Privacy Policy
Last updated: July 21, 2026
Core Promise: AI Offline Translator is a fully offline application. We do not collect, store, or upload any of your personal data. All translation processing happens on your device.
1. Information Collection
We do not collect any information. Specifically:
- Translation Content: The text you enter and the translation results are stored only on your device and are never sent to any server.
- Personal Information: We do not require account registration, and do not collect names, email addresses, phone numbers, or any other personal information.
- Usage Data: We do not use any analytics tools, crash reporting SDKs, or ad tracking SDKs.
- Device Permissions:
- Camera & Photo Library: Only requested when you actively use the photo translation feature. All images are processed locally and are never uploaded.
- Microphone: Only requested when you actively use the voice translation feature. All audio is processed locally and is never uploaded or stored.
- Speech Recognition: When you use voice translation for the first time, Apple's on-device speech recognition may download a language model directly from Apple's servers. This download contains no user data and is a one-time system-level process managed entirely by iOS. All subsequent speech recognition runs locally on your device.
2. Data Storage
All data (translation history, user preferences) is stored locally on your device using Apple's SwiftData framework. You can delete this data at any time within the app. Uninstalling the app will also delete all local data.
3. Model Download
The app offers two translation models from Alibaba's Qwen2.5 series:
- Qwen2.5 1.5B (Recommended) — ~1.1 GB, best translation quality
- Qwen2.5 0.5B (Lightweight) — ~400 MB, ideal for devices with limited storage
Models are downloaded from Hugging Face, with automatic fallback to ModelScope (a China-accessible mirror by Alibaba Cloud). The download process contains no user identifiers or tracking parameters and is triggered by the user. After download, all translation inference runs entirely on-device with no internet connection required. You can store multiple models and switch between them at any time.
4. Third-Party Services
This app contains none of the following third-party services:
- Advertising networks
- User behavior analytics
- Crash reporting services
- Social media SDKs
- Any form of data tracking
5. Children's Privacy
This app does not collect any data from anyone, including children under 13.
6. Policy Changes
If this privacy policy changes, we will notify users in the app update notes. Significant changes will be communicated via in-app notification.
7. Contact Us
If you have any questions about this privacy policy, please contact:
📧 birdflyingtodie@gmail.com
— 中文版 / Chinese —
隐私政策
最后更新：2026 年 7 月 21 日
核心承诺：AI离线翻译是一款完全离线的应用。我们不收集、不存储、不上传您的任何个人数据。所有翻译处理均在您的设备本地完成。
1. 信息收集
我们不收集任何信息。具体而言：
- 翻译内容：您输入的文字和翻译结果仅存储在您的设备本地，不会发送到任何服务器。
- 个人身份信息：我们不要求注册账号，不收集姓名、邮箱、电话号码等任何个人信息。
- 使用数据：我们不使用任何分析工具（Analytics）、崩溃报告 SDK、或广告追踪 SDK。
- 设备权限：
- 相机与相册：仅在您主动使用拍照翻译功能时由系统请求，照片全部在本地处理，不会上传。
- 麦克风：仅在您主动使用语音翻译功能时由系统请求，音频全部在本地处理，不会上传或存储。
- 语音识别：首次使用语音翻译时，苹果设备端语音识别可能从苹果服务器下载对应语言模型。此为一次性系统级操作，由 iOS 自动管理，不含任何用户数据。下载完成后所有语音识别均在设备本地执行。
2. 数据存储
所有数据（翻译历史、用户设置）均通过 Apple SwiftData 框架存储在您的设备本地。您可以随时在应用内删除这些数据。卸载应用将同时删除所有本地数据。
3. 模型下载
应用提供两种来自阿里千问（Qwen2.5）系列的翻译模型供您选择：
- Qwen2.5 1.5B（推荐）— 约 1.1GB，翻译质量最高
- Qwen2.5 0.5B（轻量）— 约 400MB，适合存储空间有限的设备
模型从 Hugging Face 下载，国内用户自动切换至阿里云 ModelScope（魔搭社区）镜像。下载过程不附加任何用户标识或追踪参数，且为用户主动触发。下载完成后，翻译推理完全在设备本地进行，无需网络连接。您可以同时存储多个模型并随时切换。
4. 第三方服务
本应用不含以下任何第三方服务：
- 广告网络
- 用户行为分析工具
- 崩溃报告服务
- 社交媒体 SDK
- 任何形式的数据追踪
5. 儿童隐私
本应用不收集任何人的任何数据，包括 13 岁以下儿童。
6. 政策变更
如本隐私政策发生变更，我们将在应用更新说明中告知。重大变更将通过应用内通知提醒。
7. 联系我们
如对本隐私政策有任何疑问，请通过以下方式联系：
📧 birdflyingtodie@gmail.com
AI Translator / AI离线翻译 · Your Privacy, Our Bottom Line / 您的隐私，我们的底线