Privacy Policy

Effective November 6, 2026 · SoundDay for iPhone and Apple Watch
Operator: 招李 (individual) · Contact: 44493547@qq.com

Recording and cloud processing

Recording, Watch-to-iPhone transfer, playback and export can be used without purchasing a subscription. Microphone access is used for a recording you start or voice enrollment you choose. Obtain necessary permission from participants; do not record secretly. Audio, timing, markers and transfer information are stored on your devices. Watch transfer is not itself an upload to our cloud. Export important originals yourself.

When you choose cloud processing, derived audio and timing information are uploaded to our backend. Audio is sent to Alibaba Cloud Model Studio / DashScope for transcription; transcript text and speaker hints are used to generate summaries. Our backend is in mainland China and the configured Alibaba AI services process data in China. Using cloud processing from another country involves a transfer to China. Availability in an App Store region does not change this processing location. HTTPS/WSS protects transport; this is not end-to-end encrypted AI processing. We do not use your recordings to train our own models or publish them. We do not promise zero retention by downstream providers.

Account and sign-in

Email sign-in uses your email to send a code through Tencent enterprise email (smtp.exmail.qq.com). A keyed email hash links your account; it is not anonymous data. Account IDs, verification records and session tokens enable authentication. Sign in with Apple uses an app-specific identifier without requiring you to disclose your name or email. An encrypted Apple authorization refresh token supports revocation on account deletion. Matching email addresses do not automatically merge accounts: use the explicit linking controls to preserve the same SoundDay account.

Google sign-in is currently disabled. Some distributed builds still contain the Google Sign-In SDK; its bundled privacy declarations cover authentication-related information. No Google account is required. SoundDay does not request access to your Gmail, contacts or Drive, and does not upload recordings to Apple or Google as part of sign-in.

Your voice: separate optional controls

Voice features are sensitive biometric information and are optional. A local voice template is distinct from an encrypted account-template backup and from remote voice registration. After separate consent, account-template sync uploads the template to our backend. Remote registration sends your enrollment sample to the configured voice service at video.lmyiot.com to create a profile. When you subsequently choose cloud processing, short speaker samples may be sent there to compare against your own registered profile. The matching request does not enroll other speakers or search other SoundDay users’ profiles. A “possibly me” match is not identity authentication.

Our remote-registration table holds account/service identifiers, model and consent versions, times and status, not the raw enrollment sample or feature vector. This does not imply that the downstream service retains no audio, features, logs or backups. We do not claim a verified zero-retention arrangement or a verified hosting region for that separate service. If those details are necessary for your use, contact us before opting in.

Deleting remote registration immediately disables new matching and queues erasure at the remote service. Network failures can delay confirmation; the App shows pending deletion. Local templates, account-template backups and remote registration have separate deletion controls. Deletion does not rewrite previously generated speaker labels or delete recordings. Contact support if deletion remains pending.

Memories you choose to save

Memories are off by default. After you explicitly enable them, you can review and save selected text from a completed recording. We store your wording, category, active or past status, source title and time, and selected transcript excerpts with their timestamps and speaker labels in your account on our server in mainland China. This feature does not upload additional audio or make an additional AI request, and does not automatically learn habits or build a profile from your recording history.

After saving, the server may compare the wording of up to 200 recent saved memories in your account and show up to three suggestions from other stories, with original excerpts and sources. This is text matching, not AI semantic judgment. Suggestions never link or rewrite memories automatically; skipping changes nothing, and pausing disables new suggestions. You can explicitly connect saved memories about the same matter; these account-scoped connections show source recordings in chronological order and can be removed without deleting the memories. Later conversations do not automatically replace earlier decisions. You can edit, mark as past, or forget a memory. Pausing memories stops new saves and new connections but keeps existing ones available. Saved excerpts are independent copies: deleting the source recording does not delete them. Forgetting a memory does not delete the original recording. Forget all removes saved memories and turns the feature off; account deletion also removes saved memories from the active service. Backups follow the retention rules below.

Apple subscriptions and usage

When subscription purchasing is enabled and you explicitly confirm an offer, Apple processes payment. We do not receive your bank card details. We verify signed transaction identifiers, original transaction identifiers, account token, purchase and expiry times, trial/subscription status and environment. These support cloud allowances, restoration, renewals and refunds. We store processing duration, task identifiers and billing-period usage. Failed accepted tasks and deleted recordings do not automatically refund usage; retrying the same admitted task does not consume it again.

The intended single monthly plan provides an Apple-eligible seven-day free trial with 300 total minutes, then 1,000 minutes per paid Apple billing period, without rollover. The actual available offer and local price are shown by Apple in the App. If the App says subscriptions are being prepared, the existing allowance policy remains in force and the App does not initiate a purchase. Cancel through Apple subscription settings; uninstalling the App or deleting a SoundDay account does not cancel renewal.

Feedback and replies

Feedback contains your selected category, message, optional contact details, account association, receipt ID, timestamps, status and our latest reply. App and OS versions are attached only if you opt in. Recordings, transcripts, voiceprints and diagnostic logs are not automatically attached. Do not include passwords or verification codes. Authorized support personnel use this information to investigate and respond. Refresh My feedback in the App to see replies; we do not promise an email or push notification.

Retention, deletion and security

Device originals remain under your control. Cloud recordings and results remain until you delete the cloud record or your account. Feedback is retained until account deletion; contact us for earlier removal. Deleting an account removes its active cloud content, feedback and usage records and queues remote-voice erasure. Subscription ownership records, including the original account token and transaction fields, are retained with restricted access to prevent reassignment or duplicate grants. This is not anonymous data; this version has no fixed automatic deletion deadline for these ownership records. Contact us with privacy requests.

Operational database backups are access-restricted and configured for a 30-day retention window with daily cleanup; expired copies may wait for the next cleanup. Deleting active data is not immediate erasure of historical backups. Nginx access logs rotate daily for up to 14 rotations and can contain IP addresses, paths, status and timing; signed audio-download paths are excluded. We do not intentionally log authorization tokens, submitted feedback text or recording contents. Provider-side logs and backups are separate; do not assume deletion from our active database deletes every provider copy. Security measures reduce risk but cannot guarantee absolute protection.

Your choices and contact

You can decline cloud processing or voice enrollment, revoke microphone access, export recordings, delete data and request access, correction or erasure at 44493547@qq.com. Rights depend on your location; you may also contact your local privacy regulator. Withdrawal does not undo processing already performed. SoundDay is not directed to children under 13 or for covert surveillance. AI and speaker labels can be wrong: check the original audio. Changes to this policy update the effective date; new purposes require appropriate notice and permission.

Provider information: Alibaba Cloud Model Studio · Tencent · Apple.

隐私政策

生效日期:2026 年 11 月 6 日 · 运营者:招李(个人)
联系邮箱:44493547@qq.com

录音与云端处理

录音、手表向手机同步、回听和导出不要求购买订阅。仅在你主动开始录音或声纹注册时使用麦克风,请告知参与者并取得必要许可,不得秘密监控。设备保存录音、时间、标记和传输信息;手表同步不等于自动上传云端,请自行导出保存重要原音。

你选择云端整理后,派生音频与时间信息上传到后台,音频发送至阿里云百炼 / DashScope 转写,转写文本和说话人提示用于生成总结。后台及当前阿里云 AI 处理位于中国大陆,境外使用涉及向中国传输;App Store 销售地区不改变处理地点。HTTPS/WSS 用于传输保护,不是端到端加密的 AI 处理。我们不使用录音训练自有模型或公开录音,也不承诺下游服务零保留。

账号与登录

邮箱验证码通过腾讯企业邮箱 smtp.exmail.qq.com 发送,邮箱关联采用带密钥哈希,该关联信息不是匿名数据。账号编号、验证码记录和会话令牌用于认证。Apple 登录使用应用专属标识,不要求提供姓名或邮箱;加密保存的 Apple 授权刷新令牌用于注销时撤销授权。不因邮箱相同自动合并账号,请使用账号绑定入口保留同一 SoundDay 账号。

Google 登录当前暂停;部分已分发版本仍包含 Google 登录 SDK,其隐私声明覆盖认证相关信息。无需 Google 账号,我们不申请 Gmail、通讯录或 Drive 权限,也不因登录向 Apple 或 Google 上传录音。

本人声纹:分别选择与管理

声纹属于敏感生物识别信息,功能可选。本机模板、加密账号模板备份、远端声纹注册是独立能力。你单独同意账号模板同步后,模板会上传到我们的后台;单独同意远端注册后,注册音频发送至当前配置的 video.lmyiot.com 声纹服务建立特征。后续主动选择云端整理时,可发送短说话片段与本人已注册声纹比较,不注册其他说话人,不检索其他 SoundDay 用户的声纹。“可能是本人”仅作提示,不用于身份认证。

我们的远端注册表保存账号及服务标识、模型及同意版本、时间和状态,不保存注册原音或特征向量;这不代表下游服务不保留原音、特征、日志或备份。我们不宣称已核实该独立服务的零保留安排或部署地域。如果这些信息影响你的使用决定,请在授权前联系我们。

删除远端注册会立即停止新匹配,并排队向远端清理;网络异常可能延迟确认,App 会展示清理中。本机模板、账号模板备份、远端注册分别删除;不会自动改写历史说话人标签或删除录音,长期清理中请联系支持。

由你确认保存的记忆

记忆默认关闭。你明确开启后,可以核对并保存已整理录音中选中的内容。我们在中国大陆的账号服务器保存你确认的正文、分类、当前或过时状态、来源标题与时间,以及选中的转写原话、对应时间戳和说话人标签。此功能不额外上传音频、不额外调用 AI,也不会自动分析录音历史来学习习惯或建立个人画像。

保存后,服务器可在本账号最近 200 条已保存记忆中比较正文措辞,最多提示三条其他事项的相似内容,并显示原话和出处。这是文字匹配,不是 AI 语义判断;提示不会自动关联或改写记忆,跳过不改变数据,暂停后不再生成新提示。你可以明确关联同一事项的已保存记忆;这些账号内的关联按来源录音时间展示,可以解除而不删除记忆。后来的对话不会自动替代之前的决定。你可以修改、标为过时或忘记记忆。暂停后停止新增记忆和关联,已有内容仍可查看和管理。保存的摘录是独立副本:删除来源录音不会同时删除这些摘录;忘记记忆也不删除原始录音。“忘记全部”会删除已保存的记忆并关闭功能,注销账号也会从活动服务删除记忆。历史备份遵循下方保留与清理规则。

Apple 订阅与用量

订阅开放且你主动确认购买后,由 Apple 处理付款,我们不接收银行卡信息。后台验证交易和原始交易编号、账号关联令牌、购买及到期时间、试用与订阅状态和环境,用于额度、恢复购买、续费及退款处理。保存处理时长、任务和账期用量;已受理失败任务或删除录音不自动退额,同一已受理任务重试不重复扣减。

单一月订阅方案为符合 Apple 资格者试用七天、共 300 分钟,之后每个 Apple 付费账期 1,000 分钟、不累计。实际可购买优惠与当地价格以 App 内 Apple 返回信息为准。显示“订阅正在准备中”时保留现行额度,不发起购买。请在 Apple 订阅设置取消续费,卸载 App 或注销 SoundDay 不会自动取消。

反馈建议

保存反馈类别、正文、自愿填写的联系方式、账号关联、回执、时间、处理状态和最新回复。只有主动勾选才附加 App 与系统版本,不自动附加录音、转写、声纹或诊断日志。请勿提供密码或验证码。授权客服用于排查并回复;请在 App「我的反馈」刷新查看,不承诺邮件或推送通知。

保存、删除与安全

设备原音由你管理,云端录音及结果保留至你删除云端记录或注销账号。反馈保留至注销,提前删除可联系支持。注销删除活动云端内容、反馈和用量,并排队清理远端声纹。为防止订阅转移或重复发放,受限订阅归属记录含原账号关联令牌和交易字段会保留,不宣称匿名,本版本未设置固定自动清除期限,相关隐私请求可联系邮箱。

运行数据库备份受限访问,配置三十天保留并每日清理,过期副本可能等待下一次清理。活动数据删除不等于即时清除历史备份。Nginx 日志每日轮转最多十四期,可含 IP、路径、状态和时间;带签名的音频下载路径不记访问日志。不主动记录授权令牌、反馈正文或录音内容。服务商日志与备份独立管理,不应假定删除我们的活动数据就清除了所有服务商副本。安全措施无法保证绝对无风险。

权利与联系

你可以拒绝云处理或声纹注册、撤销麦克风权限、导出和删除数据,访问、更正、删除等请求请联系 44493547@qq.com。权利视所在地区而定,也可联系当地隐私监管机构。撤回不追溯取消已经发生的处理。本服务不面向十三岁以下儿童,不用于秘密监控;AI 和说话人判断可能错误,请核对原声。政策变化更新生效日期,新增用途应适当告知并取得许可。