developer ID signature and Apple’s notarization, and then replaces the App when you quit it (or when you choose “Restart and Update Now” (今すぐ再起動して更新) or “Update and Relaunch” (更新して再起動)). Automatic checks stop if you turn off “Check for Updates Automatically” (アップデートを自動で確認) in the app menu; the App then connects only when you choose “Check for Updates…” (アップデートを確認…). Automatic replacement stops if you turn off “Install Updates Automatically” (アップデートを自動でインストール).

The update check sends only the same information as viewing an ordinary web page (IP address, date and time, User-Agent and so on). Video, audio, subtitle text and settings are never sent to the operator or third parties. No ID identifying the user is attached either. These accesses may be recorded in the access logs described in section 4.

Ollama, LM Studio, Codex and ffmpeg are not included in the App; they are software you install and use yourself. Their network connections and handling of information are subject to the terms and policies of their respective providers.

4. This site

This site does not use cookies, advertising or analytics tools. It counts only page views and downloads, and does not store IP addresses (only the IP address ranges of connections blocked to prevent unauthorized access are recorded). For server operation, access logs (IP address, date and time, pages viewed and so on) may be recorded for a certain period.

5. Third-party software

The App includes the following software. The license texts are bundled in the App’s Contents/Resources.

  • whisper.cpp (MIT License, © The ggml authors)
  • Silero VAD (MIT License)

6. Disclaimer

The App is provided free of charge “as is”, without any warranty as to accuracy, completeness, usefulness, fitness for a particular purpose or safety. The results of transcription, text recognition, translation and review, and the display times of subtitles, may contain errors. The quality of translations depends on the translation engine and model you choose. If you use the results for publication, delivery, distribution or the like, please play them back and check the content yourself.

Checking the rights to the videos and subtitles you create, translate or publish subtitles for is your own responsibility. The purpose of the App and what is prohibited (publishing or distributing subtitles or videos made for other people’s copyrighted works, commercial use and so on) are set out in the Terms of Use.

The operator is not liable for any damage arising from the use of, or inability to use, the App or this site, whether direct, indirect, incidental or consequential, including loss or corruption of data, equipment failure, business interruption, loss of profit or any other damage.

The distribution file is signed with a Developer ID and has been checked by Apple’s notarization service, but notarization does not mean that Apple guarantees or endorses the App’s quality or behaviour.

7. Changes

This policy may be changed without notice. The changed content takes effect from the time it is published on this page.

8. Contact

For questions about this policy and the App, please write to info@k386.sub.jp.

This English text is provided for convenience. In case of any discrepancy, the Japanese version prevails.