Version covered: This policy describes SpeechSynk version 1.0.0 (build 17), available for TestFlight testing and prepared for App Store review. Onboarding does not record audio or generate a baseline score. Any leftover onboarding recording or simulated score from an earlier version is removed when the app starts. If you use an earlier test build, update it to use the controls described below, or contact support for deletion assistance.
SpeechSynk is a product of Midbrew. This policy explains how we handle information in the app and at speechsynk.com. Contact support@speechsynk.com with privacy questions or requests.
Your voice stays on your device
Microphone audio is processed in memory for delayed auditory feedback and speech analysis. SpeechSynk does not save audio recordings or upload audio or transcripts. Speech recognition is restricted to on-device processing. If on-device recognition is unavailable or permission is denied, local audio-level analysis continues without transcription.
Local practice and optional accounts
You can practice without an account. Session scores, exercise types, dates, and durations are saved on your device. Preferences also stay locally. Scores are estimates from completed practice sessions with recognized speech, not clinical assessments. If no speech is recognized or speech recognition could not start, no score, coaching request, or practice-history record is created. Session history uses iOS Data Protection.
If you sign in, Google Firebase Authentication handles your account identifier, name when supplied, and email address, including an Apple private relay address when chosen. Firebase handles password credentials for email sign-in. Saved session records are backed up to Cloud Firestore under your account ID. If you enable notifications, an Apple push notification token may also be stored with your account to deliver reminders.
Optional AI coaching
AI coaching is off until you choose to allow it. With your permission, SpeechSynk sends your exercise type, duration, fluency score, previous score when available, and speech-pattern category through Google API Gateway and Firebase to Anthropic to generate a coaching summary. These metrics can reveal information about your speech. Firebase authenticates the request and records usage counts to limit abuse.
The Anthropic request does not contain your audio, transcript, name, email, Firebase account ID, or session ID. The app displays the returned summary in memory. This does not make all practice information anonymous. You can turn off future AI sharing in Profile → Privacy and continue with general encouragement for sessions with recognized speech.
Turning off AI coaching or deleting your SpeechSynk account does not recall requests Anthropic has already processed. Anthropic describes its API retention, normally up to 30 days with contractual, safety, and legal exceptions, in its commercial data retention policy.
Crash reports and service diagnostics
Crash reporting is off by default. If you enable Share crash reports in Profile → Privacy, Google Firebase Crashlytics receives crash stack traces, app and device information, and installation identifiers to help us fix bugs. SpeechSynk does not attach your account ID, email, audio, or transcript to these reports. Installation identifiers mean these reports should not be described as anonymous.
You can stop future uploads in Profile → Privacy. Unsent reports are discarded when reporting is disabled. Reports already uploaded follow Firebase retention policies; Crashlytics normally retains crash data and associated identifiers for 90 days before starting removal. Google Analytics collection is disabled in the app.
Firebase account and sync services also send technical metadata such as device and operating-system details, app bundle ID, platform, and SDK versions to help Google maintain its services and understand SDK adoption. Firebase states that its user-agent metadata is not linked to a user or device identifier. This service diagnostic collection is separate from optional crash reports. See Firebase's Apple-platform data disclosures.
Service providers and support
Google API Gateway authenticates coaching and deletion requests. Google Firebase provides authentication, account storage, backend processing, service diagnostics, and optional crash reporting. Apple provides sign-in, on-device speech recognition, and push delivery. Anthropic provides optional AI feedback. Authorized Midbrew personnel and our providers can access stored service data as needed to operate, secure, and support SpeechSynk.
Support email contains the contact information and message you choose to send. The website and backend hosting providers may process network information such as IP addresses and request logs to deliver and protect the service. Providers may process information outside your country. We do not sell your personal information or use it for advertising tracking.
Account deletion and retention
Open Profile → Account → Delete account, confirm, then verify your Apple sign-in or password. Apple sign-in authorization is revoked as part of that flow. The backend blocks further account data access, removes your Firestore profile and all nested session and progress records, clears usage counters, and deletes your authentication account. If cleanup fails temporarily, the server retries it and the app provides a retry option. The app confirms completion only after the server acknowledges deletion and local cleanup succeeds.
Local cleanup removes practice history, preferences, cached coaching from older versions, app keychain items, and scheduled reminders on this device. Other devices clear local practice data after reconnecting and detecting an invalid account session. An offline device can retain its copy until it reconnects or the app is removed.
A signed deletion-recovery credential is stored in the device Keychain before deletion starts and is valid for seven days. It can check or finish an already authorized deletion, but cannot start a new deletion or access practice data. A minimal server receipt containing the account ID, receipt hash, and completion status is retained for seven days after completion to block old authentication tokens and allow retries, then removed by scheduled cleanup. Already processed AI requests, previously uploaded diagnostics, support correspondence, and service-provider security logs have separate retention policies and are not erased by deleting the Firestore account. Contact us for help with those records.
Without an account, you can remove local practice data by deleting the app from the device. Contact support@speechsynk.com if you cannot access the in-app deletion flow.
Security and your choices
Network requests use encrypted HTTPS connections. Cloud account data is protected by authentication and access rules. Audio is not stored. No system can guarantee absolute security. You can choose whether to sign in, enable AI coaching, share diagnostics, or enable reminders, and can contact us to request access, correction, or deletion of information.
Adults and policy updates
SpeechSynk is designed for adults aged 18 and older. Contact us if you believe a child has provided personal information. We update this page when our data practices change and request new consent before materially expanding optional AI sharing.