Parlify Privacy Policy

Effective date: 9 October 2026

This Privacy Policy explains how Parlify handles your personal data when you use our mobile application to analyse your French pronunciation, receive estimated accent matches and pronunciation scores, or voluntarily contribute recordings to help improve the project.

1. Who is responsible for your data?
Parlify is operated by PLUS GAMES, SARL, 61 Rue de Lyon, 75012 Paris, France (“we”, “us” or “our”), acting as the data controller.
For privacy questions or requests concerning your personal data, contact [email protected].

2. Voice recordings and accent analysis
When you start a recording, Parlify uses your device’s microphone to record you reading the text displayed in the app.
Your recording is uploaded to a server operated by us. The analysis takes place on that server, rather than entirely on your device. We use our own models and open-source models, including Meta’s wav2vec 2.0, running on our server.
No recording is sent to Meta or to any external AI inference service. Our hosting provider supplies the infrastructure on which we operate the service, as described in section 6.
We process your recording and the speech features extracted from it to generate estimated accent matches, pronunciation scores and explanations.
For an ordinary analysis, the uploaded recording and temporary processing files are deleted when processing ends, whether the analysis succeeds or fails, normally within one minute of upload. Results remain temporarily available in server memory as described in section 5.
If you separately choose to contribute to the project, the recordings you submit as part of that contribution are retained under the voluntary contribution terms in section 4.
We request microphone permission before recording. You can refuse or revoke this permission in your device settings; without microphone access, the recording and accent-analysis features cannot work. Parlify records only when you initiate a recording and does not listen continuously in the background.
Please read the provided text and avoid including private information or recording other people.

3. What data we handleVoice recordings and analysis data
We process the recording you submit, acoustic features extracted from it, estimated accent percentages, pronunciation scores and the explanations generated for you.
Technical and connection data
For each analysis, the application server records a short technical log containing a random job identifier, the uploaded file’s size and type, processing time and any error code. When a contribution is created, the application also logs its contribution code.
These application logs do not contain your IP address. Your IP address is nevertheless processed when your device connects to our infrastructure.
Our HTTPS gateway, Caddy, keeps no access log. When a connection fails during a request, such as an interrupted upload, it writes a warning entry containing your IP address, the requested address and technical headers such as your device’s User-Agent, but never the audio recording.
These entries are retained in the server’s system journal as described in section 5.
For voluntary contributions, we also process the identifiers described in section 4 to limit repeated submissions.
Data stored on your device
The app stores a random device identifier and, if you contribute to the project, your contribution code in its private storage.
The app deletes its local copy of your recording as soon as it is no longer needed, at the latest when you leave the results screen.
In earlier versions, a local copy remains in the app’s private cache until it is cleared by the operating system or you clear the app’s data or uninstall it.
Analysis results are not saved persistently on your device. They disappear when you leave the results screen.
Device cloud backups are addressed in section 5.
Voluntary contribution data
If you separately consent to contribute, we collect the recordings, questionnaire answers and associated contribution information described in section 4.
Support correspondence
If you contact us, we process your email address, the contents of your message and any information or attachments you choose to provide.
Accounts, purchases, analytics and advertising
Parlify has no user accounts, no in-app purchases, and no analytics or advertising SDKs.
Deleting an audio recording does not necessarily delete technical logs, support correspondence or a separately submitted voluntary contribution. Their retention is described below.

4. Purposes and legal bases
For users in the European Economic Area, we rely on the following legal bases:

• Your consent to record, upload and analyse your voice, and your separate consent if you choose to contribute recordings for research and model improvement.
• Our legitimate interests in responding to enquiries, maintaining a reliable and secure service, and preventing abuse, limited to the data necessary for these purposes.
• Legal obligations when handling data-protection requests or retaining records required by applicable law.

You may withdraw your consent by contacting [email protected]. You can stop future recordings by revoking microphone permission in your device settings. Withdrawal does not affect the lawfulness of processing carried out before consent was withdrawn.
Ordinary accent analysis
We use your recording to provide the analysis you request.
We do not use recordings or identifiable speech features collected solely for an ordinary accent analysis to train or improve AI models. Research use requires the separate voluntary contribution process described below.
Voluntary research contributions
After an analysis, you may choose “Help the project” (“Aider le projet”). Participation is optional and requires your agreement on a separate consent screen. You must be 18 or older to contribute.
If you consent and submit a contribution, you share:

• The recording of the text you just read.
• A second recording in which you read a list of 84 words.
• Your answers to a short questionnaire about your country of birth, the country or countries where you grew up, your parents’ country of origin, your native language or languages, your French level, how many years you have been learning French, your sex and your age range.

We use these recordings and questionnaire answers only for research and to improve Parlify’s accent recognition, including training and evaluating our models.
We record the date and version of your consent and associate your submission with a contribution code.
To limit abuse, we allow one contribution per device and up to five contributions per network. For this purpose, we retain salted cryptographic fingerprints derived from a random device identifier and your IP address. These fingerprints are pseudonymous identifiers used to detect repeated submissions; they do not make your contribution anonymous.
The fingerprints, consent date and consent version are stored with your contribution and deleted together with it, after five years or earlier following a valid deletion request.
Contributions are retained for five years from submission, then deleted. They are not included in server backups.
At the end of the submission process, the app displays your contribution code. Keep this code: it allows us to locate your contribution without requiring an account.
To request access to or deletion of your contribution, or to withdraw your consent to its further use, email [email protected] and quote your contribution code.
Declining to contribute does not prevent you from using ordinary accent analysis. Uninstalling the app does not delete a contribution already submitted to our server.

5. How long we keep your data

Application logs contain a short job identifier, file size and type, processing time, error codes and, when a contribution is created, its contribution code. They contain no IP address. Caddy warning entries may contain an IP address and other connection details as described in section 3.
Deleting a contribution removes its recordings, questionnaire answers, fingerprints and associated consent records. A contribution code already recorded in a technical log may remain until that log is removed under the journal’s retention process.
Backups
Server-side recordings, analysis results and voluntary contributions are not included in backups.
From version 6 onward, the Android app’s data is excluded from Android cloud backups.
Earlier versions allowed Android cloud backup of app data, including the random device identifier and contribution code. Copies created by earlier versions may remain in existing backups, subject to your backup provider’s settings and retention rules.
Where a specific legal obligation or dispute requires us to retain particular records for longer, we restrict their use to that purpose and retain them only for the applicable period.

6. Who can receive your data?
Access is limited to authorised personnel and service providers who need it to operate Parlify, conduct the research to which you have consented, or respond to your requests.
Service providers acting on our behalf must follow our instructions and applicable data-protection obligations.
Our server hosting provider is OVH SAS, France. The server is located in France.
Our email service provider is Zoho, using its EU data centres, which processes your email address, messages and attachments when you contact us.
Using Meta’s wav2vec 2.0 model on our server does not involve sending your recordings to Meta.
If you use an earlier app version with Android cloud backup enabled, your device’s backup provider may also process the app data included in those backups, as described in section 5.
We may disclose limited information where required by law or where necessary to establish, exercise or defend legal claims.
We do not sell your personal data.

7. International transfers
The processing we carry out to provide accent analysis, store voluntary contributions and handle support emails takes place within the European Economic Area, using OVH hosting in France and Zoho’s EU data centres.
Running Meta’s wav2vec 2.0 model on our server involves no transfer of data to Meta.
For cloud backups created by earlier Android app versions, processing locations depend on your device’s backup provider and account settings. These backups are separate from our OVH hosting and Zoho email services.

8. Security
We use encrypted connections for recording uploads and appropriate technical and organisational safeguards, including restricted access to server data and deletion controls.
Voluntary contributions are associated with a contribution code. Fingerprints used to limit repeated submissions are salted, but neither these measures nor the absence of an account makes voice recordings anonymous.
No system can guarantee absolute security. We review our safeguards in light of the nature of voice data and the risks associated with its processing.

9. What your results mean
Parlify compares your pronunciation patterns with reference accent profiles.
Accent percentages represent estimated similarities within the profiles supported by the model. They are not verified statements about your nationality, ethnicity, ancestry or identity.
Accent and Rhythm scores are estimates relative to the native speakers represented in our reference corpus. They are not standardised language qualifications or definitive measures of your overall French proficiency.
Results may vary depending on the recording, background noise, pronunciation, the reference corpus and model limitations.
We do not use your voice for identity verification. The automated analysis provides informational feedback; we do not use it to make decisions producing legal or similarly significant effects on you.

10. Your choices and rights
Subject to applicable conditions, you may:

• Request access to your personal data.
• Request correction of inaccurate personal data.
• Request deletion of your personal data.
• Request restriction of processing.
• Request portability of your data where this right applies.
• Object to processing based on legitimate interests.
• Withdraw consent for processing based on consent.

Contact [email protected] to exercise these rights.
For a voluntary contribution, please include your contribution code. We may request proportionate information to locate the relevant data or verify that you are entitled to make the request.
We cannot recover recordings that have already been deleted, and we do not retain additional identification data solely to identify otherwise unlinked recordings.
Withdrawing consent does not affect the lawfulness of processing carried out before withdrawal.
Local data and uninstalling the app
No account is required to use Parlify.
You can remove the app’s locally stored device identifier, contribution code and cached files by clearing the app’s data where your operating system permits this, or by uninstalling the app. Copies previously included in device cloud backups may need to be managed separately through your backup settings.
Uninstalling does not withdraw research consent or delete a contribution stored on our server. To request this, contact us using your contribution code. Keep a copy of the code before clearing the app’s data or uninstalling it.
Response times and complaints
We respond to rights requests without undue delay and normally within one month. For complex or numerous requests, this period may be extended by up to two further months; we will explain the extension within the first month.

11. Age requirements
Parlify is intended for people aged 18 and over, in line with our Google Play target audience. The app does not verify users’ ages.
Voluntary research contributions are limited to people aged 18 and over and require separate consent.
If you believe that someone under 15 has used Parlify, or that someone under 18 has submitted a research contribution, please contact [email protected] so that we can investigate and delete any related personal data we still hold.

12. Changes to this policy
We may update this Privacy Policy to reflect changes to Parlify or our data practices.
The effective date above identifies the current version. We will communicate material changes appropriately and obtain fresh consent where required before starting new consent-based processing.