Privacy Policy
What we collect, why, who else is involved, and how to get your data deleted.
Effective date: 14 November 2026
The short version
- Your clips are rendered on your device, in your browser or through Modisign MCP. No video or photo is uploaded to us or stored by us.
- You can browse and download the free overlays without an account.
- We use no advertising, no third-party analytics and no tracking cookies. We count anonymous usage on our own server.
- If you sign in, we keep your Google name, email and profile picture, the overlays you like and bookmark, the Studio guides you chose not to be shown automatically, and, for Pro, a list of the overlay clips you finished (Studio downloads are not recorded), plus any Studio project structure you choose to save with Pro.
- We never see or store payment card details.
Who is responsible
The controller of your personal data is Viacheslav Sierbov, an individual based in Ukraine, who operates Modisign at modisign.app. Contact: support@modisign.app.
What we collect and why
When you visit. Like any website, the servers that deliver Modisign receive your IP address, browser type and the pages you request. Our hosting providers process this to deliver the site and protect it from abuse. We do not build visitor profiles from it. Legal basis: our legitimate interest in running a working, secure website.
Usage statistics. To see what is used and what breaks, we record anonymous events on our own server in the European Union: pages viewed, overlays opened, downloads started, finished, failed or canceled, MCP login outcomes and MCP render outcomes (free status and whether it is an overlay, composite or Studio edit; the overlay’s catalog id for overlays and composites), MCP installation instructions copied (which assistant’s setup was selected), views of and clicks on Pro, checkouts opened, closed or completed (only where the checkout was opened from, phone or desktop layout, and for a closed checkout how far it got, from a fixed list such as payment method chosen; never your name, email, address, payment method, price, transaction or what you were editing), share buttons used, caption generation finished, canceled or failed (including whether a finished generation found speech), Studio imports by type and length range (and whether it was the first import in the editor and whether it went to the timeline, the media bin or a missing clip), overlays placed or edited by catalog item, whether a placed overlay was started again from its default or from another starting example (and whether the result is a free overlay; never which example, its text or its colors), which editing tools were used (with broad pause-removal ranges, transition type, caption style, or each output format chosen), project saves, opens and deletes, button clicks, which page sections were viewed, FAQ items opened, the catalog view you settle on (the collection page you are on and the purpose, platform, “Examples for” niche, sort and price chosen, each from a fixed list), and the search words typed into the overlay catalog and the Studio overlay library with whether the search found results. Search words are recorded once you stop typing, in lower case and cut to 60 characters; a search that looks like an email address, a link or a number of five or more digits, or that uses symbols other than common punctuation, is recorded only as a search, without its words. In Studio, we also record when the sign-in gate or signed-in editor is ready, whether it opened after sign-in or on returning to the tab, whether the phone or desktop layout is shown, attempts to start importing and their broad outcomes, which tool panels are opened, for the built-in Studio guides which guide from a fixed list (such as the app tour, a tool tour or the guided first video), which of its fixed steps and what happened from a fixed list (opened, a step shown or completed, an action requested, skipped, closed, resumed, completed, or chosen not to be shown again), whether the Export button is shown or used, export options and fixed reasons for a wait or Pro requirement, and explicit requests to download a ready file. Caption records include starting and reaching the model, audio and transcription stages. When you leave the editor (Exit, another page or closing the page), we may record once whether the current edit has an unexported change and which broad step was active; switching tabs or apps records nothing, and leaving does not mean abandoning an edit. These events do not identify your account, connect separate page loads, contain media or typed text, or confirm that your device saved a file. We also record failures when importing media, previewing an edit, analysing pauses, generating captions or people masks, loading sounds, saving or opening projects, signing in, using account controls, opening checkout, or loading a page. When the Studio preview waits at least 1.5 seconds or stops, we record whether it recovered or failed, a broad wait range (for example 3 to 10 seconds), whether the wait was at a clip change, a seek or the start of playback, the broad video codec family (such as H.264 or HEVC) and the phone or desktop layout, never exact media timing, contents or file names. Apart from those search words, we never record file names, typed text, or project names, identifiers or content in usage statistics. Each event holds the page address without your settings or text, the overlay’s catalog id (never which example of it you opened or started from), look, format and size involved, whether it is a free overlay and whether a download uses your own video (never its name or contents), whether it was made in Studio and the counts of clips, overlays, transitions, photos and active punch-ins, and caption style (never file data), caption generation backend (WebGPU or WASM) and source duration bucket (never audio or caption text), how long a download took, the feature and processing stage that failed (for example caption placement: fonts, overlays or plan), a fixed diagnostic code, the built-in kind of error from a fixed list (for example TypeError or DOMException) for a failed download and for some unexpected errors, the website build identifier and, when available, the HTTP status, a server error reference or a first-party compiled-code position (we derive broad browser and operating-system families from the user-agent string, and never record raw error messages, stack traces, script addresses, local paths, tokens, media details other than that video codec family, file names or text you typed in these error records), which share option was chosen, your browser’s user-agent string, whether the device is a touch device, whether the visitor was signed in or on Pro (not who), and on your first page the site you came from and any campaign tags in the link. It uses no cookies and no device storage: a random identifier ties one visit’s events together and is gone when the page is reloaded, and a second random identifier ties a download’s or checkout’s start to its outcome so a repeated report is counted once (it holds nothing about the download, the checkout or your payment). Events are not linked to accounts, and IP addresses are not stored with them. We use totals of these events across all visitors, never a single visit, to order the catalog by popularity. To opt out, turn on Global Privacy Control in your browser and no anonymous usage or browser diagnostic events are recorded. Legal basis: our legitimate interest in understanding how the site is used and fixing what fails.
Service diagnostics. To diagnose failures in our API and website servers, we keep a technical error record for each server error (an HTTP 5xx answer, or a website page that failed while rendering) in our hosting logs and in the same database in the European Union. A record holds the server, the request method, a fixed route or operation type, the response status when the server knows it, a fixed diagnostic code, the deployment identifier, how long the request took and, for the website, a server error reference. It holds no request body, query string, cookie, sign-in token, account identifier, IP address, user agent, media or text you typed. Global Privacy Control does not stop these records, because they describe our servers failing, not you. Legal basis: our legitimate interest in running and repairing the service.
When you sign in with Google. Google tells us your name, email address, profile picture address and your Google account identifier. We store these to create your account and show it to you. We also store the sign-in tokens Google issues, in encrypted form. We do not get access to your Gmail, Drive, contacts or anything else in your Google account. While the Google step runs, and for one hour after it returns, we keep the address of the page you started from, so a repeated or reopened sign-in link can take you back there; it is not linked to your account. Legal basis: performing our contract with you (the account you asked for).
While you are signed in on the website. Each website sign-in creates a session record holding your IP address, your browser’s user-agent string and an expiry time. Legal basis: performing our contract with you, and our legitimate interest in keeping accounts secure.
Likes. When you like an overlay while signed in, we store that your account liked it and when. Everyone sees how many likes an overlay has, never who gave them. Clicking the button again removes your like, and deleting your account deletes your likes. Likes count toward the catalog’s popularity order. Legal basis: performing our contract with you.
Bookmarks. When you bookmark an overlay while signed in, we store which catalog overlay it is and when. A bookmark is private: it is not counted, shown to anyone else or linked to likes, and it holds no text you typed or settings you changed. Removing the bookmark deletes it, and deleting your account deletes all your bookmarks. Legal basis: performing our contract with you.
Studio guides you turned off. When you choose not to be shown a Studio guide automatically, we store which guides, from a fixed list, with your account, so the choice follows you to other browsers. Nothing about it is kept on your device. You can still open any guide yourself, and deleting your account deletes the list. Legal basis: performing our contract with you.
Clip history (Pro). When a Pro account finishes an overlay download, we save one row: the clip’s title text, the overlay name, the address of the overlay with your settings (which includes the text and colors you chose) and the time. This is what lets you reopen and re-download a clip from your account. The video file itself is not saved. Free overlay downloads made without Pro, and all Studio downloads, are not recorded. Legal basis: performing our contract with you.
Your plan. We store whether your account has Pro. When you subscribe, we give our payment provider Paddle your email address and account identifier, and Paddle tells us the status of your subscription, when it is set to end, and its subscription and customer identifiers. When you open billing from your account, we ask Paddle for the email address of that customer to confirm it is yours. Legal basis: performing our contract with you.
When you email us. We keep your message and email address to answer you. Legal basis: our legitimate interest in handling requests, or performing our contract with you.
What we do not collect
- Your videos and photos. Rendering and encoding happen on your device. The finished file goes from your browser to your downloads, never to our servers.
- Payment card data. Payments are handled entirely by Paddle. Card numbers never reach us.
- Advertising or third-party analytics data. There are no ad pixels, third-party analytics scripts or fingerprinting on the site.
- Passwords. Sign-in is through Google only.
We do not sell personal data, and we do not use it for automated decisions about you.
Modisign MCP
Modisign MCP lets an assistant render overlays and edit local video sequences, photos and captions on your machine. Configurations, including the text you type into an overlay, are sent to Modisign to validate settings and check your plan. Video, photos, audio, transcripts and local file paths are never sent to us. Calling get_transcript returns transcript text to your assistant provider so you can review or correct it. Otherwise generated transcript text stays on your machine. Media never leaves your machine through Modisign MCP. Completed Pro overlay and single-video composite renders save configuration-only clip history under the same rules as website downloads. Studio edits never enter account history. Free renders need no account and leave no history.
Connecting MCP creates a token and a connected-device record on your account: the label “Modisign MCP” with macOS, Windows or Linux, its connection date, expiry and last account API use. We store no hostname, computer username or IP address in that device session. The token lasts up to one year. Revoke it immediately from Connected devices on your account page; deleting your account deletes its devices and tokens. Device codes expire after ten minutes; expired unredeemed codes are removed on a later login request after one day. Legal basis: performing our contract with you.
Anonymous MCP usage events use a random identifier held only for the running process, separate from your account and token. Login events contain only success, failure, denial or expiry. Render events contain only free status, overlay/composite/Studio kind, success or failure, and the overlay’s catalog id for overlays and composites. No media or overlay text enters these statistics. Downloading compositions also requests Google Fonts and jsDelivr resources as described below; cached files can be reused locally. English captions download the pinned Whisper tiny.en model from Hugging Face once into the local MCP cache. Model assets persist; media and transcripts do not. The built-in sound pack and the small person-detection model and runtime used by Behind person ship inside the package and are served to the render browser from the same loopback proxy; nothing about your sounds, title text, owners, timing or events is sent to Modisign; for the Pro check only one fixed yes/no fact is sent, whether any title uses a reveal or Behind people. Render browsers can request only named local media and cached assets through a loopback proxy.
Saved Studio projects
Studio requires sign-in and is free to use. With Pro, you can choose to save a project. We store its project name, output aspect, clip file names and kinds, video source durations and in and out trim points, photo durations and source offsets, footage track assignments and V2 timeline start times, source crop percentages and destination placement boxes, each clip’s punch-in source start times, lengths and zoom factors, frame mode, centre position and zoom, photo slow zoom, video volume, mute and audio fades, transition type and duration at each cut, overlay ids, start times, and each overlay’s configuration (including its text and settings), together with its owner, project id and creation and update dates. This lets you reopen its structure and relink local files. Media never leaves your browser: we never upload or store video, photos, audio, thumbnails or file contents. Projects keep only caption style, position and size (and the Mixed caption recipe: preset, seed and colours, the phrase grouping you chose, the Scatter recipe: seed, safe zone, intensity and the allowed styles, and a captions-off flag on a clip such as a Rapid intro teaser) when captions are present or you chose a caption look before generating them; caption text, the position, size, rotation and style you give a single caption line, where lines are split or joined, and individual word styles are never stored. Titles you add in Studio are saved with the project: their text, font, colour, size, placement, backing, optional outline colour and width, optional timestamp setting, and the clip they follow or their start time and length. A title’s reveal (rectangle, direction, timing and easing) and its Behind person setting (which clip it stands behind and three quality values) are saved with it; no mask, picture or detection result is ever stored. Motion keyframes you add to a clip, photo, title or overlay are saved with it: the time, value and easing of each key for position, scale, rotation and opacity (and the phase of a title cut from a longer one). Sounds you add from the built-in sound pack are saved as structure only: the sound’s id, whether it starts at a time or follows a title, transition or overlay moment (with its offset), its start and end within the sound, volume, mute, fades and whether automatic placement made it; no audio is stored. Legal basis: performing our contract with you to provide saved projects.
Deleting a project in Studio deletes its saved structure. Deleting your account deletes all its projects.
Cookies and browser storage
We set cookies only for sign-in, so no consent banner is needed. Visitors who never sign in get no cookies from us. We do not use local storage. The one exception to keeping Studio data only in tab memory is the speech model: choosing Generate captions downloads about 43 MB from Hugging Face and caches its files in this browser using the Cache API. Captions are generated on your device in English. Video, audio and caption text are never uploaded. The cache lasts until your browser evicts it or you choose Remove in the Captions panel. Behind person runs a small person-detection model (about 250 KB, plus about 10 MB of runtime) in a worker on your device when you first prepare or export such a title. Both are static files from our own domain that your browser may cache like any other page asset. The masks it produces exist only in tab memory: they are never uploaded, saved or written to browser storage, and they are discarded when you clear them or close the tab. It finds people in the picture; it does not identify anyone. The ten built-in sound files are ordinary static files from our own domain that your browser may cache like any other page asset; where you place them is never sent to us unless you save a Pro project. Closing the tab discards media and caption text. Generation, editing and SRT/VTT downloads are free for signed-in users; expressive caption styles need Pro to export a video.
The checkout is provided by Paddle. Nothing from Paddle loads until you open it by choosing Get Pro, on the pricing section or in the editor or Studio when an export or save needs Pro, or follow a payment link from Paddle. It then loads from Paddle’s domains, and Paddle and its payment partners set their own cookies and storage on their domains, under Paddle’s privacy policy. The same applies to Paddle’s billing pages that your account links to.
- better-auth.session_token: keeps you signed in. Lasts up to 7 days and is renewed while you keep using the site; removed when you sign out.
- better-auth.state: protects the Google sign-in step. Lasts 5 minutes.
On the live site these names carry a __Secure- prefix. Google may set its own cookies on its own pages while you sign in.
Who else receives data
These companies process data for us or receive requests from your browser when you use Modisign:
- Vercel hosts the website and receives every page request (IP address, browser details, requested pages).
- Hetzner hosts our API server and database in the European Union, where all account data, sessions, likes, bookmarks, clip history, saved project structure and usage statistics are stored.
- Cloudflare provides DNS and the secure connection in front of our API, and sees the requests that pass through it.
- Google (sign-in) authenticates you when you choose to sign in, and serves your profile picture on your account page.
- Google Fonts: the overlay previews load their fonts directly from Google, so your browser sends Google your IP address and browser details whenever a preview is shown, including when a clip is rendered. The website’s own fonts are served from our site and involve no request to Google.
- jsDelivr: the previews load two open-source animation scripts from the jsDelivr content delivery network, which receives your IP address and browser details.
- Hugging Face: choosing Generate captions downloads the pinned speech model from Hugging Face and its content delivery providers, which receive your IP address, browser details and model file requests. No video, audio or caption text is sent. Cached files are reused without downloading them again. See Hugging Face’s privacy policy.
- Social networks and email: when you choose X, Facebook, LinkedIn, Reddit, Telegram, WhatsApp, Pinterest or Email under Share, that service opens with the address of the overlay, including the text and colors you set, and handles it under its own privacy policy. On a phone, Share hands the address to your device’s share sheet instead. Nothing is sent until you choose.
- Paddle is the merchant of record for payments. Paddle collects your payment details, billing country and email at checkout, calculates tax and issues invoices, and acts as an independent controller of that data under its own privacy policy.
We may also disclose data when the law requires it.
How long we keep it
- Account, likes, bookmarks, turned-off guides and clip history: until you ask us to delete your account. Deleting an account also deletes its sessions, stored sign-in tokens, likes, bookmarks, turned-off guide list, clip history and saved Studio projects.
- Saved Studio projects: until you delete the project or your account.
- Sessions: a website session expires 7 days after it was last renewed and is deleted when you sign out. MCP tokens expire after one year; revoking a connected device deletes its session immediately.
- Usage statistics: kept without a fixed time limit; they hold no name, account or IP address.
- Service diagnostics: the database records are kept like usage statistics, without a fixed time limit; the log copies stay within our hosting providers’ log limits and are removed as those logs rotate or deployments are replaced.
- Emails to support: for as long as needed to handle the request and any follow-up, normally no more than two years.
- Payment records: kept by Paddle for as long as tax and accounting law requires.
International transfers
Account data is stored in the European Union. Modisign is operated from Ukraine, so data may be accessed from there to run the service and answer requests. Vercel, Cloudflare, Google, jsDelivr, Hugging Face and Paddle operate globally and may process data in other countries, including the United States. Where personal data leaves the European Economic Area or the United Kingdom, the transfer relies on safeguards such as the European Commission’s standard contractual clauses or the provider’s participation in the EU-US Data Privacy Framework.
Your rights
Wherever you live, you can ask us for a copy of your data, to correct it, or to delete it. Under the GDPR and similar laws you also have the right to restrict or object to processing, to receive your data in a portable format, and to complain to the data protection authority where you live.
To delete your account, or to use any other right, email support@modisign.app from the address of your Google account. We reply within 30 days and do not charge for it.
Children
Modisign is not directed at children. You must be at least 16 to create an account. If you believe a child has given us personal data, email us and we will delete it.
Speech model and open-source notices
Captions use Whisper tiny.en model and licence (quantized ONNX weights, Apache-2.0), based on OpenAI Whisper (MIT, copyright OpenAI). The runtime is Transformers.js 3.8.1 (Apache-2.0) with ONNX Runtime (MIT, copyright Microsoft). The linked licenses include their copyright and permission notices.
Changes to this policy
When we change what we collect or who receives it, we update this page and its effective date. If a change is significant, we will also email account holders.
Contact
Privacy questions and requests: support@modisign.app.