{"id":"CVE-2025-49838","aliases":["EUVD-2025-21560"],"status":"published","title":"GHSL-2025-050: GPT-SoVITS Deserialization of Untrusted Data vulnerability","description":"GPT-SoVITS-WebUI is a voice conversion and text-to-speech webUI. In versions 20250228v3 and prior, there is an unsafe deserialization vulnerability in vr.py AudioPreDeEcho. The model_choose variable takes user input (e.g. a path to a model) and passes it to the uvr function. In uvr, a new instance of AudioPreDeEcho class is created with the model_path attribute containing the aforementioned user input (here called locally model_name). Note that in this step the .pth extension is added to the path. In the AudioPreDeEcho class, the user input, here called model_path, is used to load the model on that path with torch.load, which can lead to unsafe deserialization. At time of publication, no known patched versions are available.","severity":{"version":"4.0","score":8.9,"vector":"CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N/E:P","level":"high","assessor":"cna","from":"cve"},"epss":{"day":"2026-11-10","score":0.00784,"percentile":0.56389,"model":"v2026.06.15"},"facts":{"in_kev":false,"known_exploited":false,"public_exploit":false,"fix_available":false},"cwes":["CWE-502"],"published":"2025-07-15T20:36:46Z","modified":"2026-10-28T02:30:35Z","products":[{"vendor":"rvc-boss","product":"gpt-sovits"},{"vendor":"rvc-boss","product":"gpt-sovits-webui"}],"references":[{"url":"https://securitylab.github.com/advisories/GHSL-2025-049_GHSL-2025-053_RVC-Boss_GPT-SoVITS/","tags":["x_refsource_CONFIRM","Exploit","Third Party Advisory","advisory"]},{"url":"https://github.com/RVC-Boss/GPT-SoVITS/blob/165882d64f474b3563fa91adc1a679436ae9c3b8/tools/uvr5/vr.py#L216","tags":["x_refsource_MISC","Product","web"]},{"url":"https://github.com/RVC-Boss/GPT-SoVITS/blob/165882d64f474b3563fa91adc1a679436ae9c3b8/tools/uvr5/webui.py#L157","tags":["x_refsource_MISC","Product","web"]},{"url":"https://github.com/RVC-Boss/GPT-SoVITS/blob/165882d64f474b3563fa91adc1a679436ae9c3b8/tools/uvr5/webui.py#L192-L205","tags":["x_refsource_MISC","Product","web"]},{"url":"https://github.com/RVC-Boss/GPT-SoVITS/blob/165882d64f474b3563fa91adc1a679436ae9c3b8/tools/uvr5/webui.py#L64-L70","tags":["x_refsource_MISC","Product","web"]},{"url":"https://github.com/CVEProject/cvelistV5/tree/main/cves/2025/49xxx/CVE-2025-49838.json","tags":["advisory"]},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2025-49838","tags":["advisory"]}],"sources":{"cve":[{"id":"CVE-2025-49838","title":"GHSL-2025-050: GPT-SoVITS Deserialization of Untrusted Data vulnerability","description":"GPT-SoVITS-WebUI is a voice conversion and text-to-speech webUI. In versions 20250228v3 and prior, there is an unsafe deserialization vulnerability in vr.py AudioPreDeEcho. The model_choose variable takes user input (e.g. a path to a model) and passes it to the uvr function. In uvr, a new instance of AudioPreDeEcho class is created with the model_path attribute containing the aforementioned user input (here called locally model_name). Note that in this step the .pth extension is added to the path. In the AudioPreDeEcho class, the user input, here called model_path, is used to load the model on that path with torch.load, which can lead to unsafe deserialization. At time of publication, no known patched versions are available.","status":"published","published":"2025-07-15T20:36:46.859Z","modified":"2025-07-16T13:41:00.432Z","cvss":[{"version":"4.0","score":8.9,"vector":"CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N/E:P","level":"high","assessor":"cna"}],"cwes":["CWE-502"],"updated_at":"2026-11-02T14:32:24.825718Z"}],"euvd":[{"id":"EUVD-2025-21560","description":"GPT-SoVITS-WebUI is a voice conversion and text-to-speech webUI. In versions 20250228v3 and prior, there is an unsafe deserialization vulnerability in vr.py AudioPreDeEcho. The model_choose variable takes user input (e.g. a path to a model) and passes it to the uvr function. In uvr, a new instance of AudioPreDeEcho class is created with the model_path attribute containing the aforementioned user input (here called locally model_name). Note that in this step the .pth extension is added to the path. In the AudioPreDeEcho class, the user input, here called model_path, is used to load the model on that path with torch.load, which can lead to unsafe deserialization. At time of publication, no known patched versions are available.","status":"published","published":"2025-07-15T20:36:46Z","modified":"2025-07-16T13:41:00Z","cvss":[{"version":"4.0","score":8.9,"vector":"CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N/E:P","level":"high"}],"updated_at":"2026-11-02T20:32:07.830372Z"}],"nvd":[{"id":"CVE-2025-49838","description":"GPT-SoVITS-WebUI is a voice conversion and text-to-speech webUI. In versions 20250228v3 and prior, there is an unsafe deserialization vulnerability in vr.py AudioPreDeEcho. The model_choose variable takes user input (e.g. a path to a model) and passes it to the uvr function. In uvr, a new instance of AudioPreDeEcho class is created with the model_path attribute containing the aforementioned user input (here called locally model_name). Note that in this step the .pth extension is added to the path. In the AudioPreDeEcho class, the user input, here called model_path, is used to load the model on that path with torch.load, which can lead to unsafe deserialization. At time of publication, no known patched versions are available.","status":"published","published":"2025-07-15T21:15:32.593Z","modified":"2026-06-17T09:31:59.51Z","cvss":[{"version":"4.0","score":8.9,"vector":"CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N/E:P/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X","level":"high","assessor":"secondary"},{"version":"3.1","score":9.8,"vector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H","level":"critical","assessor":"primary"}],"cwes":["CWE-502"],"updated_at":"2026-11-02T14:11:18.782156Z"}],"osv":[{"id":"CVE-2025-49838","title":"GHSL-2025-050: GPT-SoVITS Deserialization of Untrusted Data vulnerability","description":"GPT-SoVITS-WebUI is a voice conversion and text-to-speech webUI. In versions 20250228v3 and prior, there is an unsafe deserialization vulnerability in vr.py AudioPreDeEcho. The model_choose variable takes user input (e.g. a path to a model) and passes it to the uvr function. In uvr, a new instance of AudioPreDeEcho class is created with the model_path attribute containing the aforementioned user input (here called locally model_name). Note that in this step the .pth extension is added to the path. In the AudioPreDeEcho class, the user input, here called model_path, is used to load the model on that path with torch.load, which can lead to unsafe deserialization. At time of publication, no known patched versions are available.","status":"published","published":"2025-07-15T20:36:46.859Z","modified":"2026-10-28T02:30:35.298336928Z","cvss":[{"version":"4.0","score":8.9,"vector":"CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N/E:P","level":"high"}],"cwes":["CWE-502"],"updated_at":"2026-11-02T15:44:50.037003Z"}]}}