Problem loading model

[SyamsQ]

Based on the following image, which engine does OmniVoice Studio actually use—Omnivoice or VoxCPM?

I currently want to use VoxCPM because Omnivoice is tied to the NC model related to the Emilia license.

[CoderDeltaLAN]

Hi, based on the screenshot, VoxCPM2 is the currently selected TTS engine—the small ACTIVE badge beside the VoxCPM2 name is what identifies the active backend. The AVAILABLE and GPU ACTIVE badges only mean that an engine is installed and can use that device; they do not indicate which engine is selected. OmniVoice TTS in the top-right “Loaded models” panel means the previous OmniVoice model is still resident in memory, not that new synthesis is being routed through it. You can click Unload beside OmniVoice TTS, or use Unload all + flush, and then generate again; VoxCPM2 should load on first use. Also check that OMNIVOICE_TTS_BACKEND is not set to omnivoice, because an environment variable overrides the engine selected in Settings. OmniVoice Studio’s VoxCPM2 backend loads openbmb/VoxCPM2 by default, and the upstream VoxCPM2 release states that its weights and inference tools are Apache-2.0 licensed, although commercial use should still be checked against the exact model card and applicable requirements.