Privacy

01.The short version

Your audio, video and text stay on your device. Every tool here runs in your browser tab: nothing you open is uploaded, and there is no server-side processing. The only things fetched from the network are the tools themselves (code, codecs and, for the AI tools, model files), which are the same for every visitor and contain nothing of yours. Once a tool has loaded, you can switch to airplane mode and it keeps working; an AI tool needs the network the first time, to download its model.

02.What does get collected

• Normal web-server logs. The host (Cloudflare Pages) records request data — IP, user-agent, which URL you hit — the same as basically every site. Used for outages and abuse, kept briefly.
• Google Analytics 4 — aggregated traffic counts (page views, country, browser), plus which tool you used, the file extension going in and coming out, whether the file was dropped or browsed for, how long a conversion took, a size band for the file, and a short category when something fails (for example "decode" or "memory"). No audio data, no filenames, no error text, no personal identifiers. To be precise about how this works: the analytics script loads on every page, and your consent choice controls whether it is allowed to store anything on your device. Decline and it runs without cookies, which means you are counted but not recognised on a later visit.

03.What is stored on your device, and only there

• localStorage: fex_consent_v1 (your consent choice), your last-used formats and settings so the controls remember them, and the last few tools you used so the homepage can offer them back to you.
• IndexedDB: the audio editor's project and its audio (so it survives a reload), audiobook chapters already read (so a long book can be finished later), a clip handed between the editor and another tool, and a file shared to the installed app. Nobody but you can read these; clear them by clearing this site's data in your browser.
• Cache Storage: the FFmpeg codec and the AI model files, so they are downloaded once rather than on every visit.

04.What does NOT get collected

• The files you convert, cut, record, transcribe or synthesise, and the text you type into the speech tools. We don't have them. They never leave your browser tab.
• Your microphone or a tab you record. The recording tools read them only while you are recording, and the audio goes into the file you download, nowhere else.
• The names of the files you process, or any of their contents. The settings you choose are counted — that a conversion was 320 kbps, say — but never tied to a file or a person.
• Account info. There are no accounts to make.

05.Advertising

This site shows ads, served by Journey by Mediavine, to pay for itself. Mediavine and its advertising partners may use cookies, pixels and similar technologies to collect information such as your IP address, browser and device type, the pages you view and how you interact with the ads, in order to choose, show and measure ads, which can include ads based on your interests. That collection is theirs and is covered by Mediavine's privacy policy; its partners are listed at mediavine.com/ad-partners.

The ad code runs in the page like any other script, but it is never given your files: they are opened inside the tool and never uploaded, so there is nothing for it to receive. Ads are not loaded at all on the pages that run AI models with isolated memory (the stem splitter, audio to text, text to speech and dictation).

To opt out of interest-based advertising, use the Digital Advertising Alliance or Network Advertising Initiative opt-out pages, your browser's tracking and cookie settings, or the consent choices shown with the ads where your region requires them. You will still see ads, but they will not be chosen from your browsing.

06.Third parties

• Google Analytics 4 — aggregated, anonymised traffic counts. Their privacy policy.
• Journey by Mediavine and its advertising partners (see Advertising above).
• Cloudflare Pages (hosting and CDN).
• unpkg, or jsDelivr if unpkg is unreachable — one 30 MB WebAssembly file (the FFmpeg core) loads from this public CDN the first time you use a tool that needs it, and is then cached on your device.
• Hugging Face — the AI tools (transcription, text to speech, voice cloning, dubbing, translation, speaker labels, the stem splitter) download their open model files from huggingface.co the first time you use them.

The CDNs and Hugging Face see your IP address at the moment of the download, as any web server would. None of the third parties sees your audio or your text. Every other library is served from this site.

07.Files the voice tools make

Speech made by the text-to-speech, voice-cloning and dubbing tools carries a short note in the file's own metadata saying it is synthetic (or a cloned voice). It is written into the file on your device and is not reported anywhere.

08.Your options

Change your analytics choice any time from "cookie settings" in the footer; for ads, see the opt-outs under Advertising. Clear this site's cookies and stored data from your browser settings to remove everything listed above. Or block JavaScript — the tools won't function, but you can still read these pages.

09.Questions

See contact.