Spaces:
Running
Running
File size: 1,285 Bytes
6c4aae6 eb75e68 6c4aae6 eb75e68 6c4aae6 eb75e68 6c4aae6 eb75e68 6c4aae6 eb75e68 6c4aae6 ab6d95c 6c4aae6 4887a61 eb75e68 6c4aae6 eb75e68 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 |
from logging import getLogger
import gradio as gr
import torch
import torchaudio
from denoisers import WaveUNetModel
LOGGER = getLogger(__name__)
MODEL = WaveUNetModel.from_pretrained("wrice/waveunet-vctk-24khz")
def denoise(inputs):
sr, audio = inputs
audio = torch.from_numpy(audio)[None]
audio = audio / 32768.0
LOGGER.info(f"Audio shape: {audio.shape}")
LOGGER.info(f"Sample rate: {sr}")
if sr != MODEL.config.sample_rate:
audio = torchaudio.functional.resample(audio, sr, MODEL.config.sample_rate)
chunk_size = MODEL.config.max_length
padding = abs(audio.size(-1) % chunk_size - chunk_size)
padded = torch.nn.functional.pad(audio, (0, padding))
clean = []
for i in range(0, padded.shape[-1], chunk_size):
audio_chunk = padded[:, i : i + chunk_size]
with torch.no_grad():
clean_chunk = MODEL(audio_chunk[None]).logits
clean.append(clean_chunk.squeeze(0))
denoised = torch.concat(clean)[:, : audio.shape[-1]].squeeze().clamp(-1.0, 1.0)
denoised = (denoised * 32767.0).numpy().astype("int16")
LOGGER.info(f"Denoised shape: {denoised.shape}")
return MODEL.config.sample_rate, denoised
iface = gr.Interface(fn=denoise, inputs="audio", outputs="audio")
iface.launch()
|