External risk intelligence

GPT-SoVITS OS Command Injection via Web UI

CVE advisorySeverity: CRITICAL (CVSS 9.3)

CVE-2026-63766

The vulnerability exists in a web-based user interface (Gradio webui.py) designed for audio processing functions. Such web applications are commonly deployed as network-accessible services to allow users to interact with the interface, making them reachable via the network in typical deployment scenarios.

OS Command Injection

Halo Surface Signal: 4 out of 5 — likely to be public-facing.

External exposure likelihood

Horizon Alert

Summary of the vulnerability and why it matters

GPT-SoVITS, a technology used for audio processing, has a critical vulnerability that allows unauthenticated attackers to execute arbitrary operating system commands. This occurs because the application directly inserts user-provided data into system commands without proper sanitization. The main concern is to confirm if this technology is in use and if so, to assess the exposure.

  • Allows unauthorized command execution.
  • Critical flaw in audio processing technology.
  • Confirm relevance and assess exposure.

Attack Path

How an attacker could exploit the issue

An attacker can exploit this vulnerability by sending specially crafted requests to the web interface, targeting functions like ASR, slice, denoise, or uvr5. Because these functions process user-supplied data from text boxes without proper sanitization and directly use it in shell commands, an attacker can inject malicious commands. This could allow them to execute arbitrary operating system commands on the server as the user running the application.

  • Network access required.
  • Inject commands into textbox inputs.
  • Execute arbitrary commands on server.

Live Threat

Current exploitation, exposure, and threat context

This vulnerability could allow an unauthenticated attacker to execute arbitrary operating system commands on the server. This is possible when the ASR, slice, denoise, or uvr5 functions are used, as unsanitized user input from Gradio textboxes is directly incorporated into shell commands.

  • Arbitrary OS command execution.
  • Via web interface path parameters.
  • Server compromise.

Operational Fix

Recommended remediation, mitigation, and detection steps

The OS command injection vulnerability in GPT-SoVITS's web UI indicates that application owners and platform teams are likely responsible for remediation. The first practical step is to identify all instances of the affected technology, confirm their network reachability and business criticality, and then assign ownership for a risk-based remediation plan.

  • Confirm application and platform ownership.
  • Verify external reachability and business criticality.
  • Plan and coordinate remediation actions.

Supplementary metadata

Validate whether this threat affects your internet-facing exposure.

Halo Threat Intelligence helps prioritize remediation with Halo Surface Signal and H/A/L/O context. Start exposure validation with a free external attack surface trial.

Frequently asked questions

What is GPT-SoVITS?

GPT-SoVITS is a software framework widely used for audio processing tasks, such as voice cloning and speech synthesis. It features a web-based user interface built with Gradio that allows users to perform complex operations like automatic speech recognition (ASR), audio slicing, denoising, and vocal removal (uvr5) through a browser.

What is the weakness in CVE-2026-63766?

This vulnerability is an OS command injection, categorized as CWE-78. It occurs when the software takes input provided by a user in the web interface and includes it in a system shell command without cleaning it first. Because the application processes this input as a command, an attacker can append their own malicious instructions, which the operating system will then execute with the same permissions as the GPT-SoVITS server process.

How can an attacker trigger this vulnerability?

An attacker triggers the bug by inputting specially crafted text into fields associated with the ASR, slice, denoise, or uvr5 functions. The vulnerability requires the application to process these inputs through its shell execution logic. Simply viewing or navigating the web interface does not trigger the flaw; the attacker must actively submit malicious path parameters or data strings designed to break out of the intended command structure.

Is my GPT-SoVITS instance at risk?

If your instance is reachable over a network, it is at higher risk. Halo Surface Signal identifies this as a critical concern because the Gradio web interface is frequently deployed as a network-accessible service. If your installation is exposed to the internet or an untrusted network, an attacker can interact with the affected input fields remotely to compromise the underlying server.

How do I respond to this threat?

Start by identifying all deployed instances of GPT-SoVITS within your environment. Once located, verify their network reachability and determine who owns or manages each server. Since this flaw allows for full system command execution, prioritize isolating any publicly accessible instances from the network until you can coordinate a secure remediation plan with your technical teams.

References