Fixed
- Model dropdown now lists all three VoxCPM models — VoxCPM-0.5B, VoxCPM1.5, and VoxCPM2 all appear in the TTS and
Voice Clone node dropdowns. Previously only VoxCPM2 was shown due to a filter that excluded older model names.
Changed
- Voice Clone node adapts to V1 vs V2 models at runtime:
- VoxCPM2 (unchanged): supports both controllable cloning (reference audio only) and ultimate cloning (reference +
transcript). - VoxCPM-0.5B / VoxCPM1.5: requires a transcript (prompt_text or ASR) — controllable cloning is a V2-only feature. A
clear error message is shown if no transcript is provided with a V1 model.
- VoxCPM2 (unchanged): supports both controllable cloning (reference audio only) and ultimate cloning (reference +
- Denoiser is now correctly applied for V1 models during voice cloning.
- Node display names updated to "VoxCPM TTS" and "VoxCPM Voice Clone" to reflect multi-model support.