Privacy Policy

SFXMint is maintained by flreey. Browsing, searching and downloading do not require a user account and connect to our service. A first generated sound needs no account; generating more and using the API or MCP server with a key requires an account, as described below.

Hosting, security and usage records

Cloudflare hosts the website, API and audio files. It processes IP addresses, request paths and browser headers for delivery and security. Our service also records API, preview, download and generation events, including the sound, format, request identifier and outcome.

We derive a hashed client identifier from connection information such as IP address and browser information. It is used for usage measurement and generation rate limits. Hashing does not make the identifier anonymous. Generation job records can retain this identifier separately from the daily rate-limit counters.

Search and sound generation

Search queries and result information are stored to diagnose missing or unsuitable results. Queries may be processed by Cloudflare Workers AI for translation or semantic search. Avoid putting personal information, passwords or confidential project details in a search.

The SFXMint plugin in ChatGPT and the SFXMint connector in Claude search the same library. What their requests contain (the search words, or the event or set the assistant asks for) is stored the same way, with the tool name and the hashed client identifier described above. We do not store the language, location or other details ChatGPT or Claude may attach to a request, and we see nothing else from your account or conversation there.

An earlier, retired on-site generator stored the prompt, requested duration, job status and related result information, and completed sounds may have been added to the library with their generation description. How the current generator handles your descriptions is described under Accounts and payments.

Accounts and payments

You sign in through Clerk, our authentication provider, which processes your email address, sign-in method (GitHub, Google or a password) and session information. If you sign in with GitHub or Google, that service shares your name, email address and profile picture with Clerk; SFXMint itself uses only the email address. We store your account identifier, email address, your Game kit purchases, and the generations you make, to count each day's free generations.

Payments, cards and receipts are handled by Stripe. We receive the payment status, amount, product and customer identifiers; we never receive your card details.

If you create API keys, we store a hash of each key (never the key itself), its name and when it was last used. Each request made with a key is logged with the endpoint, a short label such as the search words or tool name, and the status, so you can see your usage in the dashboard.

Your description and chosen length are sent to GPU servers we rent from a cloud provider, which generate the sounds. With AI Enhance on, and when you describe a change under “Not quite right?”, your words first go through those servers to a text model (DeepSeek, reached through OpenRouter), which writes them into a full sound description; that description is then what the GPU servers generate from. We store each generation (description, the rewrite, settings, status and file checksums) and the generated files with Cloudflare so you can play and download them later. Generated files are not added to the public library. Do not include personal or confidential information in a description.

For a Game kit, the description of your game and the scenes and sounds you add or change are sent the same way to that text model, which plans the kit and writes its sound descriptions. We store each kit (its description, scenes, sounds and settings) and the takes you make with Cloudflare so you can come back to them and download them.

A kit's chat is stored with the kit. Each message, with a short summary of the kit (sound names, states and today's generations, not your email or files), is sent to that text model (DeepSeek, reached through OpenRouter) to write the reply and choose the kit's next steps; we log how many messages you send and the model's token counts to keep each day's limit. The chat is deleted with its kit or your account.

Generating without an account uses a random identifier in a cookie set only on those requests, and a hash of your IP address, to count the free generation without an account and each day's per-address limit. When you create an account on the same browser, what you made moves to it. The files of a generation that no account keeps are deleted 7 days after it was made; its record (description, settings and status) stays.

With an account, the website keeps a hash of a random identifier in a cookie set only on our account and generation requests, and a hash of your network address (your IPv4 address, or the first half of an IPv6 address), with each generation, so the free tries are given once per browser and per network, not once per account. Neither the identifier nor the address is stored as it is. Accounts made with a temporary email address get no free tries.

If you delete your account in its sign-in settings, we delete your generated files and generations, revoke your API keys and remove your email address from our records. Records of payments (and of credits from before 2026-10-09) are kept for accounting and tax purposes, linked only to the account identifier.

Analytics and cookies

Cloudflare Web Analytics measures aggregate visits and page performance without analytics cookies. Separately, Google Analytics 4 collects page addresses, referrers, device information, approximate location and cookie identifiers. Google Analytics may set first-party cookies such as _ga and _ga_*. The current site loads it without a prior cookie-consent dialog.

We also run Umami, an open-source analytics tool, on a server we operate. It sets no cookies. It records page views and site events such as searches, previews and downloads, with the page address, the referring site, browser, operating system, device type, screen size, language and approximate location. Page addresses keep only search words, filters and campaign tags; test visits and automated browsers are not counted. Umami does not store IP addresses; it derives a visitor identifier from the IP address and browser information that changes each month.

You can use browser privacy controls or Google's Analytics opt-out add-on to block Analytics. Clearing cookies alone does not prevent new cookies on a later visit. Read how Google uses data from partner sites and the Cloudflare privacy policy.

Advertising choices

Display advertising is not currently enabled. Before enabling it, we will update this policy to identify the services in use and provide the applicable consent choices.

Advertising vendors such as Google AdSense may use cookies and similar identifiers to serve and measure ads based on visits to this and other websites. Google and its partners may use advertising cookies to personalize ads. Google's choices are available in My Ad Center; participating third-party vendors provide choices through YourAdChoices. See Google's advertising information. These settings are separate from Analytics opt-out controls.

Contact, retention and changes

If you email us, we receive your address and the information you include. We use those messages to handle the request. Apart from the files of free previews no account keeps, deleted after 7 days, the application does not currently apply an automatic deletion schedule to its search and generation records. We do not sell personal information.

For questions or a request concerning information you submitted, contact the maintainer with the relevant page or job identifier and approximate date. Do not send passwords or unnecessary sensitive data. Material changes to how the service handles data will be reflected here with an updated date.