OutfitSignal — Privacy Policy

Last updated: September 22, 2026

OutfitSignal ("we", "us") provides subjective AI-generated outfit-style feedback. No account is required. Submitted outfit photos and results are not stored on our servers unless you separately and explicitly opt in after a result to share a resized photo and sanitized result for grading evaluation.

Age requirement

OutfitSignal is intended only for adults 18 years of age or older. Before entering the app, each user must affirm that they are at least 18. On supported Apple devices, the app also requests an age range through Apple’s Declared Age Range framework. A system response identifying a user as under 18 blocks access. We do not receive a birth date, and we do not transmit or retain the age-range response. We store only that the current installation completed the consent screen.

Information processed

We do not require your name or email address to use outfit grading or the anonymous feedback form, and we do not collect contacts or precise location. Except for the optional Meta measurement described below, OutfitSignal does not access or send an advertising identifier. If you contact us by email because you need a reply, we receive your email address and the contents of that message.

How a submission is used

When you request style feedback, the app sends a resized copy of your photo and any optional occasion-coach brief over an encrypted connection to our Cloudflare Worker. The Worker forwards them to Google Gemini to produce feedback focused on visible garment fit, color harmony, styling cohesion, and an optional chosen style, occasion, and intended vibe. The service is instructed not to evaluate personal traits such as body shape, skin tone, face, hair, age, gender, or attractiveness.

Outfit feedback is limited to a single visible person. If Gemini reports that two or more people are visible, including background bystanders, the app rejects the photo without producing style feedback. OutfitSignal does not ask Gemini to identify anyone or infer age.

Version 1.2.0 uses the Apple install-attribution and optional Meta measurement controls described below.

Support feedback

The Support page offers an anonymous feedback form. It requires a category and message and lets you optionally provide your device model and iOS version. Please do not include photos, names, contact details, or sensitive information. When you submit the form, our Worker stores the category, message, response language, any optional device details, a random receipt ID, and the submission time in a Cloudflare D1 database so we can investigate problems and improve OutfitSignal. The receipt ID is shown after submission so you can reference the entry later.

The feedback table does not contain your name, email address, IP address, or user agent. Cloudflare may separately process standard network metadata to deliver and secure the request, as described under Service providers. The form cannot be used for a personal reply; use the email link on the Support page if you need one.

Grading activity metrics

For each fresh request to grade an outfit, OutfitSignal increments a coarse hourly aggregate in Cloudflare D1. The aggregate records whether the request succeeded, was rejected, or failed; a fixed reason category; a latency range; the app platform and version; and the grader model, prompt, and schema versions. These metrics help us understand grading volume and reliability.

The aggregate does not contain the photo, score, result, style or occasion context, exact event time, IP address, account, device identifier, advertising identifier, user agent, or request ID. It measures requests and successful grades, not unique people.

Optional grading feedback

After a fresh result, the app may ask whether the feedback was useful. Sending a rating is optional and does not require an account. We store the rating, selected reason tags, response language, controlled style/occasion/vibe identifiers, and non-sensitive grader version identifiers in Cloudflare D1. We do not attach an account, device, advertising identifier, IP address, or user agent to that record.

The outfit photo and generated result remain excluded unless you separately turn on the unchecked photo-sharing option and confirm the current consent notice. With that explicit choice, the resized JPEG and sanitized result are stored in a dedicated private Cloudflare R2 bucket and the grading-feedback record. Explicitly shared material may be accessed only by authorized OutfitSignal reviewers for grading evaluation. Choosing not to share has no effect on app features.

Storage and retention

Service providers

Google Gemini processes submitted images, optional context, and generated responses. OutfitSignal uses Gemini Paid Services through a Google Cloud project with active billing. Under Google's Paid Services terms, Google states that it does not use submitted prompts, images, or responses to improve its products. Google may log or retain this data for a limited period for abuse prevention, service safety and security, and required legal or regulatory disclosures. See the Gemini API Terms and Google Privacy Policy. Do not submit sensitive, confidential, or unnecessary personal information.

Cloudflare delivers and protects the Worker endpoint. Standard request metadata — including IP address, request time, user agent, request path, response status, and related diagnostics — may be processed for security, reliability, troubleshooting, and abuse prevention. OutfitSignal does not intentionally write submitted photos, occasion briefs, or Gemini results to Worker logs. Cloudflare hosts D1 storage for anonymous support feedback, optional grading feedback, and coarse hourly grading activity aggregates, plus a dedicated private R2 bucket for explicitly shared grading photos. Feedback records and aggregates are retained as described above; the feedback tables do not store IP addresses or user agents.

Expo Application Services. A distributed build configured to use EAS Update may send operating-system, app and runtime version, randomized installation-token, update-download, and prior fatal-error information to Expo. Over-the-air EAS Update is disabled in the live project configuration, so distributed builds do not use EAS Update to check for or download JavaScript updates. See the Expo Privacy Policy.

Apple install attribution. Once per launch while active, version 1.2.0 asks Apple's SKAdNetwork to register conversion value 0, retrying after a failed or backgrounded attempt, or uses Apple's legacy install-registration API on older supported systems. This is basic install attribution, not OutfitSignal's custom conversion event. The request does not use the Meta SDK, access an advertising identifier, or provide Apple any OutfitSignal photo, result, or user identifier. It occurs even if tracking is declined or Meta measurement is off. SKAdNetwork postbacks are delayed and aggregated; calling the API does not prove that an end-to-end postback was delivered.

Meta advertising measurement (optional; only in builds that offer it). Version 1.2.0 includes the option where it is distributed. It is not required to use OutfitSignal. OutfitSignal keeps the Meta SDK uninitialized, with automatic SDK initialization, advertising-ID collection, and automatic app-event logging disabled, unless and until a user who has completed the adult gate grants App Tracking Transparency and has not already turned measurement off in Settings. After the first successful fresh result is shown, the app may present Apple's native ATT prompt once, without a separate pre-prompt. Granting that system permission records the in-app opt-in; the Settings control remains available. OutfitSignal then uses Meta's native activateApp function for app-install/activation measurement. It sends one outfitsignal_first_check_completed event with no app-supplied custom parameters for the first subsequent successful fresh (not cached or restored) outfit check begun while those gates are open. The result that preceded the prompt, and any other result completed before permission and opt-in, is not replayed later as this custom event. It does not send the outfit photo, score, result, style, occasion, vibe, context, age or age-range response, email address, or an OutfitSignal user ID to Meta.

When enabled, the Meta SDK may receive device, app, and network metadata and the device advertising identifier, and Meta may link the events and metadata with information from other apps, websites, or services for advertising attribution and measurement. This measurement is therefore not anonymous. See Apple's User Privacy and Data Use guidance, the Meta Privacy Policy, and Meta Business Tools Terms.

We require every provider that receives user data from OutfitSignal to provide the same or equivalent protection of user data described in this policy through its applicable service terms, privacy commitments, and data-processing terms. We configure and use these providers only for the functions described above and do not sell this information. Except for Apple's aggregate install attribution and optional Meta measurement after the recorded opt-in and Apple permission described above, we do not authorize provider use for advertising attribution, advertising, or cross-app tracking. Where we control provider settings, we limit access, use, and retention to what is reasonably necessary to provide, secure, and troubleshoot OutfitSignal.

Children

OutfitSignal is not directed toward, and may not be used by, anyone under 18. We do not knowingly collect personal information from children. Users are instructed to submit only a photo of themselves with no one else visible.

Your choices

In a build that offers Meta measurement, you can decline Apple's tracking prompt or turn the measurement choice off in Settings before or after the prompt. None of those choices limits use of OutfitSignal. An earlier explicit Settings opt-out is respected. Turning the in-app choice off stops new OutfitSignal-level measurement events and disables the SDK's advertiser-tracking, advertising-ID-collection, and automatic-event settings; the next launch will not initialize the SDK unless the in-app choice and ATT authorization are again in place. Withdrawing the existing AI consent also turns Meta measurement off. Apple install attribution is separate and may still register a basic, nontracking install when Meta measurement or ATT is declined. Withdrawal cannot recall data already transmitted, shorten Meta's applicable retention, or guarantee deletion of an event the SDK had already queued for transmission.

You can stop submitting photos or occasion briefs at any time. To withdraw consent for future AI processing, open the menu from any app screen, choose Privacy & AI consent, then confirm Withdraw consent. The app immediately blocks new photo submissions, then attempts to record the withdrawal on the device. If that first record cannot be saved, the app reports an error and keeps the existing consent so you can retry; no new request is admitted during the attempt. Once withdrawal is recorded, the app deletes the local improvement album and local result cache and clears its stored consent. If local erasure cannot complete, app access remains locked and cleanup is retried before fresh consent can take effect. Withdrawal does not undo a processed request or shorten a provider's retention period.

You may ask us to delete a support-form submission by emailing the address below and including the receipt ID shown after submission. If you did not save the receipt ID, include the approximate submission date, category, and enough of the message for us to try to locate it; we may be unable to identify a particular entry if those details are insufficient. An email deletion request necessarily gives us your email address and is handled separately from the anonymous feedback table.

Subjective feedback

Scores and suggestions are AI-generated style opinions for inspiration and entertainment. They are not objective judgments or professional, medical, or other advice.

Security

Images and support-form submissions are transmitted over HTTPS. Explicitly shared grading-feedback images are kept in a dedicated private bucket accessible only to authorized OutfitSignal reviewers. We limit request frequency and size to reduce abuse. No storage or transmission method is completely secure, and retained feedback remains subject to the limits above.

Changes and contact

Material updates will be reflected by the “Last updated” date. Questions can be sent to davidhsiao510@icloud.com. Please do not email personal outfit photos.

OutfitSignal — 隱私權政策

最後更新:2026 年 9 月 22 日

OutfitSignal(以下稱「我們」)提供 AI 產生的主觀穿搭風格回饋,不需要帳號。除非你在評分後另外明確選擇分享縮小後的照片與經安全處理的結果供評分評估使用,否則穿搭評分照片與結果不會儲存在我們的伺服器。

年齡要求

OutfitSignal 僅供年滿 18 歲的成年人使用。進入 App 前,每位使用者都必須確認已年滿 18 歲。 在支援的 Apple 裝置上,App 也會透過 Apple 的「已聲明年齡範圍」架構要求年齡範圍;若系統回覆顯示 使用者未滿 18 歲,App 將阻止使用。我們不會取得出生日期,也不會傳送或保留年齡範圍回覆;只會在 裝置上儲存目前安裝已完成同意畫面的狀態。

處理的資訊

使用穿搭評分或匿名意見表單不需要提供姓名或電子郵件地址;我們也不蒐集聯絡人或精確位置。除下述選填的 Meta 成效衡量外,OutfitSignal 不會存取或傳送廣告識別碼。如果你因需要回覆而選擇寄送電子郵件,我們會收到你的電子郵件地址與郵件內容。

提交內容的使用方式

當你要求風格回饋時,App 會透過加密連線將縮小後的照片與選填的場合教練說明傳送至 Cloudflare Worker, 再轉交 Google Gemini,針對可見服裝的合身度、色彩和諧、風格完整度、 選擇的風格、場合與想呈現的感覺產生建議。系統會明確要求 AI 不得評估身形、膚色、臉部、頭髮、年齡、性別或吸引力等個人特徵。

穿搭回饋僅支援一位可見人物。若 Gemini 回報照片中有兩位以上人物(包括背景中的路人), App 會拒絕該照片,不會產生風格回饋。OutfitSignal 不會要求 Gemini 識別任何人物或推測年齡。

1.2.0 版使用下文所述的 Apple 安裝歸因與選填 Meta 成效衡量控制。

支援意見

支援頁面提供匿名意見表單。表單要求選擇類別並填寫訊息,裝置型號與 iOS 版本則為選填。 請勿提供照片、姓名、聯絡方式或敏感資訊。送出表單後,Worker 會將類別、訊息、回覆語言、選填的裝置資訊、 隨機回條編號及提交時間儲存在 Cloudflare D1 資料庫,供我們調查問題與改進 OutfitSignal。提交後會顯示回條編號, 方便你日後查詢該筆意見。

意見資料表不會儲存你的姓名、電子郵件地址、IP 位址或裝置瀏覽器資訊。Cloudflare 仍可能在資料表之外 處理標準網路中繼資料,以傳送及保護請求,詳情請見「服務供應商」。匿名表單無法讓我們直接回覆; 如需回覆,請使用支援頁面的電子郵件連結。

評分活動指標

每次新的穿搭評分請求,OutfitSignal 都會在 Cloudflare D1 的每小時彙總中增加計數。彙總資料包含請求是否成功、 被拒絕或失敗、固定原因類別、延遲範圍、App 平台與版本,以及評分模型、提示詞和資料結構版本。這些指標用於了解評分量與可靠性。

彙總資料不包含照片、分數、結果、風格或場合情境、精確事件時間、IP 位址、帳號、裝置識別碼、廣告識別碼、 裝置瀏覽器資訊或請求 ID。這些資料衡量的是請求與成功評分次數,而非不重複使用者人數。

選填的評分回饋

完成新的穿搭評分後,App 可能詢問建議是否有幫助。提交評價完全自願且不需要帳號。我們會在 Cloudflare D1 儲存評價、選取的原因標籤、回覆語言、受控的風格/場合/感覺識別碼,以及不含敏感資訊的評分器版本資料。 該紀錄不會連結帳號、裝置、廣告識別碼、IP 位址或裝置瀏覽器資訊。

穿搭照片與產生的結果預設不會包含在評價中。只有你另外開啟預設未勾選的照片分享選項,並確認目前的同意說明後, 縮小後的 JPEG 與經安全處理的結果才會儲存在專用的私有 Cloudflare R2 與評分回饋紀錄中。明確分享的資料僅限經授權的 OutfitSignal 審查人員基於評分評估目的存取。不分享不會影響任何 App 功能。

儲存與保留

服務供應商

Google Gemini 會處理提交的圖片、選填情境與產生的回覆。OutfitSignal 透過已啟用有效計費的 Google Cloud 專案使用 Gemini 付費服務。依 Google 的付費服務條款,Google 表示不會使用提交的提示、圖片或回覆 來改進其產品。Google 可能在有限期間內為防止濫用、維護服務安全,以及必要的法律或監管揭露而記錄或保留這些資料。 詳情請參閱Gemini API 條款與 Google 隱私權政策。請勿提交敏感、機密或非必要的個人資訊。

Cloudflare 負責傳送與保護 Worker 端點,並可能為安全、可靠性、疑難排解與防止濫用處理 IP 位址、請求時間、裝置瀏覽器資訊、請求路徑、回應狀態及相關診斷資訊。OutfitSignal 不會刻意將提交照片、 場合說明或 Gemini 結果寫入 Worker 記錄。Cloudflare 以 D1 儲存匿名支援意見、選填評分回饋及每小時粗略評分活動彙總,並以專用私有 R2 儲存空間保留使用者明確分享的評分照片。資料依上述期限保留;回饋資料表不會儲存 IP 位址或裝置瀏覽器資訊。

Expo Application Services。若發行版本設定為使用 EAS Update,可能會向 Expo 傳送作業系統、 App 與執行環境版本、隨機安裝代碼、更新下載狀態及先前的嚴重錯誤資訊。目前上線專案設定 已停用 EAS Update 的無線更新,因此發行版本不會使用 EAS Update 檢查或下載 JavaScript 更新。詳情請參閱 Expo 隱私權政策。

Apple 安裝歸因。1.2.0 版每次啟動會在 App 位於前景時要求 Apple SKAdNetwork 登記一次轉換值 0; 若嘗試失敗或 App 已進入背景,則會重試;在較舊的 支援系統上則使用 Apple 的舊版安裝登記 API。這是基本安裝歸因,不是 OutfitSignal 的自訂轉換事件。 此請求不會使用 Meta SDK、不會存取廣告識別碼,也不會向 Apple 提供 OutfitSignal 照片、結果或使用者識別碼。 即使使用者拒絕追蹤或關閉 Meta 成效衡量,仍會進行此登記。SKAdNetwork 回傳是延遲且彙總的;呼叫 API 並不能證明端對端回傳已送達。

Meta 廣告成效衡量(選填;僅限提供此選項的版本)。1.2.0 版在實際發行的地區提供此選項。使用 OutfitSignal 並不以加入此功能為條件。 OutfitSignal 會讓 Meta SDK 維持未初始化,並關閉 SDK 自動初始化、廣告識別碼蒐集及自動 App 事件記錄。第一次成功完成的全新結果顯示後,App 可能直接顯示一次 Apple 原生 ATT 提示,不另設預先提示。只有已完成成人資格確認的使用者允許 App Tracking Transparency,且先前 未在設定中關閉成效衡量後,功能才會啟用;允許系統權限也會記錄 App 內的選擇加入,設定控制仍會保留。 OutfitSignal 接著使用 Meta 原生 activateApp 功能衡量 App 安裝/啟用。日後在所有條件已成立時開始的第一次成功全新(非快取或相簿還原)穿搭檢查,會傳送一個不含 App 自訂參數的 outfitsignal_first_check_completed 事件;觸發提示的結果及其他在權限與選擇加入前已完成的結果, 都不會在日後重送為此自訂事件。 SDK 仍可能附加下述中繼資料。我們不會向 Meta 傳送 穿搭照片、分數、結果、風格、場合、想呈現的感覺、 情境、年齡或年齡範圍回覆、電子郵件地址或 OutfitSignal 使用者 ID。

功能啟用時,Meta SDK 可能接收裝置、App 與網路中繼資料及裝置廣告識別碼;Meta 也可能將這些事件和 中繼資料與其他 App、網站或服務的資訊連結,用於廣告歸因與成效衡量。因此,此成效衡量並非匿名。 詳情請參閱 Apple 的「使用者隱私權與資料使用」指引、 Meta 隱私權政策及 Meta 商業工具條款。

我們要求每個接收 OutfitSignal 使用者資料的供應商,依其適用服務條款、隱私承諾與資料處理條款,提供與本政策 所述相同或同等的使用者資料保護。我們只為上述功能設定與使用這些供應商,也不會出售這些資訊。除 Apple 的彙總安裝歸因,以及已記錄選擇加入且 Apple 權限已允許的選填 Meta 成效衡量外,我們不會授權供應商將資訊用於廣告歸因、廣告或跨 App 追蹤;在我們可控制設定的範圍內,會將存取、使用與保留限制在提供、保護及排解 OutfitSignal 問題 所合理需要的程度。

兒童

OutfitSignal 不以未滿 18 歲者為對象,未滿 18 歲者也不得使用。我們不會刻意蒐集兒童的個人資訊。 使用者會被要求只提交僅有自己入鏡、沒有其他人物的照片。

你的選擇

在提供 Meta 成效衡量的版本中,你可以拒絕 Apple 的追蹤提示,或在提示前後於設定中關閉成效衡量; 這些選擇都不會限制你使用 OutfitSignal。先前在設定中明確關閉的選擇會受到尊重。關閉 App 內選項後,OutfitSignal 會停止新的 App 層級 成效衡量事件,並停用 SDK 的廣告追蹤、廣告識別碼蒐集與自動事件設定;除非日後 App 內選擇與 ATT 授權再次成立, 下一次啟動時不會初始化 SDK。撤回既有的 AI 同意也會關閉 Meta 成效衡量。Apple 安裝歸因是獨立處理;即使拒絕 Meta 成效衡量或 ATT, 仍可能登記基本、非追蹤的安裝。撤回無法取回已傳送的資料、縮短 Meta 適用的保留期間, 也無法保證刪除 SDK 先前已排入傳送佇列的事件。

你可以隨時停止提交照片或場合說明。若要撤回之後的 AI 處理同意,請從 App 任一畫面開啟選單,點選 「隱私權與 AI 同意」,再確認「撤回同意」。App 會立即停止新的照片提交,並嘗試在裝置上記錄撤回。 若一開始無法儲存該記錄,App 會顯示錯誤並保留現有同意,讓你可以重試;嘗試期間不會送出新的請求。撤回記錄完成後,App 會刪除本機改造相簿與本機結果快取,並清除已儲存的同意狀態。若無法完成本機刪除,App 會維持鎖定,並在新的同意生效前重試清理。撤回同意不會取消已完成的請求,也不會縮短服務供應商的適用保留期間。

你可寄信至下方地址,並提供提交後顯示的回條編號,要求刪除透過支援表單提交的意見。若未保存回條編號, 請附上大約的提交日期、類別及足以辨識的訊息內容,讓我們嘗試尋找;若資訊不足,我們可能無法找出特定紀錄。 刪除請求電子郵件必然會向我們提供你的電子郵件地址,並與匿名意見資料表分開處理。

主觀風格回饋

分數與建議是 AI 產生的風格觀點,僅供靈感與娛樂,不是客觀判定,也不是專業、醫療或其他建議。

安全、變更與聯絡

照片與支援表單內容都透過 HTTPS 傳輸。使用者明確分享的評分照片會保留在專用私有儲存空間,僅限經授權的 OutfitSignal 審查人員存取。我們也限制請求頻率與大小以降低濫用。沒有任何儲存或傳輸方式能保證完全安全;保留的回饋受上述期限限制。政策若有重大更新,會變更「最後更新」日期。 如有問題或要提出刪除請求,請寄信至 davidhsiao510@icloud.com, 請勿透過電子郵件傳送個人穿搭照片。