Skip to content
Vulnerabilities

Critical Command Injection Flaw in GPT-SoVITS 20250606v2pro Allows Remote Code Execution

Unauthenticated attackers can execute server-side commands via manipulated output folder paths in the GPT-SoVITS voice cloning tool, according to the NVD.

Illustration: A sleek, modern desktop computer sits on a minimalist desk with its monitor displaying a blank interface.
AI-generated illustration: DataSafe Times

Key points

  • CVE-2025-71429 is rated 9.3 CRITICAL on the CVSS scale by the National Vulnerability Database.
  • The flaw affects GPT-SoVITS versions up to 20250606v2pro, specifically within the UVR5 module.
  • Attackers exploit the vulnerability by injecting shell substitutions into output folder textboxes.

The short version

  • A critical OS command injection vulnerability exists in the GPT-SoVITS voice cloning software.
  • The flaw allows unauthenticated users to run arbitrary commands on the host server.
  • Versions up to 20250606v2pro are affected, with no confirmed patch mentioned in the record.

The details

The National Vulnerability Database published record CVE-2025-71429, assigning it a CVSS score of 9.3. This rating places the vulnerability in the critical severity range. The flaw is an OS command injection weakness, classified under CWE-78. It resides in the file tools/uvr5/vr.py within the GPT-SoVITS application. The affected software is produced by RVC-Boss. The vulnerability impacts all versions of GPT-SoVITS up to and including 20250606v2pro. Users running these specific builds are exposed to the risk described in the advisory.

The context

An unauthenticated attacker can exploit this issue through the UVR5 WebUI interface. The attack vector involves the output folder textboxes used for exporting audio files. When a user selects mp3 or m4a export formats, the application processes the provided folder path. Attackers can insert shell substitutions, such as $(...), into these path fields. The server then executes these injected commands. The commands run with the privileges of the server user. This allows for potential full system compromise if the server user has elevated permissions. The vulnerability does not require prior authentication to the application.

What should you do?

The National Vulnerability Database record does not indicate that a fix or update has been released. Users of GPT-SoVITS should verify their current version against the affected list. If running version 20250606v2pro or earlier, the risk is active. Monitor vendor communications for a patched release. Until a fix is confirmed, consider restricting network access to the UVR5 WebUI. Isolate the service from untrusted networks to reduce exposure. Do not expose the interface to the public internet.

What to do and how to stay safe: GPT-SoVITS

  • Verify the installed version of GPT-SoVITS to determine if it is at or below 20250606v2pro.
  • Restrict network access to the UVR5 WebUI to trusted internal networks only.
  • Monitor vendor announcements for a security update addressing CVE-2025-71429.
  • Avoid exporting audio files via the web interface if the service is publicly accessible.

Step-by-step guide: Automatic Software Updates: 8 FAQs on Security and Risk

General security guidance from the DataSafe Times newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

What is the CVSS score for CVE-2025-71429?

The National Vulnerability Database rates CVE-2025-71429 as 9.3, which is classified as critical.

Which versions of GPT-SoVITS are affected?

All versions of GPT-SoVITS up to and including 20250606v2pro are affected by this vulnerability.

How do attackers exploit this flaw?

Attackers inject shell substitutions like $(...) into output folder paths in the UVR5 WebUI to execute commands.

Sources

  1. CVE Program
GPT-SoVITSRVC-BossCVE-2025-71429OS command injectionUVR5

Related stories

Cybersecurity news without the noiseGet Alerts