Privacy Policy

Summary

• Audio and video files are processed in your browser, except when you choose AudioCutter Pro's server transcription or server vocal separation.
• We do not upload, store, listen to, inspect, or sell your media files. The only upload is the audio you send when you choose AudioCutter Pro's server transcription or server vocal separation, and we do not keep it.
• We do not collect file names, audio contents, video contents, waveform data, transcripts, or exact media metadata.
• We use privacy-safe product analytics to understand whether tools work, fail, or complete successfully.
• Google advertising and analytics services may use cookies or similar technologies when their scripts are active on the site.

Local file processing

When you select a file in AudioCutter, the browser reads it locally through standard web APIs. Processing such as splitting, trimming, merging, converting, and extracting audio happens on your device. The file itself is not sent to AudioCutter, Axleno, Supabase, Cloudflare, Google, or any third-party storage service by our application code. The only exceptions are AudioCutter Pro's server features, when you choose them: server transcription sends the audio through Cloudflare to DeepInfra, and server vocal separation sends it through Cloudflare to Replicate, as described below.

You are responsible for saving processed files to your device. Temporary browser memory can be cleared when you refresh or close the page.

Product analytics

AudioCutter records coarse product events so we can see whether the application is actually usable. These events include page views, time spent on a page (as a range), file selection, decode success, processing start, success, cancellation and errors, result previews, downloads, switching a result to another tool, and the type of unexpected script errors.

Analytics events may include coarse values such as the tool name, input and output format, file size bucket, duration bucket, browser and operating system family with their major version, device type, a rough range of the device's memory and processor cores, whether the browser supports WebGPU (graphics acceleration that in-browser tools can use), language, the country of the request, the kind of site that referred you (for example "search" and the search engine's name, or another site's domain), and campaign tags (utm) present in the link you followed. They do not include file names, full URLs or referrer paths, full user agents, IP addresses, exact screen sizes, or audio/video contents.

To count visitors without cookies or a persistent identifier, the analytics service derives an anonymous value from your IP address and browser type combined with a random value that changes every day and is deleted after two days. Only that value is stored, never the IP address, so it cannot be linked to you or across days. Events from automated browsers and crawlers are not counted.

Feedback

If you send feedback, we receive the message you write, the optional email address you provide, the page where the feedback was sent, and limited technical context needed to diagnose bugs. Do not include private media content, passwords, or other sensitive information in feedback messages.

Advertising and cookies

AudioCutter may use Google AdSense and related Google advertising services. Third-party vendors, including Google, may use cookies to serve ads based on a user's prior visits to this website or other websites. Google's use of advertising cookies enables Google and its partners to serve ads based on visits to this site and other sites on the Internet.

You can opt out of personalized advertising through Google Ads Settings. You can also learn how Google uses data on partner sites at Google's partner sites policy page.

Depending on your region, Google or other consent tools may show additional privacy choices before personalized ads are used.

Infrastructure providers

The site is delivered through Cloudflare. Privacy-safe analytics and feedback endpoints use Supabase Edge Functions and database storage. These providers process technical request data as needed to deliver, secure, and operate the service.

AudioCutter in Claude and ChatGPT

When you use AudioCutter inside Claude or ChatGPT, the editor runs on your device inside the chat and your file is not uploaded to us. Our server at mcp.audiocutter.app, delivered through Cloudflare, receives only the settings of each request, such as the start time, end time and output format. When a file is attached in ChatGPT, ChatGPT also sends a temporary link to that file; we do not use or store it, and the editor loads the file directly from ChatGPT. The conversation itself is handled under the privacy policy of the assistant you use.

AudioCutter Pro purchases

AudioCutter Pro is sold through Link, Stripe's checkout service, which acts as the seller (see the AudioCutter terms). Card and payment details go to Link and Stripe (and, for buyers in China, to Alipay through Stripe), never to AudioCutter.

After payment, Axleno's license server at license.axleno.com receives your email address and the checkout reference from Stripe, signs your license key and emails it to you through Resend. It keeps a record of the license so a refund can cancel it.

The key is kept in your browser's local storage and checked there. About once a day the browser downloads the list of cancelled licenses from license.axleno.com; that request carries no key, email address or file information.

Product analytics count when the Pro offer is shown, when a checkout is opened and when Pro is turned on, with the plan (7-day pass or 1-year license) and nothing about the buyer.

Server transcription (AudioCutter Pro)

AudioCutter Pro includes server transcription (speech to text). It is used only when a Pro user chooses it in the Transcribe tool. With the in-browser option, your audio stays on your device: the in-browser model downloads its model files from models.audiocutter.app and processes everything locally.

When you choose server transcription, your browser converts the audio to 16 kHz mono and sends it over HTTPS, in pieces of up to 5 minutes, to AudioCutter's own Cloudflare function at audiocutter.app/api/transcribe. That function forwards each piece to DeepInfra, Inc. (United States), which runs the Whisper large-v3-turbo model and returns the text and timestamps. The transcript goes back only to your browser.

Neither AudioCutter nor DeepInfra stores the audio or the text. DeepInfra processes it only to return the transcript and keeps no copy (zero data retention). The function's logs contain only event names and numbers: no audio, text, key or IP address.

Your Pro key is sent with each request to prove the purchase. To count each license's transcription time (8 hours in total with the 7-day pass, 40 hours with the 1-year license), a database at Cloudflare (D1) keeps a one-way hash of the license id, the seconds used and timestamps. Nothing else is stored.

Before audio is sent, Cloudflare Turnstile checks that the request comes from a person. Turnstile may process browser and device signals for that check, and it runs only on the transcription and vocal separation steps.

Server vocal separation (AudioCutter Pro)

AudioCutter Pro includes server vocal separation. It is used only when a Pro user chooses Fast in the Vocal Separator. When you choose Private, your audio stays on your device: the in-browser model is downloaded from models.audiocutter.app and does all the work locally.

When you choose it, your browser converts up to 10 minutes of the audio to a 192 kbps MP3 and sends it over HTTPS to AudioCutter's own Cloudflare function at audiocutter.app/api/separate. That function passes it to Replicate (United States), which runs the Demucs model and returns the vocal and instrumental tracks. They go back only to your browser.

AudioCutter does not store the audio or the tracks. At Replicate, the uploaded copy is deleted when the run ends and expires by itself within 24 hours at the latest; the run's inputs and outputs are deleted after 1 hour. The function's logs contain only event names and numbers: no audio, key or IP address.

Your Pro key is sent with each request to prove the purchase. To count each license's separation time (2 hours in total with the 7-day pass, 10 hours with the 1-year license, each file counting at least 1 minute), a database at Cloudflare (D1) keeps a one-way hash of the license id, the seconds used and timestamps, and each run's reference at Replicate, removed the next time the feature is used after 2 days, or after 30 days if nobody came back to see how the run ended. Nothing else is stored.

Before audio is sent, Cloudflare Turnstile checks that the request comes from a person, as it does for server transcription.

Contact

AudioCutter is an Axleno product. For privacy questions or product support, contact [email protected].

Last updated: November 6, 2026.