How we handle your data

1. Who we are and who this applies to

This policy describes how Salven ("we", the "Service") collects, uses, shares and stores data from users of salven.ai and the @salven_bot Telegram bot.

It applies to anyone who visits the site, signs up, tops up their balance, publishes work to the Community, contacts support or takes part in the partner program, and to people we have sent a partner program invitation to. The rules for using the Service are set out in the agreement at /terms.

The legal grounds on which we process data are set out in section 3.

2. What we collect

We collect what the Service needs in order to work:

• Account: email, password (stored in protected form, see section 10), the site language and price currency you chose, sign-up date, credit balance and its history.
• Profile: display name, handle, avatar, headline, About text and social links, whatever you choose to fill in.
• Google sign-in: your Google account ID, email, name, profile photo and the sign-in keys Google issues. Telegram sign-in: your Telegram ID, username (@username) and the name on your Telegram profile.
• API keys and connected apps. For an API key: the name you gave it, the first characters of the key and its hash (the key itself is shown once, when it is created, and is not stored), and when it was created, last used and revoked. For a connected app (for example, Claude): its name and web addresses, when you allowed it access, when it last made a request to the Service and when it was disconnected; the connection codes and access tokens we issue to it are stored only as hashes. A generation records that it was started through the API, and by which key or app.
• One-time upload links and request IDs. For a link that a connected app, or a program using your key, creates so that you can add files from your device: a hash of the link, its expiry time and the list of files added through it (the file's storage address, type, size and file name). For a paid API request: the request ID the program or app sends, a hash of the request's contents and our answer to it, so that a repeated request does not charge your credits a second time.
• Content: prompt text (as typed and as automatically enhanced); files you upload, meaning images, video, audio (for example, a voice track for a talking video in Lipsync) and 3D models, including files copied from a web address at your request; generation settings and results; requests sent through the Telegram bot; and your AI chat history, if you used the chat before it was retired.
• Community: post titles, the prompt visibility setting, comments, likes, follows, view counts and reports on works (reason and comment).
• AI personas: the persona's name, @name and description, photos of the person, clothing and location photos, the "always" and "never" rules, the chosen voice, text for voice-overs and the voice tracks made from it.
• Characters: a character's name, its @name and one to four photos of one person.
• Support: ticket subject, category and text, your conversation with operators, and attached images.
• Partner program: your application text and links to your channels, the application status and decision, your referral code, users who signed up through your link, earned rewards, how you receive rewards (rubles or credits), payout requests, and the payout details you provide in a support ticket.
• Payments: amount, currency, payment method, credits issued, the chosen pack or plan, the paid period of a plan, the payment's status, its ID at the payment service and the service's fee. You enter card or account details on Freekassa's page or confirm the payment in your own bank: we never ask for or store them.
• Technical data: IP address, browser and device type, language, server and event logs (sign-ins, errors, requests to the site), and the log of your actions in the Service while you are signed in (section 7).
• Cookies and browser local storage, detailed in section 8.
• Partner program invitations: the email address we sent an invitation to. We don't save it in the Service's database; the email is sent through Resend (section 6). If you received such an email and don't want any more, write to support@salven.ai.

3. Why we use it

• To provide the Service: sign you in, run generations (including through the Telegram bot), store and show your work, keep your credit balance and automatically refund credits for failed generations. On the home page we look up the country of your IP address to show prices in rubles or dollars; this uses the DB-IP database stored on our own server, and the address is not sent anywhere.
• To carry out requests made through the API and by connected apps: check the key or token, run generations on behalf of your account, return to the program or app what it asked for (section 6), and avoid charging credits twice for a repeated request.
• To process payments: create a payment, receive the payment service's confirmation, add credits and extend a paid plan.
• To send service messages: the sign-up code and password reset link by email and, if Telegram is linked to your account, Telegram notifications about support replies and about your plan ending soon or having ended, as well as the results of generations started in the bot.
• To help you: handle support tickets and problems with payments and generations.
• To protect the Service and people: prevent fraud and abuse, review reports, and unpublish works that break the rules.
• To run the partner program: work out whose link a user came from, calculate rewards, make payouts and invite creators and site owners to the program.
• To improve the product: investigate errors, look at aggregated visit statistics and use the activity log to find where the Service is inconvenient.
• To comply with the law: keep payment records and respond to lawful requests.

We don't train AI models on your content.

We process data on these grounds:

• Performance of the agreement at /terms to which you are a party: your account, generations, payments, support, the partner program and service messages.
• Compliance with legal obligations: keeping payment records and responding to lawful requests from public authorities.
• Our legitimate interests, where they do not override your rights: protection against fraud and attacks, server logs, handling reports, visit statistics and the account activity log.
• Your consent, where the law requires it, for example for showing profile data and published works publicly (section 4).

Processing is automated and includes collection, recording, organization, accumulation, storage, updating, retrieval, use, transfer (provision, access), anonymization, blocking, deletion and destruction of data.

4. What others can see

Your generations are not visible to others until you publish them. Published content is open to anyone on the internet, including people without an account, and may be indexed by search engines:

• Your author page (an address like salven.ai/u/handle): name, handle, avatar, About text, follower, work and like counts, and your published works.
• A post: the work itself, its title, date, likes, comments and views. The prompt is visible unless you hide it; a hidden prompt is removed from public pages on the server, not just hidden in the interface.
• Comments: the text together with your name and avatar, to anyone who opens the post.

The partner whose link you signed up through sees the first two characters of your email together with its domain (or the start of your handle if there is no email), your sign-up date, whether you have topped up, and how much they earned from your top-ups.

The team may show published works as examples on the home page and the sign-in pages. If you don't want this, or you have unpublished the work, write to /support and we will remove it from there.

Support tickets, payments, uploaded files and the prompts of unpublished works are not shown to other users. Characters, AI personas and their photos stay hidden too, unless you publish works made with them. If a published work shows its prompt, the @names you wrote in it are visible.

Unpublished works and uploaded files don't appear in the feed or on the Service's pages. The files themselves are stored under long, hard-to-guess links, so don't forward a direct file link to anyone you don't want to see the file. A one-time upload link is an exception: while it works (30 minutes), anyone who opens it sees your handle and the names and sizes of the files already added through it, can get the links to those files, and can add files to your uploads. Give it only to people you trust.

You decide what to fill in on your profile and what to publish, and you can unpublish a work at any time. You can also ask us through /support to stop showing your data publicly, and we will remove it from public view within 3 working days. The team may unpublish works that have been reported.

5. Photos and voices of other people

If you upload photos, video or voice recordings of another person, for example for a character, an AI persona, a try-on or a talking video (Lipsync), you confirm that you have their consent or another legal right to do so. If the person is under 18, you need the consent of their legal guardian.

We use this material only for the generations you run, and we don't use it to recognize or identify people.

Some video effects made from photos of people are for people aged 18 and over. By running such an effect, you confirm that you are at least 18 and that the photo shows you or someone who agreed to it, and we check the photos you uploaded for it automatically with OpenAI (section 6): how many people are in them and whether the person looks under 18, and for some effects also whether the face is clearly visible and not covered, how large it is, how long their hair is, and on old ("archive") photos whether there is a face and where, so the photo can be cropped. The check does not establish who anyone is. Depending on its result the effect either runs or does not run, and then no credits are taken. The age estimate is not recorded with the order; the hair length and the face positions on archive photos are kept with the order, and the photos themselves as the generation's source files (section 9).

If your photos or voice are being used without your consent, report the post or write to support@salven.ai with "Legal" in the subject line, and we will look into it.

6. Who we share data with

We use external processors to run the Service. Each receives only what its function requires:

• fal.ai runs the generation models. It receives the prompt, generation settings and links to the images, video, audio and 3D models you upload, including photos for AI personas and try-on. When you write the @name of a character or an AI persona in a prompt, fal.ai also receives its photos, and some models also get its name in the prompt text. Your email, name and payment data are not shared.
• PiAPI (piapi.ai) runs some of the generation models. It receives the same as fal.ai: the prompt, generation settings, links to the files you upload and the photos of any character or AI persona whose @name is in the prompt. For the Seedance 2.5 and Seedance 2.0 models PiAPI keeps copies of these files, because those models take photos of people only from PiAPI's own library. Copies of the photos of characters and AI personas are kept while they are in use: a copy is deleted within an hour after you delete the character or the photo, and after 3 days without use. Copies of a generation's other files are deleted as soon as it ends; if that fails, PiAPI deletes them itself after 7 days. The one exception is a 480p Seedance 2.5 video: it is made as a draft from which the same video can be made in 1080p for 7 days, so PiAPI keeps the copies of its files for those 7 days and then deletes them. Your email, name and payment data are not shared.
• OpenRouter (openrouter.ai) runs some of the generation models: Seedance 2.0 at 4K, Nano Banana 2.1, Google Veo 3.1 Fast at 4K and ElevenLabs speech. It receives the prompt or the text to be spoken, generation settings and links to the pictures you add: frames, references and, for Nano Banana 2.1, the photos of any character or AI persona whose @name is in the prompt. OpenRouter keeps a finished video briefly, while we copy it to our storage.
• OpenAI enhances the prompt before generation and checks the photos before an 18+ video effect runs (section 5). To enhance a prompt, it receives the prompt text only (for AI persona shoots, that text includes the persona's description, its "always" and "never" rules and the location description). For the check, it receives links to the photos uploaded for such an effect, and those photos. No email, name or payment data is sent to OpenAI. Text for voice-overs is not sent to OpenAI.
• Cloudflare stores uploaded files and generation results in the cloud. It receives the uploaded files and generation results themselves.
• Freekassa (freekassa.net) accepts payments via SBP and by bank card. It receives the amount, the order number, the chosen payment method, your account email (or, for an account without one, a service address made of your internal Salven account ID and @salven.ai) and the IP address you pay from. You enter card or account details on its page or in your banking app. Until October 13, 2026 SBP payments were accepted by paypear.ru.
• Resend sends email: sign-up verification codes, password reset links and the partner program invitations we send to creators and site owners. It receives the recipient's address and the message text.
• Google handles sign-in with a Google account, if you choose it. It passes us the account details listed in section 2.
• Telegram handles sign-in through the bot, account linking and notifications (support replies, plan reminders). It receives your Telegram ID and the text of the messages you exchange with the bot.

Support operators see your tickets and the account data needed to answer them: name, email, avatar, balance, total top-ups, number of generations and sign-up date.

Service administrators can view accounts, generations (including unpublished ones, their prompts and uploaded files), payments, the prompt enhancement log and the account activity log, for moderation and for handling reports, errors and payment problems.

An app you connected to your account yourself (for example Claude, by Anthropic) receives from us only what it requests on your behalf: your account's handle and display name, your credit balance, the catalogue of models and effects, your generations, including those made on the site (the prompt, the model or effect, status, cost and links to the result files), the names and @names of your characters and AI personas and how many photos each has (plus links to the photos of a character the app itself creates), and links to the files added through that app, with each file's type, size and file name. With finished images and videos we also send the app small preview pictures. A program in which you use your own API key gets the same through the API, plus your internal account ID, the key's name and links to the photos of your characters and AI personas. Your email, password, payment records and support tickets are not available through the API or to connected apps. We send an app nothing on our own initiative: we only answer its requests, until you disconnect it or revoke the key in Settings, under Claude & API. What the app then does with what it received is governed by its provider's own terms, not by this policy. What is available through the API is described in detail at /developers.

If you ask for a file to be copied from a web address (a connected app or a program using your key can do this), our server requests the file: the site at that address sees a request from Salven's server, not from your device, and the request carries no data from your account. The copied file is saved among your uploaded files.

We disclose data to public authorities only in response to a lawful request and within the limits set by law. We do not sell data or pass it to third parties for their own purposes.

7. Analytics

For visit statistics we use Umami, running on our own server. It records pages viewed, the referring page, browser, device and operating system type, screen size and language.

Umami data stays with us, is not shared with anyone and is not linked to your account. Umami uses no cookies and no advertising identifiers. There are no third-party ad trackers or external counters on the site.

When you are signed in, we also keep a log of your actions in the Service, tied to your account: how you signed up, which pages and sections you open, which model you choose, when you start typing a prompt or add a file (not their contents), when you see there are not enough credits, when a generation could not start and why, when you open the top-up panel or payment, and your device type, window size and the site you came from. Prompt texts and file contents never go into this log. It is kept on our server for 180 days, is not shared with anyone and is used to find where the Service is inconvenient and to help you in support.

8. Cookies and local storage

We use only functional cookies and browser local storage. Without them, sign-in doesn't work and settings aren't saved:

• __Secure-authjs.session-token keeps you signed in, for up to 30 days. Service cookies for sign-in and protection against request forgery are also set: until you close the browser, or for a few minutes when you sign in with Google.
• salven_locale stores the interface language you picked, for up to 1 year. It is set only when you change the language yourself.
• salven_ref stores the code of the partner whose link brought you here, for 30 days or until you sign up.
• salven_studio_ followed by the section and mode (for example, salven_studio_image_generate) stores the model and settings you picked in the studio, for up to 1 year.
• salven_inf_prefs and salven_inf_last store your AI persona studio settings and the last persona you opened, for up to 1 year.
• sv marks that you have already opened a post so a view isn't counted twice, for 6 hours.
• Browser local storage (keys salven:composer:…) keeps a draft of your prompt and links to files you already uploaded in the studio, so they survive a page reload. The draft itself stays on your device. While a tab is open, it also holds a technical flag for page transition animations.

There are no advertising or third-party analytics cookies. You can delete cookies in your browser settings, but you will then need to sign in again and your studio choices will reset.

9. How long we keep data

We keep data for as long as it is needed for the purposes in section 3:

• Account, profile, content, characters, AI personas, support tickets and partner program data: while your account exists or until you delete them. If you ask us to delete your account (section 11), the data is removed within 30 days of our confirming that the request came from the account owner, except for what the law requires us to keep longer.
• Payment records (amount, date, status and payment ID): for the periods required by tax and accounting law (currently 5 years), including after the account is deleted.
• The prompt enhancement log (original and rewritten text): while your account exists. It is not removed when you delete a single generation.
• Server logs: for as long as they are needed to investigate errors and protect against attacks. The web server request log is kept for 14 days.
• Visit statistics: for as long as they are needed to analyze traffic; they are not linked to your account.
• The account activity log (section 7): 180 days, after which entries are deleted automatically. It is deleted together with the account.
• API keys and connected apps. A revoked key and a disconnected app stop working at once (apart from an upload link already issued, which works until it expires; see the next item), but the record of them (the name, the key's first characters and hash, the dates) stays until the account is deleted. A connected app's access tokens are deleted the moment you disconnect it. While an app is connected, an access token works for 1 hour and the token used to renew it for 90 days; expired tokens and connection codes are deleted at the next clean-up, no sooner than a day after they expire.
• One-time upload links and request IDs. A link works for 30 minutes and takes up to 10 files; its record is deleted at the next clean-up, no sooner than a day after it expires, while the files added through it stay in storage as your uploaded files until the account is deleted. An API request ID and the stored answer are kept for 24 hours; after that the record no longer applies and is deleted at the next clean-up.

When you delete a generation, it disappears from your gallery and its files are removed from storage. A record of it stays in our accounts until the account is deleted: which model made it and when, whether it succeeded, how many credits it took and what it cost us, and when it was deleted — without the prompt, the settings or the files. The source files you uploaded for it and technical copies (such as video preview frames and watermarked copies) are kept until the account is deleted. Copies saved to an AI persona are deleted separately, in the persona itself.

A character is kept until you delete it. We store separate copies of its photos, while the source files you uploaded stay in storage until the account is deleted. When you delete a character or remove a photo from it, the copies are deleted, except those already used in your generations: they remain the source files of those generations and are kept until the account is deleted.

We cannot delete copies of published works that other people have already saved.

10. Where data is stored and how we protect it

The Service's main servers and database run in a data center in the EU. Uploaded files and generation results are stored in Cloudflare's cloud storage, and copies are cached on our server for faster delivery.

Our servers are in the EU, and the processors from section 6 are foreign companies (fal.ai, OpenAI, Resend, Google and Cloudflare are based in the US), so data is transferred outside Russia. This transfer is needed to perform our agreement with you: without it we cannot run a generation, store files, send an email or let you sign in with Google.

Passwords, email verification codes, password reset links, API keys (apart from their first characters, which tell one key from another), and connected apps' codes and tokens are stored only as hashes, connections to the site are encrypted, and staff access to data is limited by role. The record of a one-time upload link also holds only its hash, but the link is the address of a page, so, like the address of any page you open, it gets into the web server request log and the visit statistics (sections 7 and 9); the account activity log records only that such a page was opened, not the link itself. The link stops working 30 minutes after it is created.

11. Your rights

You can:

• Find out what data we process about you, why and on what legal basis, who we share it with and how long we keep it, and get a copy.
• Correct inaccurate data. You can change your name, handle, avatar, headline, About text and links yourself on your profile and in Settings. You can also require us to block or delete data that is incomplete, outdated, unlawfully obtained or no longer needed.
• Hide a prompt, unpublish a work, and delete your generations, comments and characters. Section 9 explains what remains after a generation or a character is deleted.
• Delete your account and the data associated with it. There is no delete button in the interface: write to us, and we will delete the account within 30 days of confirming that the request came from its owner.
• Withdraw consent where processing relies on it (for example, public display of profile data): we will stop that processing within 30 days. Data needed to perform the agreement, or that the law requires us to keep, will continue to be processed while those grounds apply; to end the rest, ask us to delete your account.
• Stop Telegram notifications: unlink Telegram in Settings (unless it is your only way to sign in), or stop or block the bot.
• Complain about how we handle your data to Roskomnadzor (the Russian data protection authority) or to a court.

To exercise these rights, write to support@salven.ai from the address linked to your account, or open a ticket at /support. We may ask you to confirm that the account is yours. We will respond within 10 working days of receiving the request; if we need more time, we may extend this by up to 5 working days and will tell you why in advance.

If your balance still holds paid, unused credits, we will offer to refund them under the rules at /terms before deleting the account.

12. Children

The Service is not intended for anyone under 16. If we learn that an account belongs to a child, we remove it. If you are a parent and believe your child has given us their data, write to support@salven.ai.

13. Changes to this policy

We may update this policy, for example when new features or processors are added. The new version is published on this page, and the last-updated date is at the top. We will try to give advance notice of material changes on the site, and by email if your account has one. If a new version adds processing that requires consent, we will ask for it separately.

14. Contact

For any data-related questions, write to support@salven.ai, open a ticket at /support or message us on Telegram at t.me/salven_help.

Support replies Monday to Friday, 10:00 to 19:00 Moscow time. Send personal data requests (access, correction, deletion, withdrawal of consent) to support@salven.ai with "Personal data" in the subject line, and legal requests with "Legal" in the subject line. Other contacts are on the /contacts page.