Subtitle Companion / 字幕伴侣 — Privacy Policy / 隐私政策
Effective date / 生效日期: 2026-10-03
Developer / 开发者: feixuwu
Contact / 联系邮箱: feixuwu@gmail.com
English
Subtitle Companion creates subtitles from playback audio that other Android apps permit to be captured. No app account is required.
Audio and translation
In offline mode, playback audio is recognized on the device by SenseVoice for the supported languages. Text is translated on the device by Google ML Kit. The app does not open the microphone or save audio files. Only audio shared under Android playback-capture rules is available; protected content may be unavailable.
Azure mode is optional. When you select it and provide your own Azure Speech Key and Region, playback audio is transmitted to your configured Microsoft Azure Speech service for recognition and translation. Microsoft processes this data under its applicable service terms and privacy policy. Charges, retention and service-region choices depend on your Azure account and configuration. Offline mode does not send playback audio or subtitle text to the developer.
Stored information and deletion
Subtitle history is enabled by default and stores recognized text, translations and session times only in the app’s private local database. You can disable future history, delete individual sessions or clear all history in the app. Disabling history does not delete existing sessions. No audio recordings are stored. Azure Key and Region are encrypted with Android Keystore before local storage. Language, download and caption appearance preferences are stored locally. App data is excluded from Android cloud backup and device transfer. Clearing app storage or uninstalling deletes the app’s local data; Azure server-side processing remains subject to Microsoft’s policies.
Model downloads and external services
Offline processing requires speech and translation models. Speech models may be bundled in the installation or downloaded from Hugging Face and GitHub; translation models are downloaded through Google ML Kit. These providers receive network requests, which can include an IP address and routine device or service information, and may keep logs under their own policies. Model requests are distinct from sending playback audio for recognition. The app contains no developer-operated analytics or advertising integration. Opening this public policy page uses GitHub Pages and is subject to GitHub’s website privacy practices. Sending support email shares the information you choose to include with the developer and email providers; avoid including audio, private subtitles or credentials unless needed.
Third-party SDK diagnostics
Google ML Kit also uses SDK diagnostics and usage analytics. Google may receive device and app information, installation identifiers, performance and error metrics, feature events, and configured source and target languages. Translation uses Firebase installations and remote configuration services for diagnostics. This is separate from local subtitle processing and from model-file download requests. The developer does not operate an analytics backend or advertising service. Google handles this SDK data under its own privacy policies; do not interpret offline translation as an absence of all SDK network activity.
Local history remains on the device until you delete it, clear app storage or uninstall. Support emails are kept only as needed to handle your request. Third-party service data is retained under the relevant provider policies and your Azure account configuration; deleting local app data does not guarantee deletion of data already processed by a provider. Contact the support email for requests about information you sent directly to the developer. The app has no user account, and local history can be deleted without contacting support.
Android permissions
RECORD_AUDIO is required by Android for playback capture; this app’s capture configuration does not use microphone input. MediaProjection requires Android’s confirmation when starting a session. The foreground service and notifications show the running session and provide pause/stop actions. Picture-in-picture places subtitles over another app using Android’s PiP controls; the app does not request the SYSTEM_ALERT_WINDOW overlay permission. Internet and network-state access support model downloads and optional Azure processing.
Your choices and contact
You can stop capture at any time, use offline mode, disable optional history, remove credentials by clearing app storage, and delete local history. For privacy questions or requests concerning information you send directly to the developer, contact feixuwu@gmail.com. The app has no registration flow and does not ask for a birth date; the developer does not intentionally solicit children’s personal information. If a parent or guardian believes such information was sent to the developer, please contact this address. Applicable rights depend on your location and the relevant service provider.
Changes and third-party policies
Changes to data processing will be reflected in this policy’s effective date and, where appropriate, in the app’s explanation or consent flow.
Microsoft: https://privacy.microsoft.com/privacystatement
Google: https://policies.google.com/privacy
GitHub: https://docs.github.com/en/site-policy/privacy-policies/github-general-privacy-statement
Hugging Face: https://huggingface.co/privacy
中文
字幕伴侣根据其他 Android 应用允许共享的播放音频生成识别和翻译字幕,无需注册应用账户。
音频与翻译
离线模式在设备上使用 SenseVoice 识别支持的语言;文本在设备上使用 Google ML Kit 翻译。应用不打开麦克风,也不保存音频文件。只能捕获符合 Android 播放音频共享规则的内容,受保护内容可能无法捕获。
Azure 模式为可选功能。用户选择该模式并填写自己的 Azure Speech Key 和 Region 后,播放音频会发送到用户配置的 Microsoft Azure Speech 服务进行识别和翻译。Microsoft 按适用的服务条款和隐私政策处理这些数据;费用、保留期限和服务地区取决于用户的 Azure 账户及配置。离线模式不会向开发者发送播放音频或字幕文本。
本地保存与删除
字幕历史默认启用,只在应用私有的本地数据库保存识别原文、译文和会话时间。用户可以关闭后续保存、删除单个会话或清除全部历史;关闭保存不会删除已保存的历史。应用不保存音频录音。Azure Key 和 Region 使用 Android Keystore 加密后在本地保存,语言、下载及字幕外观偏好也在本地保存。应用数据不参与 Android 云备份和设备迁移。清除应用数据或卸载会删除应用本地数据;Azure 服务端的数据处理仍受 Microsoft 的政策约束。
模型下载与外部服务
离线处理需要语音和翻译模型。语音模型可能随安装包提供,或从 Hugging Face、GitHub 下载;翻译模型通过 Google ML Kit 下载。这些提供方会收到网络请求,可能包含 IP 地址及常规设备或服务信息,并可能按各自政策保留日志。下载模型与上传播放音频进行识别是不同的数据处理。应用没有集成开发者运营的统计分析或广告服务。访问此公开政策页面会使用 GitHub Pages,并适用 GitHub 的网站隐私规则。发送支持邮件会向开发者和邮件服务提供方提供用户主动填写的信息;除非必要,请不要发送音频、私密字幕或凭据。
第三方 SDK 诊断
Google ML Kit 还使用 SDK 诊断及使用情况分析。Google 可能收到设备和应用信息、安装标识、性能和错误指标、功能事件,以及配置的原文和目标语言。翻译 SDK 使用 Firebase 安装及远程配置服务进行诊断。这与设备上的字幕处理、模型文件下载请求是不同的数据处理。开发者不运营分析后台或广告服务。Google 根据其自身隐私政策处理这些 SDK 数据;离线翻译不代表 SDK 完全没有网络活动。
本地历史保留在设备上,直到用户删除、清除应用数据或卸载。支持邮件仅在处理请求所需的时间内保留。第三方服务数据按服务商政策及用户的 Azure 账户配置保留;删除本地应用数据不保证删除服务商已经处理的数据。有关用户直接向开发者发送的信息,可通过支持邮箱提出请求。应用不创建用户账户,删除本地历史不需要联系开发者。
Android 权限
Android 要求录音权限(RECORD_AUDIO)以控制播放音频捕获;本应用的捕获配置不使用麦克风输入。每次启动会话时,媒体投射(MediaProjection)由 Android 请求用户确认。前台服务和通知展示运行状态,并提供暂停、结束操作。画中画通过 Android PiP 控件在其他应用上方显示字幕,应用不请求 SYSTEM_ALERT_WINDOW 悬浮窗权限。网络和网络状态权限用于模型下载及可选的 Azure 处理。
用户选择与联系
用户可以随时停止捕获、使用离线模式、关闭历史保存、清除应用数据以移除凭据,以及删除本地历史。有关隐私或用户直接向开发者提供的信息,请联系 feixuwu@gmail.com。应用没有注册流程,也不要求出生日期;开发者不会主动索取儿童个人信息。如家长认为儿童向开发者发送了个人信息,请通过此邮箱联系。具体权利取决于用户所在地及相关服务提供方。
变更与第三方政策
如数据处理发生变化,将更新本政策的生效日期,并在适用时更新应用说明或同意流程。第三方政策链接见上方英文部分。