Privacy Details

1. About This Page

This page supplements the Studio Sadola Privacy Policy and describes how tranext (the "Software") handles information. It does not override that policy, and anything not covered here is governed by it.

• Studio Sadola Privacy Policy

2. Communications for License Activation and Starting a Trial

When the User performs Activation online, the Software sends the License Key, a Machine Code generated from device hardware information, and the operating system type, and receives a signed activation code. The Machine Code hashes hardware identifiers; the underlying hardware information itself is not sent. Choosing offline activation avoids this communication from the Software. The routine license check at startup is performed on the device and does not involve network communication for this check (see Section 5 for retrieving notices).

When the User starts the free trial, the Software sends the Machine Code and a value indicating the start operation to register the trial, and receives its status, start time, remaining days, and configured duration. This communication does not occur unless the User starts the trial, but the trial cannot be started offline. After the trial has started, the startup check is performed on the device and does not involve network communication.

For online License Activation, starting a trial, and other communications with our APIs, we process the source IP address for rate limiting to prevent abuse. We do not retain this source IP address.

3. What the Software Does Not Send

The Software does not automatically send audio data, transcription results, AI conversation content, materials generated by AI, usage statistics (except for version and other information specified in Section 5), operation history, or error logs (except when the User chooses to send them under Section 7).

4. Catalog and Update Checks and Downloads

At startup and every 24 hours thereafter while the Software is running, the Software retrieves a signed model catalog from its distribution source using a GET request. No User data is sent in this communication. This communication cannot be disabled through the normal settings.

By default, the Software automatically checks for engine updates at startup and automatically downloads an available update. It retrieves update information, engine binaries, and their hashes from engine distribution sources; for some engines, it also connects to the GitHub API or a distribution source to check or download candidates. Automatic update checks and automatic downloads can each be disabled separately in Settings.

By default, at startup the Software queries the app-update distribution source and retrieves update information for the selected channel. This automatic check can be disabled in Settings, and the User can also check manually. The app update itself is downloaded as an operating-system-specific package only when the User chooses to download it. No such download occurs without that action, and a download in progress can be canceled.

5. Retrieving Notices

The Software queries our API to retrieve notices from the Studio (such as updates, maintenance, and replies to inquiries) at startup and approximately every hour thereafter while it is running. On devices with an activated license, it sends the License Key and Machine Code, together with the version of the Software, the operating system type, and the edition of the Software. We use this information to track which notices have been read, answer inquiries, determine applicable versions, and provide support, and retain only the latest values for each device. Devices without an activated license (including those in a Trial) send only the product and operating system type, and do not send information that identifies the device. This communication cannot be disabled through the normal settings.

6. Model Search and Downloads and Importing from a URL

Only when the User searches for an AI model using a non-empty search query does the Software send that query to Hugging Face. Searches within the catalogs of transcription and speaker-separation models are performed on the device and do not send the query. When the User selects and downloads a model, the Software retrieves the selected file from Hugging Face or another distribution source. These communications do not occur unless the User performs the relevant action, and a download in progress can be canceled.

Only when the User imports from a YouTube HTTPS URL (youtube.com, www.youtube.com, m.youtube.com, music.youtube.com, or youtu.be) in Pro or during an active Trial does the Software connect to YouTube and its delivery destinations to retrieve audio or video. This communication does not occur unless the User performs the import, and a download in progress can be canceled.

7. Feedback and Diagnostic Information

When sending feedback, the Software displays all the information to be sent on a review screen before sending, including the entered content, basic operating environment, and any attached diagnostic information. Diagnostic information is attached by default, but the User can remove the attachment. Information is sent only when the User presses “Send with this content.” If diagnostics are attached, the Software sends the same zip shown on the review screen (up to 2 MiB).

If diagnostic generation fails or the generated diagnostics exceed the overall size limit, the Software notifies the User. The User can remove the attachment and send after reviewing the content again on the review screen. The Software does not silently remove diagnostics and send. If communication fails, the entered content is kept, and the User can resend after going through the review screen again.

The Software sends the product name (tranext), category (bug, feature request, question, or other), subject (up to 200 characters), message (up to 5,000 characters), email address only if entered, source app platform (Windows, macOS, or Linux), and Software version. It does not silently truncate entered content before sending. Basic operating environment information also includes the Software version, OS type and version, architecture (such as x64 or arm64), GPU vendor category, and computing method. This information is sent even if the diagnostic attachment is removed, and includes no device names or model names.

On devices with an activated license, the Software also sends the License Key, Machine Code, and OS classification to verify the license. The review screen indicates that these credentials will be sent for verification, but the key and Machine Code values do not appear on screens, in diagnostic information, or in logs. If the activation state changes after review, the Software does not send and asks the User to review again. During support submission, the server hashes the key and Machine Code for verification and stores a reference to the activation in the inquiry record. It does not store the key or Machine Code received for verification in their original form.

Attached diagnostic information includes the Software and Electron versions (Electron is the app runtime), OS, architecture, CPU core count, GPU count, GPU vendor categories, types, VRAM (GPU memory) capacity ranges and computing methods, memory size, free disk space, engine installation status, and license plan and status. Settings include boolean values, numbers, and selected options for settings defined by the app. Free-form strings such as paths, model names, device names, and URLs are represented only by whether they are set, without their values; arrays are represented only by their item counts. Diagnostics also include self-check items, results, reason codes, and time taken, together with whether the previous session ended normally and its version.

Diagnostic information includes operating records from the last seven days (such as timestamps, operation categories, result and error codes, time taken, counts, and memory or speed measurements), and app-update records for the same period (stages, results, and error codes). Operating records also contain random identifiers used to link related processing records (such as an ID for each app launch). These are not user or device names. When operating and app-update records exceed their size limits, newer records are kept in preference to older ones, which are omitted. The amount omitted is recorded in the diagnostic information. The overall limit for diagnostic information is 2 MiB. Computer names are removed, and file names are replaced with only their extensions (for example, *.m4a). Audio, video, and document contents, transcript text, and AI prompts or responses are not included. Any record line containing a License Key or Machine Code pattern is excluded in its entirety. Generation time and covered period are also included. The app discards data prepared for review within five minutes.

Diagnostic information we receive is stored in storage provided by Cloudflare, our service provider. Diagnostic information is automatically deleted 90 days after submission or when the inquiry record is deleted, whichever comes first. It can be viewed only by those Studio support staff who need it to respond to the inquiry. The message and email address stored in the inquiry record are deleted by a daily process once one year (365 days) has passed since resolution. The same process also deletes basic operating environment information. Email addresses are stored encrypted. Source IP addresses are not stored in inquiry records and are used for rate limiting in 60-second windows to prevent abuse.

8. AI features and external communication

The release version of the Software has no AI features that communicate externally (Web search or web-page retrieval).

9. Contact

For questions about how the Software handles information, contact support@sadola.dev. Requests for disclosure, correction, suspension of use, or deletion, and the handling of updates to this page, follow the Studio Sadola Privacy Policy.