Views
No views yet
1import librosa
2import torch
3import soundfile as sf
4from torchaudio import transforms as T
5from neucodec import NeuCodec, NeuCodecOnnxDecoder
6
7model = NeuCodec.from_pretrained("neuphonic/neucodec")
8model.eval()
9compiled_model = NeuCodecOnnxDecoder.from_pretrained("neuphonic/neucodec-onnx-decoder-int8")
10
11y, sr = torchaudio.load(librosa.ex("libri1"))
12if sr != 16_000:
13 y = T.Resample(sr, 16_000)(y)[None, ...] # (B, 1, T_16)
14
15with torch.no_grad():
16 fsq_codes = model.encode_code(y)
17 # fsq_codes = model.encode_code(librosa.ex("libri1")) # or directly pass your filepath!
18 print(f"Codes shape: {fsq_codes.shape}")
19 recon = compiled_model.decode_code(fsq_codes) # (B, 1, T_24)
20
21sf.write("reconstructed.wav", recon, 24_000)