End User License Agreement, Privacy Policy & Support

§1Licence grant

Subject to your compliance with this Agreement, we grant you a personal, non-exclusive, non-transferable, revocable licence to install and use the Software on devices you own or control, for your own purposes, including commercial ones.

Where you obtained the Software through the Microsoft Store, the number of installations and devices is governed by the Microsoft Store Terms of Sale in addition to this Agreement.

§2Restrictions

You may not:

• reverse engineer, decompile or disassemble the Software, except to the extent that applicable law expressly permits it despite this restriction;
• rent, lease, lend, sell, sublicense or distribute the Software, or make it available to third parties as a service;
• remove, obscure or alter any copyright, trademark or licence notice contained in the Software, including third-party model licences and attributions;
• circumvent, disable or interfere with any licensing, subscription or entitlement check;
• use the Software in violation of any applicable law, or in violation of §6 below.

§3Ownership

The Software is licensed, not sold. We and our licensors retain all right, title and interest in the Software, including all intellectual property rights. This Agreement grants you no rights to our trademarks or trade names.

The Software incorporates third-party components — including open-source libraries and machine-learning model weights — which remain the property of their respective owners and are licensed to you on their own terms. Those terms are reproduced in the Software under Settings → About App → Third-party Licenses, and where they conflict with this Agreement in respect of the component they cover, they prevail.

§4Your content and your output

You retain all rights to the video, audio, text and other material you supply to the Software ("Input"), and to the transcripts, translations, synthesised speech and rendered video the Software produces from it ("Output"). We claim no ownership of either, and we do not receive either — see Part II.

You are responsible for holding the rights necessary to process your Input, including any rights in a speaker's voice or likeness, and for how you use the Output. See §6 for restrictions that apply to both.

§5Third-party AI services and your own API keys

The Software can route transcription, translation and speech synthesis either on your device or through third-party AI services (currently OpenAI, Google, xAI, Soniox, ElevenLabs, Inworld, and any OpenAI-compatible endpoint you configure). Which of these runs — if any — is your choice.

Where you choose a third-party service:

• you supply your own API key and contract directly with that provider;
• your use of that service is governed by that provider's terms and privacy policy, not by this Agreement;
• your Input is transmitted to that provider and processed under their terms, including their retention and model-training policies, including by any subprocessors they use to provide the service. We are not a party to that relationship and make no promises on the provider's behalf;
• you are responsible for all charges the provider bills to your account.

We are not liable for the availability, accuracy, pricing, or conduct of any third-party service, nor for any loss of Input or Output caused by one.

§6AI models and use restrictions

The Software incorporates machine-learning models supplied by third parties. Some of these models are licensed to us on terms that impose use-based restrictions which we are contractually required to pass on to you, and which you accept by using the Software. You agree not to use the Software, any model it incorporates or downloads, or any output produced by them:

1. in any way that violates any applicable national, federal, state, local or international law or regulation;
2. for the purpose of exploiting, harming, or attempting to exploit or harm minors in any way;
3. to generate or disseminate verifiably false information or content with the purpose of harming others;
4. to generate or disseminate personal identifiable information that can be used to harm an individual;
5. to generate or disseminate information or content, or to place such information or content in any context, without expressly and intelligibly disclaiming that it is machine generated;
6. to defame, disparage, or otherwise harass others;
7. to impersonate or attempt to impersonate any person, including by synthesising that person's voice or likeness, without their consent;
8. for fully automated decision making that adversely impacts an individual's legal rights or otherwise creates or modifies a binding, enforceable obligation;
9. for any use intended to, or which has the effect of, discriminating against or harming individuals or groups based on online or offline social behaviour or known or predicted personal or personality characteristics;
10. to exploit any vulnerability of a specific group of persons based on their age or social, physical or mental characteristics, in order to materially distort the behaviour of a person belonging to that group in a manner that causes or is likely to cause that person or another person physical or psychological harm;
11. for any use intended to, or which has the effect of, discriminating against individuals or groups based on legally protected characteristics or categories;
12. to provide medical advice or interpretation of medical results; or
13. to generate or disseminate information for use in the administration of justice, law enforcement, immigration or asylum processes, including predicting that an individual will commit a crime or fraud.

You are accountable for the output you generate with the Software and for its subsequent use. These restrictions apply in addition to, and do not replace, any other term of this agreement; they bind any person to whom you make the Software or its output available, and they survive termination of this agreement. The full text of the applicable model licences is available in the Software under Settings → About App → Third-party Licenses.

§7Downloaded models

Some features require downloading model weights from third-party hosts (currently GitHub and Hugging Face). These downloads can be large. They are stored on your device, they are governed by their own licences as described in §3, and their availability depends on those hosts rather than on us. Downloading a model discloses your IP address to the host, as any download does; see Part II §6. The licence texts reproduced in the Software are those published by the model authors as of the date of the Software release you are running; where an author later changes the terms for new downloads, the updated terms apply from the catalogue update that adopts them.

On macOS 26 and later, the Software can also use Apple's on-device speech recognition. When a language's assets are missing, the Software asks for your approval and then requests that macOS download and install them from Apple. They are system components: macOS manages their storage, updates and removal and shares them with other apps, and Apple's applicable terms govern them rather than the model licences in §3.

On macOS 26 and later, where Apple Intelligence is available and turned on, the Software can also use Apple's on-device language model (Apple Foundation Models) for translation and for short answers in the in-app Help panel, grounded in help passages bundled with the Software. It is likewise a system component. The Software does not request or start any download of it: macOS manages it. If it is not ready, translation reports it as unavailable and Help continues with local documentation search. On Windows, Help uses local documentation search only.

On macOS 26 and later, the Software also offers Apple Translation as a translation model. It runs on your Mac using language assets managed by macOS. The Software cannot download missing translation languages: install the source and target languages in System Settings → General → Language & Region → Translation Languages before using this model. Apple's applicable terms govern these system components, and macOS manages their storage, updates and removal.

On macOS 15 and later, the Software can also synthesize speech on your Mac using Apple system voices already installed on the device. The Software does not request or start downloads of these voices; macOS manages their installation, updates and removal. Personal Voice and third-party speech-synthesis extensions are not used by this option.

§8Subscriptions and payment

The free tier exports the original full-length video with its complete source or background audio, but adds dubbed voice only for the first 10 subtitle segments. An active subscription removes that segment allowance and adds dubbed voice for every available segment. That is the only feature difference between the free and paid tiers.

On macOS the subscription is offered as a weekly or a monthly auto-renewable plan with the same benefit; on Windows as a monthly plan. The price and renewal period are shown before you confirm a purchase, in your local currency, and may differ by country. There is no free trial: the free tier above is available without one. A subscription renews automatically until you cancel it in the account of the store you bought it from.

All purchases, renewals, cancellations and refunds are handled by the store you obtained the Software from — by Microsoft under the Microsoft Store Terms of Sale, or by Apple under the Apple Media Services Terms and Conditions. We do not process payments and never receive your payment details. Manage or cancel a subscription through the account of the corresponding store.

A subscription is bound to the store where you bought it. We operate no user accounts, so there is no mechanism by which a purchase made in one store can grant access in another: a Microsoft Store subscription does not unlock the macOS build, and an App Store subscription does not unlock the Windows build. If you use the Software on both platforms, each requires its own subscription.

Subscription entitlement is verified against that store. If verification is unavailable, premium features may be temporarily unavailable.

Costs you incur with third-party AI providers under §5 are separate from any subscription and are billed to you by those providers.

§9Updates

We may release updates that add, change or remove features, including support for particular models or providers. A model or provider that is available today may become unavailable — for example if its licence terms change or a third-party host withdraws it.

§10Disclaimer of warranties

In particular, and without limiting the above: machine transcription, machine translation and synthesised speech are probabilistic and frequently imperfect. We do not warrant that any Output is accurate, complete, faithful to the Input, or suitable for any particular use. Do not rely on Output in any context where an error could cause harm — including medical, legal, financial, safety-critical, or official use — without competent human review.

Nothing in this Agreement excludes or limits any warranty or right that cannot be excluded or limited under the law that applies to you as a consumer.

§11Limitation of liability

These limits do not apply to liability that cannot be limited under applicable law, including liability for death or personal injury caused by negligence, or for fraud.

§12Indemnity

You agree to indemnify and hold us harmless against any claim, demand, loss or expense arising from your Input, your Output, your use of third-party AI services, or your breach of this Agreement — in particular of §6.

§13Term and termination

This Agreement takes effect when you first install or use the Software and continues until terminated.

It terminates automatically if you breach any of its terms. You may terminate it at any time by uninstalling the Software and ceasing to use it. On termination you must stop using the Software; §3, §4, §6, §10, §11 and §12 survive.

Termination of this Agreement does not by itself cancel a store subscription — cancel that through your Microsoft or Apple account.

§14Export compliance

You may not use or export the Software in violation of any applicable export control or sanctions laws, and you represent that you are not located in, or a national or resident of, a jurisdiction subject to such restrictions where they would prohibit your use.

In particular, you represent that you are not located in a country that is subject to a U.S. Government embargo, or that has been designated by the U.S. Government as a "terrorist supporting" country, and that you are not listed on any U.S. Government list of prohibited or restricted parties.

§15Changes to this Agreement

We may update this Agreement to reflect changes to the Software or to legal requirements. The current version is always the one published at https://aivideotranslator.app/legal.html, with its "Last Updated" date; where a store listing links to these terms, that link points here. Continuing to use the Software after an update takes effect means you accept the revised Agreement. If you do not accept it, stop using the Software.

§16General

If any provision of this Agreement is held unenforceable, the remainder stays in force. Our failure to enforce a provision is not a waiver of it. This Agreement, together with the Privacy Policy below and the third-party terms it references, is the entire agreement between you and us regarding the Software.

This Agreement is governed by the laws of the Province of Alberta and the federal laws of Canada applicable therein, without regard to conflict-of-law rules, and without depriving you of the protection of mandatory consumer-law provisions of your country of residence.

§17Additional terms for the Apple App Store

This section applies only where you obtained the Software from the Apple App Store, and prevails over anything else in this Agreement to the extent of a conflict.

• This Agreement is with us, not Apple. It is concluded between you and us alone. Apple is not a party to it, and we — not Apple — are solely responsible for the Software and its content.
• Scope of the licence. The licence granted in §1 is limited to a non-transferable licence to use the Software on any Apple-branded products that you own or control, and as permitted by the Usage Rules set out in the Apple Media Services Terms and Conditions.
• Maintenance and support. We are solely responsible for providing any maintenance and support for the Software. Apple has no obligation whatsoever to furnish any maintenance or support services.
• Warranty. We are solely responsible for any product warranties, whether express or implied by law, to the extent not effectively disclaimed. If the Software fails to conform to any applicable warranty, you may notify Apple, and Apple will refund the purchase price of the Software to you; to the maximum extent permitted by applicable law, Apple has no other warranty obligation whatsoever with respect to the Software.
• Product claims. We, not Apple, are responsible for addressing any claim by you or a third party relating to the Software or your possession and use of it, including product liability claims, any claim that the Software fails to conform to an applicable legal or regulatory requirement, and claims arising under consumer protection, privacy or similar legislation.
• Intellectual property claims. In the event of a third-party claim that the Software or your possession and use of it infringes that third party's intellectual property rights, we, not Apple, are solely responsible for the investigation, defence, settlement and discharge of that claim.
• Legal compliance. You make the representations set out in §14.
• Third-party terms. You must comply with any applicable third-party terms of agreement when using the Software — including the provider terms referenced in §5 and the model licences referenced in §6 and §7.
• Third-party beneficiary. Apple and Apple's subsidiaries are third-party beneficiaries of this Agreement, and upon your acceptance of it Apple will have the right — and is deemed to have accepted the right — to enforce this Agreement against you as a third-party beneficiary of it.
• Contact. Questions, complaints and claims regarding the Software should be directed to us at ai.video.translators@gmail.com.

§1The short version

For the purposes of this policy, the controller of personal data is Oleh Tieliushchenko, Ukraine. Contact for privacy requests and questions about this policy: ai.video.translators@gmail.com.

AI Video Translator is a desktop application that processes your media on your own computer. We operate no remotely hosted media-processing servers, we have no user accounts, and your video, audio, transcripts, translations and generated speech are never sent to us.

These are the things that can be sent outside the Software, and each is described in full below:

We do not sell, rent or trade any data to anyone, and we run no advertising.

§2What stays on your device

The following stays on your computer except for the cloud processing you select (§3), automation you request (below), or files you deliberately send or share:

• your source video and audio files;
• projects, transcripts, translations, subtitles and generated speech, including all cached TTS audio;
• downloaded model weights;
• your API keys, which are transmitted only to the provider they belong to. On macOS they are stored in the Keychain. On Windows they are stored in a file in the Software's application-data folder, encrypted with Windows Data Protection (DPAPI) so that only your Windows user account can decrypt it;
• your application settings.

The Software can run its entire pipeline fully offline, using on-device models for speech recognition, translation, speech synthesis and source separation. Used that way, the Software does not send media or translation text to an AI provider; automation requests and returned information are separately handled as described below. This includes Apple's on-device speech recognition on macOS, which processes your audio on your Mac and works without an internet connection while the language's assets are installed. macOS downloads missing assets from Apple after you approve it in the Software (Part I §7) and manages later updates itself; the Software does not upload your audio or transcripts to Apple for this recognition. The same applies to translation with Apple's on-device language model (Apple Foundation Models): the text runs through the model on your Mac, it works without an internet connection while macOS has the model ready, and the Software does not use Apple's server-side options for it, such as Private Cloud Compute. Apple Translation and speech synthesis with installed Apple system voices likewise run on your Mac without an internet connection while the required language assets or voices are installed. The Software does not send the text or generated audio to Apple for them, and does not download translation languages or voices; missing translation languages are reported with instructions for installing them in macOS System Settings.

The in-app Help panel searches English documentation bundled with the Software. On supported Macs, after you submit a question, Apple Foundation Models can generate a short answer using only selected help passages. Your question and those passages are processed on your Mac; the Software does not send them to us, Apple, another AI provider or Private Cloud Compute. Help questions, generated answers and passage excerpts are not recorded in application logs or included in crash reports or usage statistics, and the Help panel adds no analytics events. If the model is unavailable, local search still works. On Windows, Help uses local search only. Opening a link in a help passage opens its published destination in your browser; the Software does not add your question to that link.

Automation through Siri, Shortcuts and MCP. MCP is off by default and starts only when you enable it in Settings → MCP. Turning it off closes host connections and removes the endpoint; it does not cancel an already accepted app-owned job. On macOS, Shortcuts can request video translation, and Siri and Shortcuts can search bundled Help. A local MCP server lets an assistant host request Help, provider/language/model/voice information, translation jobs, status and cancellation on macOS and Windows. The app owns the translation job. Siri receives matching help passages when you search Help. Depending on the operation, Shortcuts or an MCP host receives matching help passages, job identifiers, status and progress, coded errors, warnings (including free-tier limits), source and destination file paths, and the location of the result file. MCP export replies return the result path, media type and size without embedding video bytes. The get_frame tool separately returns a source-video frame as JPEG image bytes to the requesting host, with dimensions and the requested time. A host or shortcut with its own file access can subsequently read or share a result file under its own permissions.

MCP hosts can list and inspect any registered project accessible to the app, including projects you have not opened. The list_projects, get_project and list_segments tools return project content and verified take-duration information without selecting or editing a project. The get_frame tool inspects an accessible project's source video and returns an image without changing the project or its undo history. Frame bytes are not recorded in application logs, crash reports or usage statistics. The edit_segments, edit_project, even_out_acceleration and project_history tools can immediately apply free transcript, timing, settings and speaker edits and undo or redo project changes without opening the project; edit responses include operation receipts and created row IDs. Each changed batch has one shared undo entry, including partial effects. History and accepted edit-request identities last only for the running app. Free edits never automatically translate, voice or start paid work. Enabling MCP grants this read and edit access to connected local hosts; there is no separate permission per project or per host. Turn MCP off to close those connections.

MCP hosts can explicitly request translation or voice generation for selected segments, retranscription of an existing project, or export of current voiced segments through retranslate_segments, revoice_segments, retranscribe_project and export_project. These calls start without a GUI confirmation dialog, may incur charges from cloud providers, and their published results are not undoable. Successful retranscription clears project history. Export cannot be undone and does not repair missing translations or voices. Jobs report discarded stale results and possible incurred cost when concurrent human edits win; already published partial results remain after failure or cancellation. Accepted request identities survive restart, unfinished jobs become interrupted, and paid processing never automatically resumes. Assistants should start paid operations only when you explicitly ask, and must not automatically repeat paid work after an error.

Provider API keys, endpoints and provider configuration maps never cross either automation channel: they remain in the Software and credentials are supplied only to the selected provider. A path supplied by an assistant is not a file-access grant; where permission is missing, you must select the file or folder in the app. Automation still follows external-AI consent, which is checked again for each provider request and never granted by an MCP call, and the same free-tier export rules as the app; it adds no separate premium requirement. The normal export allowance produces a free-tier warning; an explicitly required full dub refuses an over-allowance export. That consent does not control an MCP host's processing of returned project content. Local execution of the translation pipeline does not mean the assistant's request or the information returned to it stays on your computer. Apple handles Siri and Shortcuts under Apple's terms; an MCP host and its assistant process requests and responses under their own terms and privacy policies, which may include cloud processing. We receive neither those automation requests nor their results through these channels.

We have no ability to access, recover or delete any of this. Deleting a project or clearing the cache removes it from your machine and nowhere else.

Uninstalling removes the Software but not your data. Your projects live in the projects folder you choose, and downloaded models in the models folder shown in the Software's settings (on macOS, by default inside the Software's Application Support folder); uninstalling does not touch either, so if you want everything gone, delete them by hand. Apple's speech-recognition assets, Apple Translation language assets, its on-device language model and installed system voices on macOS are stored and managed separately by macOS: deleting those folders does not remove them, and macOS controls their retention and removal. Whether the install identifier described in §4 survives a reinstall depends on whether uninstalling also removes the Software's stored settings; §4 describes both outcomes.

§3Third-party AI providers (only if you choose them)

If you configure a cloud provider, the relevant Input — audio for transcription, text for translation, text for speech synthesis — is sent directly from your computer to that provider, authenticated with the API key you supplied. It does not pass through any system of ours.

No Input is sent to a cloud provider without your consent. The Software asks for it before you can enter an API key: new users see the request during onboarding, existing users once after upgrading. You may continue without consenting and grant it later in Settings → About App. You can withdraw consent at any time by unchecking the external AI services checkbox in Settings → About App. Without consent, API keys for cloud providers cannot be entered, keys already stored stay in the encrypted local storage described in §2 without being supplied to any provider, and every cloud step is refused before a request is made. Withdrawal stops an already-running cloud batch before its next request; requests already sent may complete, and withdrawal does not undo processing already performed by a provider. On-device processing, Apple's system providers on macOS (on-device speech recognition, Apple Foundation Models, Apple Translation and installed Apple system voices), and an OpenAI-compatible endpoint at localhost, 127.0.0.1 or ::1 do not need consent, because the data stays on your computer; an endpoint elsewhere on your network counts as external.

Connection checks on Windows. Each time the Software starts on Windows, it uses the curl program built into Windows to send one HTTPS HEAD request to each of these hosts: github.com and huggingface.co (each for one published model file, following the redirect to that host's download server), and api.openai.com, api.x.ai, generativelanguage.googleapis.com, api.elevenlabs.io, api.soniox.com, tts-rt.soniox.com and api.inworld.ai. api.mixpanel.com is added only if you have allowed usage statistics (§5), and the Sentry host only if crash reporting is on (§4). These requests let Windows verify, and if needed update, its trusted root certificates before the Software opens its own secure connections; Windows may download missing root certificates from Microsoft while doing so. They are made whether or not you have consented to external AI services and whether or not you use those providers. They carry no Input, API key, file name or other personal data, only curl's standard request headers (such as its version in User-Agent); no response body is downloaded, and the Software keeps only the host name and whether the check succeeded in its local log. Like any connection, each request reveals your IP address to the host. The macOS build makes no such requests.

The providers the Software can be configured to use are: OpenAI, Google, xAI, Soniox, ElevenLabs, Inworld, and any OpenAI-compatible endpoint you enter yourself (a hosted aggregator, or a local server such as Ollama or LM Studio, in which case the data stays on your own machine or network).

Because you contract with these providers directly using your own account:

• their privacy policy, retention period and model-training policy govern that data, not ours;
• we cannot see what you sent, and we cannot delete it on your behalf;
• to exercise data rights over it, contact that provider.

Please read the privacy policy of any provider before enabling it.

§4Diagnostics and crash reports (Sentry)

The Software sends crash reports, error diagnostics, performance traces and performance profiles to Sentry so we can find and fix defects. This is enabled in released builds, and you can switch it off with Crash reports in Settings → About App. The setting is read when the Software starts, so turning it off takes effect from the next launch, and nothing is sent from that point on.

What is sent: the exception and stack trace, application version, operating system and device characteristics, timing measurements, and the application's own log breadcrumbs leading up to an event.

What is deliberately removed before sending. Sentry's "default PII" collection is switched off, and every outbound event additionally passes through a scrubber that redacts, from all fields including log text and stack traces:

• API keys, bearer tokens, private keys and credentials in URLs;
• email addresses (except a name and email you type into the feedback form yourself — see below);
• IP addresses;
• the user-name component of file paths (C:\Users\<name>\…, /Users/<name>/…, /home/<name>/…).

Identifier. Reports carry a random identifier ("install ID") generated on your device on first launch. It is a random UUID — it is not derived from and does not correlate with your name, email address, account, machine name or any hardware identifier. Its only purpose is to let us tell "one user hit this crash 40 times" apart from "40 users hit it once". A feedback report you send (see below) carries neither it nor the trace identifier your crash reports share, so what you write — and a name or email you add — is not tied to your installation's crash reports by any identifier the Software attaches. (Anything you type or attach yourself is sent as you wrote it.) A reinstall that clears the Software's stored preferences generates a new one; a reinstall that leaves them in place keeps the old one.

Feedback you choose to send

Everything above is automatic and deliberately impersonal. The feedback form is the opposite, and it is worth reading before you use it.

When you open Send feedback and submit it, the Software sends to Sentry: the message you wrote, a name and an email address if you choose to fill them in (both are optional and blank by default), any files you attached, and — if you leave the box ticked — the last 500 entries of the application log.

The form will not send until you tick a confirmation describing what the report contains; the tick is remembered, visible on every submission, and can be removed. The form uses the same reporting channel as crash reports; if you have switched Crash reports off, submitting feedback sends only your report — that switch stays off and nothing else is sent.

That log is anonymized before it is attached: file paths, user names, email addresses, IP addresses, long keys and tokens, and store and transaction identifiers are replaced. It travels as an attachment rather than as event fields, so it is redacted by the logging component's own rules and not by the outbound scrubber described above — the two overlap but are not identical, and a credential in an unusual format can survive. If the log might contain something you would rather not send, untick the box.

Two further consequences follow, and neither is covered by any scrubbing:

• The fields are yours to fill. Nothing forces you to identify yourself, and nothing is pre-filled from your system. If you type a name and an email, you have chosen to give them to us, and they arrive attached to the report.
• Attachments are sent as you provided them. The scrubber described above rewrites event fields; it does not open your files. A screenshot, a project file or a log you attach yourself goes to Sentry unmodified, so please check what is in it — and never attach an API key.

Feedback reports are readable by us as support. To have one deleted, write to the address in §10.

§5Product analytics (Mixpanel)

Product-usage events are sent to Mixpanel only if you choose "Share usage statistics" during onboarding or later enable Usage statistics in Settings → About App. Collection is off until you make that choice. We use these events to see which features are used and where the pipeline fails.

What is sent: the name of the action or screen (for example project_created, demo_download_failed, Settings), and accompanying properties describing the context of that action — application version, operating system and version, interface language, locale, country code, time-zone offset, which onboarding step or control was used, the provider, voice and source/target languages selected for an operation, subscription status, and measurements of the operation such as durations and downloaded byte counts.

For the translation pipeline, lifecycle events record when transcription, translation, speech synthesis and final audio overlay start, and whether each stage completes or fails. Their properties are limited to selected provider and model identifiers where applicable, source or target language codes, segment counts, duration, and a fixed machine-readable failure code. For the onboarding funnel, events record guided-demo model-download start and outcome (total/completed byte counts and percentage), the guided-demo transcription start and outcome (provider, model, language, duration and failure code), display of an FFmpeg availability warning and its fixed reason, display of a missing-API-key warning during first real-project setup (provider and fixed reason), and a click on the privacy-policy link. When the free export allowance actually omits segments, an event records the original segment count.

Player and interface events record source-video play or pause, starting or stopping sequential or continuous accelerated playback, a settled timeline seek, subtitle-segment auditions, subtitle-filter changes, and opening Settings. Their event-specific properties are limited to the resulting playing or enabled state, an enum-backed filter identifier, and the number of segment auditions coalesced into a short window. Settled seeks and Settings openings carry no event-specific properties. Timeline seeks are sent only after a one-second quiet period, and segment auditions are grouped into at most one counted event per five-second window.

A failure is reported by the event name — demo_download_failed, purchase_failed — and at most by a short reason code chosen from a fixed list written into the Software, such as sentry_disabled. The text of the underlying exception is deliberately not included: that text is not scrubbed on this path and can contain a file path or a URL, so it goes only to the crash reporting described in §4, where the scrubber does apply. For the same reason the name you give a speaker in a project is not sent either; only whether a segment's voice assignment was set or cleared.

One of those failures concerns this policy directly. If you send feedback and it does not reach us, the Software records that the attempt failed and why — the reason code only, never your message, your name, your email address or your attachments. It does so because the feedback form is our only support channel, and every other way it could tell us it is broken travels through the very service that just failed. If you have switched Usage statistics off, this is not sent either.

What is not sent: your media, file names, file paths, transcripts, translations, generated audio, API keys, exception messages, your name, or your email address.

Identifier. Analytics events carry a random install ID generated by the same mechanism described in §4: a UUIDv4 generated locally by the Software and stored on your device. It is not derived from and does not correlate with your name, email address, account, machine name or any hardware identifier, and it survives a reinstall exactly as long as the stored preferences do (see §4). If local preferences are temporarily unavailable, the Software uses a fresh random UUID for that session instead; it never substitutes an account- or device-derived identifier or a placeholder shared with other installations.

This makes analytics data pseudonymous rather than anonymous: it is not linked to your identity, but events from one installation can be grouped together.

Mixpanel processes this data as our service provider, under its own privacy policy. You can stop collection at any time by switching off Usage statistics in Settings → About App. Sending stops as soon as you switch it off and events still queued on your device are discarded rather than delivered; the change takes full effect from the next launch, when the setting is read again at startup. Switching collection off does not delete events Mixpanel has already received; §10 explains how to request that.

§6Model downloads

When you download a model, the request goes to the host that publishes it — currently GitHub or Hugging Face. As with any download, the host sees your IP address and the file requested. We receive nothing. These hosts are not our service providers; they act as independent controllers under their own privacy policies.

On macOS, after you approve it in the Software, macOS downloads missing assets for Apple's on-device speech recognition from Apple. The Software supplies no audio, transcript or other Input in that installation request; Apple sees it as it sees any system download and handles it under its own privacy policy. We receive nothing through it.

Apple's on-device language model used for translation and grounded in-app Help answers on macOS is provided and updated by macOS as part of Apple Intelligence. The Software does not download it and sends no translation text, Help questions or bundled help passages to Apple through it. Help does not initiate model downloads; if the model is unavailable, local Help search remains available.

Apple Translation language assets are supplied and managed by Apple through macOS. You install missing languages in System Settings → General → Language & Region → Translation Languages. The Software does not initiate those downloads. As with any system download, Apple receives your IP address and handles the request under its own privacy policy; the Software sends no transcript or translation text in a language installation request.

Apple system voices used for speech synthesis are supplied and managed by macOS. The Software uses only voices already installed on your Mac, does not request voice downloads, and sends no text or generated audio to Apple through this synthesis.

§7Subscriptions

Subscription entitlement is verified against the store the Software came from — the Microsoft Store on Windows, the App Store on macOS. The Software asks that store whether the signed-in account holds a valid licence and receives a yes/no answer with its validity period. Purchases and payment are handled entirely by Microsoft or Apple under their own privacy statements; we never receive your name, billing address or payment details.

§8What we never collect

We do not operate accounts, and nothing in the Software requires your name, email address, postal address or phone number — the only place we accept a name or an email is the feedback form in §4, where both fields are optional and blank unless you fill them in. We do not use advertising identifiers, tracking cookies, cross-site tracking, or any analytics beyond the two services named in §4 and §5. We do not read your files, and we do not perform automated decision-making or profiling that produces legal or similarly significant effects.

§9Retention and international transfers

We hold no database of users. Retention of the limited data described above is determined by each processor: Sentry and Mixpanel retain event data according to their own retention schedules.

Each of Sentry and Mixpanel processes this data solely as our service provider, under contractual commitments — Sentry's Data Processing Addendum and Mixpanel's Data Processing Addendum — that restrict use of the data to providing the service to us and require protection no less protective than this policy.

These service providers may process data on servers outside your country, including in the United States, under the safeguards set out in their respective privacy policies.

§10Your rights

Depending on where you live, you may have rights to access, correct, delete or restrict processing of your personal data, to object to processing, or to lodge a complaint with a supervisory authority.

Because we hold no account and no identifiable profile, in most cases we hold nothing that identifies you. Two exceptions are worth knowing:

• a feedback report you sent may contain whatever you wrote, attached, or filled into the optional name and email fields;
• diagnostic and analytics events are tied to the identifiers described in §4 and §5. You can stop future analytics collection with Usage statistics in Settings → About App, and you can stop Sentry from initializing with Crash reports in the same place; both switches take full effect from the next app start. These switches do not delete events already received. If you want existing events located or deleted, write to us by email and we will help you retrieve the relevant identifier and act on it — without it nothing links those events to you, which is the point of the design but also its limitation.

To make a request, or to ask anything about this policy, write to ai.video.translators@gmail.com.

§11Children

AI Video Translator is not directed at children under 13, and we do not knowingly collect personal information from them.

§12Changes to this policy

We will update this policy when the Software's data handling changes, and revise the "Last Updated" date at the top. Material changes will be reflected in the app listing where you obtained the Software.

• Provided by: Oleh Tieliushchenko, Ukraine
• Support email: ai.video.translators@gmail.com
• In-app: press the feedback button in the title bar and send us a report from there (see Part II §4) — it reaches the same people, and it can carry the log and any files you attach.

For any concern, question or dispute, contacting support is the fastest route to a resolution. When reporting a problem, including the application version from Settings → About App and the approximate time it happened helps us find the matching diagnostics.

Third-party component and model licences are reproduced in full in the Software under Settings → About App → Third-party Licenses.

Parts I–III describe the desktop Software. This website is separate and processes data of its own: it uses cookies and analytics, which you control through the cookie banner, and the feedback form on this site sends the email address and message you enter to us. Nothing on this website has access to anything inside the Software, and nothing inside the Software reports to this website.

In detail, and only with your consent:

• Cookies and consent. On your first visit a banner lets you accept or reject analytics and marketing. Nothing non-essential runs until you choose. Your choice is stored locally in your browser (localStorage, key avt_consent), not in a cookie, and you can change it at any time through Cookie settings in the footer.
• Analytics (Mixpanel). If you allow analytics, we use Mixpanel for aggregate page views and basic interaction events. It loads only after you consent. IP addresses are not collected, it stores its data in localStorage rather than cookies, and it honours your browser's "Do Not Track" setting.
• Tag manager and Google Consent Mode v2. We use Google Tag Manager to manage measurement tags. Under Consent Mode v2 all storage signals default to denied, so Google tags stay off until you grant the matching consent (analytics or marketing).
• Feedback form (Web3Forms). If you send feedback from this site, your email address (required), name (optional) and message are submitted to Web3Forms, which delivers them to us by email, solely so we can reply. We do not use them for marketing.
• No data is sold, and no advertising runs without your consent.

Feedback messages sent from this site are kept only as long as needed to handle your request and improve the product; write to ai.video.translators@gmail.com to have yours deleted. This website is not directed to children under 13, and we do not knowingly collect their personal data.