Here is the article:
CVE-2026-63766: Unauthenticated Command Injection in GPT-SoVITS WebUI
A critical OS command injection flaw in the GPT-SoVITS voice-cloning WebUI lets unauthenticated attackers run arbitrary commands on the host, earning a CVSS score of 9.8.
What Is It
CVE-2026-63766 is an OS command injection vulnerability (CWE-78) in GPT-SoVITS, the open-source voice synthesis and cloning toolkit maintained by RVC-Boss. The flaw lives in webui.py, where the ASR, slice, denoise, and uvr5 functions interpolate unsanitized Gradio textbox values directly into shell commands executed with shell=True. Because path parameters are passed straight through, an attacker can inject shell metacharacters to execute arbitrary OS commands as the server process user; with no authentication required.
Why It Matters
The vulnerability carries a CVSS 3.1 base score of 9.8 (CRITICAL) with a network attack vector, low attack complexity, and no privileges or user interaction needed (CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H). A CVSS 4.0 score of 9.3 (CRITICAL) is also assigned. Successful exploitation grants full confidentiality, integrity, and availability impact; effectively complete control over the affected host. Given that GPT-SoVITS WebUI instances are often exposed for remote access, an internet-reachable deployment offers a trivial, pre-authentication path to remote code execution.
What's Vulnerable
- Vendor: RVC-Boss
- Product: GPT-SoVITS
- Affected versions: All versions up to and including
20250606v2pro - Repository: https://github.com/RVC-Boss/GPT-SoVITS
Patch Status
The supplied source material does not identify a fixed release or specific remediation steps. The issue is tracked publicly in the project's GitHub issue tracker (issue #2793) and detailed in a VulnCheck advisory. Operators should restrict network exposure of the WebUI and monitor the referenced sources for a patched version. This CVE does not appear in the supplied CISA KEV data, so there is no confirmation of active exploitation at this time.