Privacy notice

This notice explains which personal data I collect through maslabs.it and its products, why I collect them, who I entrust them to, how long I keep them and what you can ask me to do with them. It is written under articles 13 and 14 of Regulation (EU) 2016/679 (GDPR).

Who processes your data

The data controller is Marco Andrea Sala (MAS Labs), Via N. Rusca 30/b, 23023 Chiesa in Valmalenco (SO) — VAT IT04007420161. For any request about your data write to [email protected]: I answer personally.

I have not appointed a data protection officer (DPO) because the cases set out in article 37 of the Regulation do not apply.

Which data I collect, and only when you give them to me

Idea Audit and the "Tell me your idea" form. Name, email address and the text in which you tell me your idea, or the voice note if you choose to record one. The voice note is always optional: recording only starts when you press the button and allow the microphone. Text and voice note reach my inbox, attached to the email: I do not publish them and I do not store them on the website's server. I use them to answer you and to prepare the report.

How I prepare the report: to transcribe the voice note and for a first analysis (market, costs, rules) I use an artificial intelligence (the supplier is listed below), with an account that does not use your data to train models. Then I reread, check the numbers, correct and sign the report myself: the conclusions are mine, not a machine's. Legal basis: pre-contractual measures at your request (art. 6.1.b) and, for keeping the correspondence only, my legitimate interest in keeping track of business contacts (art. 6.1.f).

The MAS Labs account. Email address, the name you want me to call you by, the access codes I send you (valid for a few minutes), the password and PIN if you choose to set them (stored only as an encrypted fingerprint that nobody can read back), your billing details if you enter them and, if you top up credits, the history of payments and usage. They are needed to let you in, to keep track of credits and to issue invoices. Legal basis: performance of the contract (art. 6.1.b) and, for payment and invoice data, legal obligations (art. 6.1.c).

If you open Voxpert without an account, I recognise you as a guest with the technical cookie described below, so you do not lose what you fill in. If you do not come back and have not created anything, the guest is deleted automatically after 45 days.

What you put into Voxpert. Information about your business (profile, products, prices, opening hours, the "memory" Voxpert builds), texts, plans and the calendar, the photos, documents, videos and audio recordings you upload to your Materials (videos and recordings are watched and listened to by Google, Gemini, to write down what they contain and the words spoken: that text stays with the file), the messages you dictate. In the memory I also keep what you tell me about yourself that applies to all your businesses (who you are, how you like to work, how you want to be spoken to): you see it in Memory, under “You”, where you can correct or remove it, and it is deleted with your account. What you tell me for a single job stays tied to that job. If you ask Voxpert to read your website or to study how competitors communicate, it reads public pages and keeps the useful facts. The reviews you give me in “How you communicate today”, pasted or as screenshots, I read to understand what your customers praise and which doubts keep coming up: the screenshots are read by the artificial intelligence provider, without copying the names of the people who wrote the reviews, and I don't keep them; I keep the text with the analysis. If your profile includes your professional Instagram account, Voxpert also reads its public posts, images included (the text written in the images and the style of the feed), to prepare the plan and the conversation, and may use some of them as a style reference for new images; it does not keep your images, except the photos it suggests for your library (see the next point). All of this is used to prepare content for you. Legal basis: performance of the contract (art. 6.1.b). If other people's data appear in the materials (for example your staff in a photo), you must be entitled to use them: for those data I act on your behalf (see "When I process data on your behalf").

The photos on your website and your Instagram. To start from your real photos, Voxpert looks for the photos published on your website and on your professional Instagram account, the ones in your profile (never other people's pages), downloads them and shows them to you in your library as suggestions. They become your materials only if you tap «Keep». The ones you discard I delete straight away; the ones you don't decide on are deleted automatically after 7 days. Of a discarded photo I keep only the address it came from and a numeric fingerprint of the image, so as not to suggest it again. Legal basis: performance of the contract (art. 6.1.b).

The tutorial video of your website. If you ask Voxpert for a video that shows how to use your website or your app, a browser on Voxpert's server (Contabo, in the European Union) opens the public pages of your website, as someone visiting it from a phone sees them, and takes screenshots; if there is a cookie banner it closes it as a visitor would, in a browser that is then deleted. It doesn't enter private areas, fill in forms or buy anything. The text of the pages and the screenshots are read by the artificial intelligence provider that writes the script; I don't keep the text, the screenshots become the scenes of the video and stay with the video in your account. If data about other people appears on the pages, you must be entitled to use it (see "When I process data on your behalf"). Legal basis: performance of the contract (Art. 6.1.b).

Photos and videos from your phone. You can also send photos, videos and recordings to your library from your phone without signing in: you scan the QR code that Voxpert shows you on your computer. The link is temporary: it is valid for a few minutes (the duration is written under the code) and for a limited number of files. While it is valid, anyone who has it can upload files to your library: don't share it. The files arrive just like the ones you upload from your computer. Legal basis: performance of the contract (art. 6.1.b).

Filming yourself or recording your voice from the Studio. With «Record a video» and «Record your voice» the browser uses your camera and microphone, only after you allow it and only while you record. The recording stays in your browser until you tap «Use this»: if you redo it or close, it reaches no one. With «Use this» it is uploaded like any other file. Legal basis: performance of the contract (art. 6.1.b).

Dictation in Voxpert. When you press the microphone inside the Studio, the audio is turned into text by the artificial intelligence provider listed below and then discarded: I keep only the text, which you can correct.

The voice meeting with Voxpert. In the voice meeting you are talking with an artificial intelligence: it is also written on the screen. If you choose to prepare your plan by talking it through with Voxpert, the microphone turns on only when you press «Turn on the microphone» and turns off when the conversation ends. Your voice goes in real time to the artificial intelligence provider listed below, which answers you by voice and transcribes the conversation. I record the conversation for you, so you can listen to it again (see the next point). For the plan I keep the decisions written into the form (which you can correct) and the facts about your business that enter the memory as «to be confirmed». As soon as the conversation ends, I reread the transcribed text to pick out your ideas about your communication, how your work goes for your customers, your stories (episodes from your work, things your customers tell you, doubts you hear, real numbers) and a few of your sentences copied word for word, which I write into the summary, where you see them and can correct or remove them before generating the plan, and the facts you told me, which enter the memory as «to be confirmed». The ideas, the how it works, the stories and the sentences you leave in the summary go into the plan's brief and are used to write the posts. In the same reading I also clean up the text of your copy of the conversation (see the next point): I remove hesitations like «um» and correct obvious transcription errors, without changing what you said. In the same reading I also pick out the decisions we made together (what you asked for, chose or ruled out, and the suggestions of mine that you approved), each with your own words as proof: you see them again in the outline and in the plan, under «What you asked me for». I keep the text of the conversation together with the outline and the plan that come from it, as well as the text of the conversations in which you ask to change the outline: the steps that write the outline, the plan and the posts read it, so that they all start from what you said. It passes through the server and the artificial intelligence provider, it is encrypted on the server and I don't use it for anything else. It stays until you delete it from your Materials («Your conversations with Voxpert»), even if you delete the outline or the plan, and it disappears if you close your account. If the conversation is paused, or you change page while talking, the text so far and the summary stay on the server, encrypted, so you can pick up where you left off: for 24 hours at most, then the conversation closes and its text is kept like that of the other conversations. Legal basis: performance of the contract (art. 6.1.b).

The transcript of a briefing you upload. In the new communication plan you can upload the transcript of a briefing (downloaded from Voxpert, or of a meeting held elsewhere) instead of talking to Voxpert. I read it like the text of a conversation: I take your ideas, the decisions and the brief from it, and the facts about your business go into the memory as «to be confirmed». I don't keep the file: I keep the text, encrypted, like the text of conversations; the outline and the plan that come from it use it, and it stays until you delete it or close your account. If the file is a PDF, the artificial intelligence provider reads it first to get the text out. If other people speak in the transcript, only upload conversations you have the right to use: the people who spoke must know. Legal basis: performance of the contract (Art. 6.1.b).

The welcome. At the entrance of the Studio you can talk or write to Voxpert to tell it what you need; then it takes you where the work happens (the communication plan, a post, a video, the outline, the report), with what you have already told it. What you say or write there goes through the artificial intelligence provider so that Voxpert can answer; I record the spoken conversation for you like the others (see below), and I also keep the written one, encrypted: what you write to me and what I answer, with the time. I keep in the same way what you ask me in writing on an outline, a post, a plan, a report or a storyboard. You find it in your Materials, under «Your conversations with Voxpert», and it stays until you delete it or close your account. If I give you an opinion or an idea there, when the conversation ends the artificial intelligence provider reads it again and writes its key points, which you find in "Your work" together with what you had asked me: they stay until you delete them or close your account. If the work you ask for includes a communication plan, when you start it the text of the conversation (written and spoken) stays on the server, encrypted, together with the outline and the plan that come from it, like that of the plan briefing: the steps that write the outline and the plan read it, and it stays afterwards too, until you delete that conversation or close your account. If the spoken conversation is paused, or you change page while talking, the text so far stays on the server, encrypted, so you can pick up where you left off: for 24 hours at most, then the conversation closes. If you ask for something Voxpert doesn't do today (for example a long YouTube video or a newsletter), I keep that request in your words and it reaches me by email, so I know what to build (the artificial intelligence provider reads it again to tell me what kind of work it would take); if I reply, the reply reaches you by email and you also find it in the Studio. With the request I also keep what I need to understand it without asking you to repeat it: why my team can't do it and what it's missing, the things you had put on the table with what I had read in them, and the last lines of the conversation, encrypted. I read the conversation only in Voxpert's admin panel: it doesn't reach me by email. The request and the reply delete themselves after 12 months, and they also disappear if you close your account. Legal basis: performance of the contract (Art. 6.1.b) and, for the requests, my legitimate interest in improving the service (Art. 6.1.f).

The table. In the conversation with Voxpert, and then on the outline and plan pages, you can put photos, posters, documents, PDFs, videos, audio recordings and links (YouTube, websites, social pages, Google listings) on the table. I read them to understand what they contain and what you want done with them, and I use them to write the outline, the plan and the posts. Photos and documents are read by the artificial intelligence provider (Anthropic) and also go into your Materials; videos, audio recordings and YouTube links are watched and listened to by Google (Gemini); I open linked pages from the server, and the ones that don't open (Google listings, many social pages) I look up on the web. When you ask me for an opinion on something of yours, or how a part of it is made, I look at the whole of it again, with the same providers, together with your question and a few lines about your business: only the answer remains, in the conversation. If in the conversation you say you'll send me something later, you'll find a slot on the table waiting for it. What you put on the table is kept until you delete it, even if you delete the outline or the plan: photos and documents in your Materials, videos, audio recordings and links under «Your conversations with Voxpert». You can remove anything whenever you like, and everything goes if you close your account. Legal basis: performance of the contract (art. 6.1.b).

The conversation video. If you tap "Record the conversation", your browser records the Studio tab (or the window or screen you choose when Chrome asks) with your voice and Voxpert's. It stops when the conversation is paused and starts again when you resume it. When the conversation ends, or when you stop it yourself, the video downloads to your computer and a copy goes into your Materials, on the Voxpert server (Contabo, in the European Union): it stays there until you delete it or close your account. No artificial intelligence provider watches or listens to this video, unless you put it on the table yourself. If other people can be seen or heard in the recording, you are the one who must be allowed to record them (see "When I process data on your behalf"). Legal basis: performance of the contract (Art. 6.1.b).

Your video storyboards. If the work you ask for includes a video with a storyboard, when you start it the text of the conversation at the entrance (written and spoken) stays on the server, encrypted, together with the storyboard: the steps that read or write the storyboard, and those that change it when you ask me to, read it again, and it stays until you delete that conversation or close your account, even if you delete the storyboard. The storyboard you give me (a text, a PDF, a document, the frame images) is read by the artificial intelligence provider, like the other things on the table. When you approve the storyboard I make the video with the usual providers: the images (OpenAI, or Higgsfield when I choose it) and the animated scenes (fal.ai or Higgsfield), the voices (Google Gemini, with OpenAI as a backup), the music (Google Lyria), the editing on Voxpert's server; they receive the descriptions of the scenes and the text of the lines, not the conversation. If the storyboard uses a video or an audio of yours that you put on the table, I keep a copy of it with the storyboard, which is deleted with it; only the part the storyboard says goes into the video. I never put the recordings of your conversations with Voxpert in a video on my own: if you want to use them, you download them and put them on the table yourself. If in the storyboard you point to things that are on public pages of the websites you name (a piece of work to show, an animation, a photo, an audio), a browser on Voxpert's server (Contabo, in the European Union) opens those pages, presses the buttons that show them and takes them (a screenshot, a screen recording, the file of the photo or the audio): it doesn't enter restricted areas and doesn't fill in forms. To find the right piece, the artificial intelligence provider looks at a screenshot of the page. What I take stays with the storyboard and is deleted with it; you must be allowed to use that material too, like what you upload. If you ask to animate a photo of yours, the artificial intelligence provider first checks that there are no people in it (a real person is never animated), then the photo goes to the animated-scene provider. The storyboard and the video stay in your account with your other work until you delete them or close your account. If other people appear in your audio or video, it's up to you to be allowed to use them (see "When I process data on your behalf"). Legal basis: performance of the contract (art. 6.1.b).

A video where you speak. If you ask Voxpert for a video where you speak (a lesson, a course, a presentation), you give me a photo or a video of you and, if you like, a recording of your voice, and you tell me whether the person is you or a person created with artificial intelligence: I keep this statement of yours with the work and with the video. The person can only be you or a person created with artificial intelligence, never another real person. HeyGen makes the video: it receives the photo (or a frame of your video), the recording of your voice (at most the first three minutes, audio only) and what you say in the video; from the photo it makes an avatar that speaks with lip-sync, and from the recording a copy of your voice, only for this video. As soon as the video is ready, or if it fails, I delete from HeyGen the avatar, the copy of the voice, the uploaded files and the video. If you don't give me the recording, the text is read by a native-speaker synthetic voice from Google (Gemini), or from OpenAI if Google doesn't respond; if you ask me to polish the text so it's easier to say, the artificial intelligence provider (OpenAI) rewrites it. I use your face and your voice only to make the video you ask for, never to recognise you. For its whole length the video says it was created with artificial intelligence, and it stays in your account with your other work until you delete it or close the account; the things you had put on the table follow the rules of the table, and to work on them I keep a copy on the server for the time of the work, 2 days at most. Legal basis: performance of the contract (Art. 6.1.b).

A video course or tutorial. If you ask Voxpert for a video course or tutorial, you give me the text (your outline or your notes): I split it into lessons with code or, if you ask me to turn it into a spoken talk, the artificial intelligence provider (OpenAI) writes it. You choose who speaks: you, from your photo or from your "digital twin" made from a video of you; a person created with artificial intelligence whose image you have; or a HeyGen presenter, in which case your face isn't needed. Never another real person. Your face and your voice are biometric data: when the course uses them, before starting it I ask for your explicit consent (Art. 9(2)(a) of the Regulation) and keep it with the course; you withdraw it whenever you want by stopping the course. HeyGen receives your photo or your video (two and a half minutes at most), the recording of your voice when you give it to me and the text of the lessons; from them it makes an avatar (or your digital twin) and a copy of your voice, which stay on its servers only while the course is being made, 7 days at most, then I delete them; according to its biometric privacy notice, HeyGen destroys the face and voice data of people in the European Economic Area within 60 days of deletion. Before you pay for the digital twin, Google (Gemini) watches and listens to your video to check that it works (one person only, the face in view, clean audio), and I tell you what to change; Google receives it even if you then don't approve the course, like any video you put on the table. For the digital twin HeyGen asks you to record a consent sentence on its page, in front of the camera: it's its check that the person is you. When it's your turn I send you an email; if HeyGen rejects the recording, you can try again. If you don't give me the recording, the text is said by the voice taken from your video or by a native synthetic voice from Google (Gemini), or from OpenAI if Google doesn't respond. To check that each lesson says its script, I transcribe what is heard with the artificial intelligence provider (OpenAI). The lessons say for their whole length that they were created with artificial intelligence and stay in your account with your other videos; the copies of your photo, your video and your recording stay with the course and go when it's finished or when you delete it. Legal bases: performance of the contract (Art. 6(1)(b)) and, for your face and your voice, your explicit consent (Art. 9(2)(a)).

Subtitles for your videos. If you ask Voxpert for the subtitles of a video (a video you put on the table, or a public YouTube video you give me the link to), I write down the spoken words with their timings: the video you put on the table is listened to by the artificial intelligence provider that transcribes (OpenAI, which receives only the audio of the video) and, if nobody speaks in it or the words need correcting, watched by Google (Gemini); a YouTube video is watched by Google (Gemini); if you want them in another language, the artificial intelligence provider (OpenAI) translates them; if you want them on the video, Voxpert's server (Contabo, in the European Union) remakes your video with the captions. I don't download a YouTube video: Google watches it from the link. I delete an uploaded video from Google as soon as it has been read. The subtitles and the captioned video stay in your account with your other work until you delete them or close your account; the video you had put on the table follows the rules of the table; to work on it I keep a copy on the server for the time of the work, 2 days at most. If other people appear in the video, or you didn't make it, it's up to you to be allowed to use it (see "When I process data on your behalf"). Legal basis: performance of the contract (art. 6.1.b).

The table for your videos. If you put one of your videos on the table, to cut it, change its order, format, light or sound, join it to other things on the table or write on it, Voxpert's server (Contabo, in the European Union) measures its shots and pauses, and the artificial intelligence provider that transcribes (OpenAI, which receives only the audio of the video) writes down the spoken words with their timings, once per video, so you can cut a sentence by saying it. The words stay in the server's memory until it restarts: I don't keep them anywhere else. Voxpert's server does the editing: the edited video stays in your account (in your Materials and on the table, next to the original) until you delete it or close your account; the captions and the end card drawn for the edit stay on the server for the time of the work, one day at most. If you ask me to find a moment described in words ("when she opens the door"), Google (Gemini) watches the video on the table: it receives it, and I delete it from Google as soon as it has been read. If you ask for a model retouch on a stretch (removing or adding an object, changing the background or the time of day), that piece of video, without audio, goes to fal.ai with the request and comes back redone: it stays on its servers for one hour at most. I never change a person's face, body or voice. A voice-over is read by a native-speaker synthetic voice from Google (Gemini), or from OpenAI if Google doesn't answer, which receives only the text to say; the music is composed by Google (Lyria) from a description of the style, which contains none of your data; subtitles in another language are translated by OpenAI, which receives only their text; an image to place over the video, if you ask for one to be created, is made by OpenAI, which receives only the description of what it should show (one of your own photos is placed on the video by the Voxpert server, with no provider involved). If you ask to adjust just the voice or just the rest of the audio, the two parts of a video made by Voxpert are rebuilt by the server from its pieces, with no provider involved; those of any other video are separated by fal.ai, which receives only the audio. The redone pieces, the voices, the music, the created images and the audio parts stay on the server, linked to your account, until you close it; an edited video that contains one of them says in its metadata that it contains parts made with artificial intelligence. If other people appear in the video, or you didn't make it, it's up to you to be allowed to use it (see "When I process data on your behalf"). Legal basis: performance of the contract (art. 6.1.b).

Custom texts, voices and searches. If you ask Voxpert for a text that isn't a post (a letter, an email, a speech, a script…), the artificial intelligence provider (OpenAI) writes it with what Voxpert knows about your business. If you ask for a text to be read aloud, a native-speaker synthetic voice from Google (Gemini) reads it, or one from OpenAI if Google doesn't answer: they receive only that text and how it should be said. If you ask for a search, the artificial intelligence provider (OpenAI) looks for what you ask in public web pages (businesses, organisations, places, events, never information about private individuals) and I keep only what is written in the pages it read, with their address. Texts, audio files and lists stay in your account with your other work until you delete them or close the account; you do the sending. Legal basis: performance of the contract (Art. 6(1)(b)).

The recording for you, to listen to the briefing again. While you talk with Voxpert I record the conversation for you: both voices, yours and Voxpert's, and the transcribed text. It is there for you, to listen to and read again whenever you like, even from another device: you find it under «Listen to the briefing again», in the summary after the conversation, on the plan page and among your Materials, together with the links you gave Voxpert. During the conversation you see «Recorded for you» at the top: if you press «Don't record», the recording of that conversation is deleted. The recording is encrypted in your browser before it leaves; the key to open it stays on the server, itself encrypted, and is used only to show it to you when you open it from your account. I do not use it for anything else: I do not listen to it, I do not give it to anyone and I do not use it to train artificial intelligence models. It stays until you delete it, with «Delete the recording» or from your Materials, and it disappears if you close your account; when you delete it, I also delete the text of that conversation that I keep to work. Legal basis: performance of the contract (art. 6.1.b).

Recording conversations during the trial period, only with your permission. For a few days, while Voxpert is in its trial period, before the microphone turns on I ask whether I may also keep a copy of your conversations with Voxpert for myself, separate from the one for you: both voices, yours and Voxpert's, the transcribed text and a few moments of the conversation (for example when you press «Next» or when it pauses). I use them only to understand where Voxpert gets things wrong and to improve it: I do not publish them, I do not give them to anyone and I do not use them to train artificial intelligence models. This copy is encrypted in your browser before it leaves, and it can only be opened with a private key that I keep, outside the server: the server stores only encrypted data and cannot read it. If you say no, Voxpert works just the same. During the conversation you can press «Don't record»: the recording in progress is deleted, together with the copy for you. You can see and change your permission at any time in the Studio, under Account: if you withdraw it, I do not record you for the trial from the next conversation on. You can also ask me to delete the ones already made. Legal basis: your consent (art. 6.1.a), which you can withdraw at any time; withdrawing it does not affect what was done before (art. 7.3).

Waiting list for a tool in trial. If you join the waiting list of a tool that is not open to everyone yet, I process your name, email address and the line describing your business. I use them only to write to you when the tool is ready and to understand which businesses to start with. You can leave the list at any time by replying to the confirmation email. Legal basis: pre-contractual measures at your request (art. 6.1.b).

Support and proper functioning. From the admin panel I can see the state of your account and of what you do in Voxpert (plans, posts, publications, credit usage). I look at it only to support you, to check that the service works, to fix it when it gets things wrong and to prevent abuse. Legal basis: performance of the contract (art. 6.1.b) and my legitimate interest in running the service well (art. 6.1.f).

Technical data. The site runs through Cloudflare's network, which protects it from attacks and makes it faster: to do so it processes the IP address and the technical data of each request. The server records the IP address and the date of requests for security reasons and to rate-limit automated submissions by bots. Legal basis: legitimate interest in the security of the site (art. 6.1.f).

Visit statistics. I do not use Google Analytics or any other statistics or advertising profiling tool. Cloudflare only shows me aggregated traffic figures (how many visits, from which countries), without cookies and without identifying people.

Social channels connected to Voxpert

Today you can connect a Facebook Page and a professional Instagram account to Voxpert; other channels, such as LinkedIn, TikTok or YouTube, will follow. When you connect a channel:

• what I receive: the access key (token) the platform issues when you authorise Voxpert; the identifier, name and picture of the pages or profiles you choose to connect; basic public data on the posts published through Voxpert (for example views, reactions and clicks), to show you how they went;
• what I do with them: I publish only the content you have approved or scheduled, on the channels you have chosen, and I show you the results. I do not read your private messages, I do not write to anyone in your name, I do not use platform data for advertising, I do not sell them and I do not pass them on;
• how I protect them: access keys are stored encrypted on the server and used only by the publishing service;
• how long I keep them: as long as the channel stays connected. When you disconnect it, or remove it from the platform's settings, I delete the channel's keys and data within 30 days. The instructions are on the Data deletion page.

Legal basis: performance of the contract (art. 6.1.b), on the basis of the authorisation you give on the platform. The use of data received from Google APIs (for YouTube) complies with the Google API Services User Data Policy, including the Limited Use requirements.

When I process data on your behalf

Some Voxpert features collect or store data about other people on your behalf:

• the contact pages you create with Voxpert, where your customers leave their name, email and the other information you ask for, with their optional consent to receive your communications;
• the tracked links in your posts, whose clicks I count by recording only the date and the type of device (no IP address);
• the photos, documents and information about other people that you upload, including from your phone, or that you keep among the photos found on your website and your Instagram.

For these data you are the controller: I am the processor (article 28 of the Regulation) and I process them only to run the service for you, following the rules in the Terms of service. People who leave their details on one of your contact pages find your privacy notice there, with your name as controller. The contacts collected are deleted automatically after 24 months.

Artificial intelligence

Voxpert is an artificial intelligence: it writes texts, creates images and runs the voice meeting. When you talk with Voxpert it is written on the screen, and if you ask, it confirms it. The images it creates carry in the file's metadata an indication that they were generated with artificial intelligence. No decision about you is taken solely by a machine with legal or similarly significant effects (article 22 of the Regulation): what Voxpert prepares are proposals that you approve, and I sign the Idea Audit report myself.

The videos where you speak, made from your photo and with your voice, and the lessons of video courses say for their whole length that they were created with artificial intelligence.

Cookies and browser storage

This site uses no profiling cookies and no third-party cookies. That is why there is no banner: there is nothing to consent to. I use only technical tools that the site needs to work:

• studio_s — recognises your account, or you as a guest, while you use the site and Voxpert. It contains only a random code. It lasts 90 days, or until you sign out.
• lingua — remembers the language you chose. It lasts 12 months.
• browser storage — if you record a voice note in the form, your browser keeps it on your device until you send it, so you do not lose it if the connection drops (7 days at most). The browser also remembers a few interface preferences. These data stay on your device and never reach me.

Who I share them with

Your data are never sold or transferred to anyone for commercial purposes. They are processed, on my behalf and on my instructions, only by the technical suppliers needed to run the site and the products. They are processors under article 28 of the Regulation:

• Contabo (Germany): hosts the site, the database and Voxpert's files, on servers in the European Union.
• Cloudflare (United States): the network the site runs through, for security and speed.
• OpenAI (United States): Voxpert's artificial intelligence. It receives the information needed to write texts, create images, read documents, transcribe dictation, run the voice meeting, review the plan and every piece of work before delivery and prepare videos (the script, the subtitles and, if Google doesn't answer, the voice-over and the voice of the texts you ask to have read aloud), and transcribe and translate the subtitles of your videos when you ask for them (it receives the audio of the video and the text of the subtitles), and transcribe with their timings the words of the videos you put on the table (it receives only the audio), and search public web pages for what you ask me to find. Videos are edited on Contabo's servers. Through its APIs, data are not used to train the models and OpenAI keeps them only for the limited time it needs to monitor abuse.
• fal.ai (United States): turns some of the images created for videos into short animated clips (the animated scenes). It receives those images and the description of the movement; a photo of yours only when you ask to animate it in a storyboard, and never if there are people in it. On the table for videos it redoes the piece of your video you ask to retouch: it receives that piece, without audio, and the request, and returns it redone. If you ask to change the volume of just the voice or just the rest of a video's audio, it receives that audio, without the images, and separates the voice from the rest, without changing or remaking it. The files stay on its servers for one hour at most.
• ElevenLabs (United States), through fal.ai: listens to the synthetic voices of the videos I create and writes down their words with timings, for the subtitles. It receives only those voices, created by artificial intelligence, reading the video's script: never your voice or the audio of your videos. The files stay on fal.ai's servers for one hour at most.
• Higgsfield (United States): with the Seedance models it makes the animated scenes of videos, when I choose it instead of fal.ai, and makes the presenter of ads speak: an invented person, created by Voxpert, who says some lines of the script to the camera. It receives only the images created for those scenes, the portrait of the invented person, the piece of voice-over it speaks and the description of the movement; a photo of yours only when you ask to animate it in a storyboard, and never if there are people in it. Never your voice. When I choose it instead of OpenAI it also creates images and retouches photos: then it receives the description of each image and the photos it starts from, including yours when an image starts from a photo of yours.
• Anthropic (United States): the artificial intelligence I use to transcribe voice notes and prepare the first analysis for Idea Audit. For Voxpert it rereads the communication plan before you see the outline and improves it: it receives the conversation with Voxpert, the brief, what your business does and the communication plan. And, before the quote, it rereads the work you ask for at the entrance of the Studio and adjusts it: it receives the conversation at the entrance (I don't keep it for this reread), what Voxpert knows about your business and the work requested. And, before you see it, it rereads every piece of work the team makes for you (a post, an image, a video, a text, the storyboard of a video) and says what to improve: it receives the work with its images, what you asked for, the conversation it comes from and what Voxpert knows about your business. And it reads the photos and documents you put on the table, also when you ask me for an opinion on them: it receives the file, your question and a few lines about your business. Through its APIs, data are not used to train the models and are kept by Anthropic only for the limited time it needs to check for abuse.
• Google (United States): with Gemini it watches the videos and listens to the audio recordings you put on the table or in your Materials, and the YouTube videos you give me the link to, to tell me what they contain, to give me an opinion when you ask for one (it also receives your question and a few lines about your business) and, when you ask me, to write their subtitles with the timings; it watches the video you put on the table when you ask me to find a moment in it; and it reads the script of the videos with a native-speaker voice (the voice-over) and the texts you ask me to have read aloud; with Lyria it composes the background music of the videos from a description of the style written by Voxpert, which contains none of your data. It receives only those videos and recordings, or their link, the text of the script or of the text to be read and the description of the music. Through the paid service of its APIs, data are not used to train the models or to improve its products; I delete an uploaded video as soon as it has been read.
• HeyGen (United States): makes the videos where you speak. It receives your photo (or a frame of your video), the recording of your voice when you give it to me, and what you say in the video; from them it makes a speaking avatar and a copy of your voice, only for that video. As soon as the video is ready, or if it fails, I delete the avatar, the copy of the voice, the files and the video from its servers. For video courses it also receives your video (for the digital twin) and the text of the lessons, and keeps the avatar and the copy of the voice while the course is being made, 7 days at most.
• Resend (United States): delivers the emails I send you (access codes, notices, replies).
• Aruba (Italy): the mailbox of maslabs.it, where your messages arrive.

Others process data as independent controllers, for their own purposes and under their own privacy notices:

• Stripe: processes card details when you top up credits, also for its anti-fraud and anti-money-laundering obligations. I never see card details.
• The social platforms you connect (Meta, LinkedIn, TikTok, Google): they receive the content to publish on your channels.

Transfers outside the European Union

Cloudflare, OpenAI, fal.ai, ElevenLabs, Higgsfield, HeyGen, Anthropic, Google and Resend are established in the United States. The transfer takes place on the basis of the supplier's participation in the EU-US Data Privacy Framework or, where this does not apply, of the standard contractual clauses approved by the European Commission.

How long I keep them

• Ideas, messages and voice notes received through the form: 24 months from the last exchange, then deleted. If you ask me to delete only the recording, I do it right away and confirm in writing.
• Voxpert account and content: as long as the account is active. After 24 months without any sign-in I write to you; if you do not sign in within 30 days, I delete the account. If you have credits you bought and did not use, I write to you first to refund them.
• Guests without an account: 45 days from the last visit, if they have not created anything.
• Recordings of conversations during the trial period (only if you allowed them): 90 days at most, then they delete themselves.
• Recordings of your conversations with Voxpert kept for you: until you delete them or close your account.
• Text of your conversations with Voxpert, spoken and written: until you delete it or close your account, even if you delete the outline, the plan or the storyboard.
• Text of the briefings you upload: like the text of conversations. The uploaded file is not kept.
• What you put on the table (photos, documents, videos, audio, links and what I read in them): until you delete it or close your account.
• The storyboard of a video, with the copies of your videos and audio it uses: until you delete it or close your account.
• A video course, with the consent you gave me: until you delete it or close your account; the copies of your photo, your video and your recording until the course is finished. On HeyGen, the avatar and the copy of your voice: 7 days at most.
• The copy of one of your videos taken to work on it (the subtitles, a video where you speak): 2 days at most, then it deletes itself.
• The parts of a video made by Voxpert (the images of the scenes, the clips, the voice, the logo), to check it and correct it without redoing it: 30 days at most, then they delete themselves; straight away if you delete the video.
• Photos found on your website and your Instagram and not yet chosen: 7 days, then they delete themselves; straight away, if you discard them.
• Contacts collected by your contact pages: 24 months, then they delete themselves.
• Connected social channels: as long as they stay connected; after disconnection, 30 days at most.
• Payments and invoices: ten years, as the law requires, even after the account is closed.
• Database backups: 30 days at most, then they are deleted. Data you have deleted disappears from them too within that time.
• Server technical logs: 30 days at most.

What you can ask me

You have the right to ask me for access to your data, rectification, erasure, restriction of processing, portability, and to object to processing based on legitimate interest (articles 15-22 of the Regulation). Where processing is based on your consent, you can withdraw it at any time. Write to [email protected]: I answer within thirty days, free of charge. The steps to disconnect channels and delete the account are on the Data deletion page.

If you believe I am processing your data improperly you can contact the supervisory authority: in Italy it is the Garante per la protezione dei dati personali (www.garanteprivacy.it).

Changes

If I change anything in this notice I update the date at the top of the page. Previous versions remain available on request. The use of the products is also governed by the Terms of service.