InsideVoice Privacy Policy
Effective Date: 2026-07-05 Last Updated: 2026-07-05
InsideVoice ("the App", "we", "us", or "our") is a personal English learning application developed and operated by Yang Jiandong as an individual developer. We take your privacy seriously. This Privacy Policy explains what information we collect, how we use it, and your choices regarding that information.
In one sentence: We collect as little as possible, store almost everything locally on your device, require no account, and never sell your data.
1. Information We Collect
1.1 Locally Stored on Your Device (Never Sent to Us)
The following data is stored only on your iPhone using Apple's standard local storage (AsyncStorage). We do not operate a server that holds it, and we cannot access it:
- Your learning progress (study days, streak count, level)
- Vocabulary cards and spaced-repetition schedule
- Sentences you write for practice
- Achievement unlocks and diamond balance
- App preferences (selected English level, theme, etc.)
Audio you record during practice is created on your device and used only to produce an immediate result. Depending on the feature, it is either processed entirely on your device or sent transiently to a speech provider for scoring (see 1.2 and 1.4). We do not store your audio after grading.
1.2 Information Sent to Third Parties (Only When You Use Certain Features)
When you use AI-powered features, the relevant input is sent to third-party services for processing. By default these requests pass through our own secure proxy (hosted on Cloudflare), which attaches the necessary credentials and forwards them to the providers below; the proxy does not store your content.
| Feature | Data Sent | Third-Party Service |
|---|---|---|
| Sentence grading (造句评分) | The English sentence you wrote + the target word | OpenRouter (LLM) |
| Free conversation (自由对话) | Your typed or transcribed messages | OpenRouter (LLM) |
| Word lookup (词汇查询) | The English word you tap | OpenRouter (LLM) |
| Pronunciation scoring (发音评分) | Your recorded audio clip | Microsoft Azure (Speech) |
| Real-voice speech (真人语音朗读) | The text to be read aloud (e.g. AI replies, example sentences) | Inworld (text-to-speech) |
| Conversation speech-to-text | Audio is transcribed on your device by Apple; only the resulting transcript is sent onward | Apple (on-device) + OpenRouter |
| Audio playback (词音 / 故事) | Your IP address (standard for content delivery) | Cloudflare (R2 CDN) |
We do not send any personally identifying information with these requests — no name, no email, no device ID, no account. Each request is anonymous.
1.3 Information NOT Collected
We do not collect, request, or have access to:
- Your name, email, phone number, or any account information
- Your contacts, photos, calendar, or location
- Your browsing history outside the App
- Crash reports, analytics, or usage tracking
- Advertising identifiers
- Any data that would identify you as an individual
The App does not require an account. There is no sign-up, no login.
1.4 Microphone and Audio
The App requests microphone access to record your spoken English. There are two separate paths, and which one runs depends on the feature you use:
- Conversation and sentence practice: your speech is transcribed into text by Apple's on-device speech recognition. The raw audio stays on your device; only the transcript may be sent onward to an AI service (see 1.2).
- Pronunciation scoring (发音评分): to score your pronunciation at the phoneme level, the recorded audio clip itself is sent, over an encrypted connection, to Microsoft Azure's speech service (via our Cloudflare proxy). This is required because detailed pronunciation assessment cannot be performed on-device.
In both cases, your audio is used only to produce the immediate result and is not retained by us afterward. If you never use pronunciation scoring, no audio ever leaves your device. You can revoke microphone access at any time in iOS Settings › Privacy & Security › Microphone › InsideVoice.
2. Third-Party Services
Requests to the services below are, by default, routed through our own proxy hosted on Cloudflare, which attaches the necessary API credentials and forwards the request. This keeps provider keys off your device; the proxy is a pass-through and does not store your content.
2.1 AI Language Model Services (text)
For AI grading and conversation, your text input is processed by a large language model:
- By default, requests go through our proxy to OpenRouter (Privacy Policy), which routes them to a suitable model.
- If you configure your own API key in Settings, requests go directly to the provider you choose, which may be OpenRouter, DeepSeek (Privacy), Anthropic Claude (Privacy), or Alibaba Qwen (Privacy).
These services temporarily process your text to generate a response and may retain logs for safety and abuse-prevention purposes per their own policies. We do not control or have access to those logs.
2.2 Speech Assessment (audio)
Pronunciation scoring sends your recorded audio clip to Microsoft Azure's Cognitive Speech service (Microsoft Privacy Statement) for phoneme-level assessment. Azure processes the audio to return a score. We do not store the audio, and no name, email, or account is attached to the request.
2.3 Real-Voice Speech Synthesis (text)
The "real human voice" feature sends the text to be spoken aloud (such as an AI conversation reply or an example sentence) to Inworld (Privacy) to generate natural-sounding audio. Only text is sent — no microphone audio, and no identifying information.
2.4 Content Delivery and Proxy (Cloudflare)
Pre-recorded audio (word pronunciations, story narrations) is delivered via Cloudflare R2, and the AI requests above are routed through our Cloudflare proxy (Cloudflare Privacy). Your device's IP address is visible to Cloudflare for the duration of each request, as is standard for any internet service.
2.5 Apple Services
The App is distributed through the Apple App Store and TestFlight. Apple may collect anonymous usage data per its own privacy policy (Apple Privacy). In-app subscriptions are processed entirely by Apple; we never see or receive your payment details.
2.6 Optional: Obsidian Integration
If you choose to enable Obsidian sync, the App writes selected vocabulary and conversation data to a local Obsidian vault folder you specify. This is a file-system write only — no data is sent to Obsidian Inc. or any server. You can disable this feature anytime in Settings.
3. Children's Privacy
The App is rated 4+ and is suitable for general audiences. We do not knowingly collect any personal information from users of any age, including children under 13. If you are a parent or guardian and believe your child has provided information to the App, please contact us — although, as stated above, no such information should exist in our possession.
4. Your Rights and Choices
Because we do not run a server that stores your data, you have complete control:
- View your data: All your data lives on your device. You can see it within the App.
- Delete your data: Uninstall the App. All locally stored data is permanently deleted along with the App.
- Disable AI features: Use Settings to enable/disable specific AI-powered features. You can also opt to provide your own API key, removing reliance on our proxy.
- Keep audio on-device: If you do not use pronunciation scoring, your recorded audio never leaves your device.
- Disable microphone: Revoke microphone permission in iOS Settings › Privacy & Security › Microphone › InsideVoice.
We do not have an "account deletion" option because there is no account to delete.
5. Data Security
Because virtually all your data stays on your device, traditional server-side breach risks do not apply to the App. We rely on Apple's standard iOS data protection (which encrypts data at rest when your device is locked with a passcode).
For data sent to third-party services (see Section 2), security depends on those providers' practices. All requests use HTTPS encryption in transit.
6. International Users
The App is available globally through the Apple App Store. The third-party services we use by default (OpenRouter, Microsoft Azure, Inworld, Cloudflare) primarily process data on servers in the United States. If you configure your own AI provider key, your data may be processed wherever that provider operates, which may include other countries. By using these features you understand that your input may be processed outside your country of residence.
7. Changes to This Policy
We may update this Privacy Policy from time to time. The "Last Updated" date at the top reflects the most recent revision. Material changes will be announced through the App's update notes. Continued use of the App after changes means you accept the revised policy.
8. Contact
If you have any questions about this Privacy Policy or how the App handles your data:
Developer: Yang Jiandong Email: 138011988yjd@gmail.com
InsideVoice 隐私政策（中文版）
生效日期： 2026-07-05 最后更新： 2026-07-05
InsideVoice（以下简称"本应用"）由个人开发者 Yang Jiandong 独立开发与运营。我们高度重视您的隐私。本政策说明我们收集什么信息、如何使用、以及您对此的选择权。
一句话：我们收集尽可能少的信息，几乎所有数据只存在您的设备本地，不需要账号，永远不会出售您的数据。
1. 我们收集的信息
1.1 仅存储在您设备上的数据（不发送给我们）
以下数据使用 Apple 标准本地存储（AsyncStorage）保存在您的 iPhone 上。我们没有存储它的服务器，也无法访问：
- 学习进度（总天数、连续天数、等级）
- 单词卡 + 记忆曲线安排
- 您输入的造句练习内容
- 成就解锁记录、钻石余额
- 应用偏好（选定的英文等级、主题等）
您在练习中录制的音频在设备上生成，仅用于产出即时结果。根据所用功能，它要么完全在设备本地处理，要么被短暂发送给语音服务用于评分（见 1.2 与 1.4）。评分完成后，我们不保存您的音频。
1.2 在使用特定功能时发送给第三方的数据
使用 AI 驱动功能时，相关输入会被发送到第三方服务进行处理。默认情况下，这些请求经由我们自己的安全代理（部署在 Cloudflare）转发给下列服务商；代理不存储您的内容。
| 功能 | 发送的数据 | 第三方服务 |
|---|---|---|
| 造句评分 | 您写的英文句子 + 目标词 | OpenRouter（大语言模型） |
| 自由对话 | 您输入或语音转写的消息 | OpenRouter（大语言模型） |
| 词汇查询 | 您点击的英文单词 | OpenRouter（大语言模型） |
| 发音评分 | 您录制的音频片段 | 微软 Azure（语音服务） |
| 真人语音朗读 | 需要朗读的文本（如 AI 回复、例句） | Inworld（文本转语音） |
| 对话语音转文字 | 音频先在设备本地由 Apple 转写，仅转写文本发送给后续服务 | Apple（设备端）+ OpenRouter |
| 音频播放（词音 / 故事） | 您的 IP 地址（内容分发的标准行为） | Cloudflare（R2 CDN） |
所有请求都不包含任何身份识别信息——没有姓名、邮箱、设备 ID、账号。每个请求都是匿名的。
1.3 我们不收集的信息
我们不收集、请求或访问：
- 您的姓名、邮箱、电话或任何账号信息
- 您的通讯录、照片、日历、位置
- 应用外的浏览历史
- 崩溃日志、分析数据、使用追踪
- 广告标识符
- 任何能识别您身份的数据
应用不需要账号。无注册、无登录。
1.4 麦克风与音频
应用申请麦克风权限用于录制您的英语跟读。这里有两条相互独立的路径，具体走哪条取决于您使用的功能：
- 对话与造句练习： 您的语音由 Apple 的设备端语音识别转写为文字。原始音频保留在您的设备上，只有转写文本可能被发送给 AI 服务（见 1.2）。
- 发音评分： 为了在音素级别给发音打分，录制的音频片段本身会通过加密连接发送给微软 Azure 的语音服务（经我们的 Cloudflare 代理）。这是必要的，因为详细的发音评估无法在设备本地完成。
无论哪条路径，您的音频都只用于产出即时结果，之后不被我们保留。如果您从不使用发音评分，就没有任何音频离开您的设备。您可以随时在 iOS 设置 › 隐私与安全性 › 麦克风 › InsideVoice 中撤销麦克风权限。
2. 第三方服务
下列服务的请求，默认经由我们部署在 Cloudflare 上的代理转发——由代理附加所需的 API 凭证再转发出去。这样可以让服务商密钥不落到您的设备上；代理只做透传，不存储您的内容。
2.1 AI 大语言模型服务（文本）
AI 评分和对话功能会用大语言模型处理您的文本输入：
- 默认情况下，请求经我们的代理发送给 OpenRouter（隐私政策），由其路由到合适的模型。
- 如果您在"设置"里配置了自己的 API key，请求会直接发送给您选择的服务商，可能是 OpenRouter、DeepSeek（隐私）、Anthropic Claude（隐私）或阿里 Qwen（隐私）。
这些服务会临时处理您的文本以生成响应，并可能按其各自政策为安全与防滥用目的保留日志。我们不控制也无法访问这些日志。
2.2 语音评估（音频）
发音评分会把您录制的音频片段发送给微软 Azure 的认知语音服务（微软隐私声明）做音素级评估。Azure 处理音频以返回分数。我们不存储该音频，请求也不附带任何姓名、邮箱或账号。
2.3 真人语音合成（文本）
"真人音色"功能会把需要朗读的文本（如 AI 对话回复或例句）发送给 Inworld（隐私）以生成接近真人的音频。只发送文本——不发送麦克风音频，也不含任何身份信息。
2.4 内容分发与代理（Cloudflare）
预录音频（单词发音、故事朗读）通过 Cloudflare R2 分发，上述 AI 请求也经由我们的 Cloudflare 代理转发（Cloudflare 隐私）。每次请求期间，您的 IP 地址对 Cloudflare 可见，这是任何互联网服务的标准行为。
2.5 Apple 服务
本应用通过 Apple App Store 和 TestFlight 分发。Apple 可能按其隐私政策收集匿名使用数据（Apple 隐私）。应用内订阅完全由 Apple 处理，我们绝不会看到或收到您的支付信息。
2.6 可选：Obsidian 集成
如果您启用 Obsidian 同步，应用会将选定的词汇和对话数据写入您指定的本地 Obsidian 仓库文件夹。这只是文件系统写入操作——不向 Obsidian Inc. 或任何服务器发送任何数据。 您可以随时在设置里关闭。
3. 儿童隐私
应用分级为 4+，适合一般受众。我们不会有意收集任何年龄段用户的个人信息，包括 13 岁以下儿童。如果您是父母或监护人，认为孩子向应用提供了信息，请联系我们——但如上所述，我们手中不应存在任何此类信息。
4. 您的权利与选择
由于我们没有存储您数据的服务器，您拥有完全控制权：
- 查看数据：所有数据都在您的设备上，可在应用内查看。
- 删除数据：卸载应用。所有本地存储数据会随应用一起被永久删除。
- 关闭 AI 功能：在设置里启用/关闭特定 AI 功能。也可以提供自己的 API key，不依赖我们的代理。
- 让音频不离开设备：如果您不使用发音评分，录制的音频永不离开您的设备。
- 关闭麦克风：在 iOS 设置 › 隐私与安全性 › 麦克风 › InsideVoice 撤销权限。
我们没有"删除账号"选项，因为没有账号可删。
5. 数据安全
由于几乎所有数据都在您的设备上，传统的服务器端泄漏风险对本应用不适用。我们依赖 Apple 标准的 iOS 数据保护（在设备用密码锁定时对静态数据加密）。
发送到第三方服务的数据（见第 2 节），其安全性依赖于这些服务商的实践。所有请求在传输中使用 HTTPS 加密。
6. 国际用户
应用通过 Apple App Store 全球分发。我们默认使用的第三方服务（OpenRouter、微软 Azure、Inworld、Cloudflare）主要在美国的服务器上处理数据。如果您配置了自己的 AI 服务商 key，您的数据可能在该服务商运营的任何地区被处理，可能包括其他国家。使用这些功能即表示您理解：您的输入可能在您居住国以外被处理。
7. 政策变更
我们可能不时更新本隐私政策。顶部"最后更新"日期反映最近一次修订。重大变更将通过应用更新说明公告。在变更后继续使用应用，即表示您接受修订后的政策。
8. 联系方式
如有任何关于本隐私政策或应用数据处理方式的问题：
开发者： Yang Jiandong 邮箱： 138011988yjd@gmail.com