batikse/voxcpm

1import soundfile as sf2from voxcpm import VoxCPM3 4model = VoxCPM.from_pretrained("openbmb/VoxCPM-0.5B")5 6wav = model.generate(7    text="VoxCPM is an innovative end-to-end TTS model from ModelBest, designed to generate highly expressive speech.",8    prompt_wav_path=None,      # optional: path to a prompt speech for voice cloning9    prompt_text=None,          # optional: reference text10    cfg_value=2.0,             # LM guidance on LocDiT, higher for better adherence to the prompt, but maybe worse11    inference_timesteps=10,   # LocDiT inference timesteps, higher for better result, lower for fast speed12    normalize=True,           # enable external TN tool13    denoise=True,             # enable external Denoise tool14    retry_badcase=True,        # enable retrying mode for some bad cases (unstoppable)15    retry_badcase_max_times=3,  # maximum retrying times16    retry_badcase_ratio_threshold=6.0, # maximum length restriction for bad case detection (simple but effective), it could be adjusted for slow pace speech17)18 19sf.write("output.wav", wav, 16000)20print("saved: output.wav")