Privacy Policy

SpeechBuddy is a communication assistant for people whose speech is hard for strangers to understand. Because we listen to a real person's voice to do that work, we take privacy seriously. This document explains — in plain language — what we collect, why, where it goes, and what you can do about it.

1. Who "we" are

SpeechBuddy is operated by LiberaVoice LLC, an Ohio limited liability company. When this policy says "we," "us," or "our," we mean LiberaVoice LLC. You can reach us at Speechbuddy33@gmail.com for any privacy question.

Registered address: 562 Legends Row, Avon Lake, OH 44012

2. Who this policy is for

SpeechBuddy accounts are set up and managed by a caregiver — usually a parent, spouse, sibling, or adult child of the person who uses the app to communicate. The daily user — the person whose speech SpeechBuddy interprets — may be a minor or an adult, and may or may not read this policy directly.

By creating a SpeechBuddy account, the caregiver confirms that they are at least 18 years old and either (a) the daily user themselves, (b) the parent or legal guardian of the daily user, or (c) an authorized caregiver acting with the daily user's or their guardian's consent. This policy binds both the caregiver and, to the extent applicable, the daily user.

3. What we collect

Account information

• Email address (stored as a one-way hash for lookup and a reversibly-stored copy for password reset emails)
• Password (stored bcrypt-hashed; we never see it in plaintext)
• Username and, optionally, the daily user's display name (e.g. "Maggy")
• Signup date and last-login timestamp

Communication content

• Phrases you save to your library, including the text, the categories you assign, and any image you attach
• Voice recordings the daily user makes of themselves speaking those phrases — stored as audio files on our servers, associated with the specific phrase they were recorded for
• Voice model (a small LoRA adapter, roughly 25 MB) derived from those recordings, used to improve how well SpeechBuddy understands that user's speech
• Utterances — every time the daily user says something to SpeechBuddy, we log what was heard (Whisper's transcription), what phrase matched (if any), and what SpeechBuddy spoke back
• Caregiver notes you optionally write about the daily user in Settings → About (used to personalize AI suggestions only if you opt in — off by default for AI use, always visible to you)

Usage & telemetry

• Which features you use, in what mode (Phrase Match, Sentence Builder, Free Form), and how the results turned out
• Suggestion impressions and taps (which suggestions we showed, which you chose, which you dismissed)
• Anonymous performance metrics (page load, transcription latency, speed to complete a sentence)
• Client-side error reports (JavaScript errors, network failures)

Support & coaching

• Conversations with the in-app support agent and coaching cards you saw or acted on

What we do not collect

• We do not collect precise location.
• We do not use third-party advertising or marketing cookies.
• We do not run analytics beacons to Google Analytics, Meta, Amplitude, Mixpanel, or similar third-party platforms.
• We do not record the daily user's microphone continuously — audio is only captured while a recording session is explicitly active (waveform showing, "recording" state visible).

4. Why we collect it

We collect each of the above for one of four reasons:

• To provide the service. An account + a library of phrases + voice recordings is what makes SpeechBuddy work at all. Utterances are what let the app match what the daily user said to the correct saved phrase.
• To improve the service for you. Voice recordings train a per-user voice model that stays on our servers and is used only to interpret your daily user's speech, not anyone else's. Usage data lets us surface the right next-phrase suggestions and coaching cards.
• To improve the service overall. Aggregated, de-identified metrics (transcription accuracy, feature adoption, latency) tell us where SpeechBuddy is working and where it isn't. We do not use identifiable voice recordings for aggregate improvement without your explicit opt-in via Settings → "Help SpeechBuddy learn."
• To keep the service safe and running. Error logs and usage patterns help us catch bugs, prevent abuse, and monitor ourselves against runaway costs.

5. Where your data goes (third parties)

Running SpeechBuddy requires four categories of third-party processors. Each one sees only what it needs to do its job.

• Anthropic processes text prompts for AI features (next-phrase suggestions, phrase categorization, image search hints, caregiver helper chat, and the support agent). We send text context — phrases, categories, session state — not voice audio. Anthropic's data-handling terms are at anthropic.com/legal/privacy.
• ElevenLabs synthesizes text-to-speech audio when SpeechBuddy speaks a phrase aloud in a caregiver-selected voice. We send text; we receive audio. We do not upload the daily user's voice recordings to ElevenLabs. ElevenLabs' policy is at elevenlabs.io/privacy.
• Fly.io hosts the SpeechBuddy application and database. All account data, phrases, voice recordings, and voice models are stored on Fly.io infrastructure. Their policy is at fly.io/legal/privacy-policy.
• Email delivery. Password resets, support escalations, and account-related notifications are delivered through Google (via Gmail's SMTP relay). The recipient's email address and the message body are the only data shared. Google's policy is at policies.google.com/privacy.

What we do not share

• We do not sell personal information to anyone, for any purpose.
• We do not share your data with advertisers, data brokers, or marketing platforms.
• We do not use your voice recordings to train models available to other users or to third parties.

6. How long we keep it

See the separate Data Retention Statement for exact retention windows for each data type. In short:

• Your account, phrases, voice recordings, and voice model stay until you delete them.
• Most usage telemetry is kept on a rolling window (30–90 days depending on category), then automatically pruned.
• A minimal cost and health log is kept longer for our own accounting and safety monitoring.

7. Your rights & how to exercise them

Regardless of where you live, you have the following rights over your SpeechBuddy data. We honor these for every user, not only those in states or countries that legally require them:

• Access. Ask us what we have on you. Reply to a request within 30 days at Speechbuddy33@gmail.com.
• Correction. Edit your username, display name, caregiver notes, and library content directly in Settings and the Library tab.
• Deletion. Delete your entire account and all associated data from Settings → Danger Zone → Delete account. Deletion is immediate and cascades to voice recordings, voice model, phrase library, translation history, and telemetry.
• Portability. Ask us for a machine-readable export of your account content (phrases and settings). We'll send a JSON export within 30 days. Voice recording exports are supported on request.
• Opt-out of "Help SpeechBuddy learn." Toggle at the top of Settings. When off, identifiable usage data is not used for improving matching accuracy across users.
• Object & restrict. If you want a specific type of processing halted (e.g. stop AI features but keep your account active), tell us — we'll accommodate what we can technically support.

California residents: the rights above satisfy your CCPA / CPRA rights to know, delete, correct, and opt out of sale (there is no sale). We do not process sensitive personal information for advertising or profiling.

8. Children's data

Many SpeechBuddy daily users are children. In every case, the account is created by an adult caregiver acting under one of the consent categories in Section 2. Because the caregiver is the account holder and the consenting party, we treat all account-related consents and communications as being with the caregiver.

If we learn we have collected personal information from a child under 13 without verifiable parental consent, we will delete it. Contact Speechbuddy33@gmail.com for concerns.

9. Security

Passwords are hashed with bcrypt. All traffic between your device and SpeechBuddy is served over HTTPS. Account emails are stored using a keyed hash for lookup plus reversible ciphertext for password-reset delivery. Voice recordings are stored on encrypted volumes.

No system is perfect. If you believe your account or someone else's has been compromised, contact us at Speechbuddy33@gmail.com immediately.

10. International users

SpeechBuddy is operated from and hosted in the United States. If you use SpeechBuddy from outside the U.S., you consent to your data being processed in the U.S. under U.S. law.

11. Changes to this policy

If we make material changes to how we handle your data, we'll tell you in-app before the change takes effect, and we'll bump the version number at the top of this page. Minor clarifications (typos, better phrasing) may be made without notice.

12. Contact

Questions, requests, or complaints: Speechbuddy33@gmail.com.

If you're not satisfied with how we've handled a privacy request, you may also contact your state attorney general or the Federal Trade Commission.