Spaces:
Running
Running
File size: 2,905 Bytes
183b9ba 68647bf 183b9ba 68647bf 183b9ba 68647bf 183b9ba afe296a 183b9ba 68647bf 183b9ba d2d6202 68647bf d2d6202 183b9ba |
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 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 |
import gradio as gr
import torch
from torchaudio.backend.common import AudioMetaData
from df.enhance import enhance, load_audio, save_audio
from df.io import resample
from libdf import DF
from df.model import ModelParams
from df import config
import moviepy.editor as mp
import numpy as np
try:
config.load('config.ini')
except Exception as e:
print(e)
DEVICE = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
p = ModelParams()
df = DF(
sr=p.sr,
fft_size=p.fft_size,
hop_size=p.hop_size,
nb_bands=p.nb_erb,
min_nb_erb_freqs=p.min_nb_freqs,
)
print("Device - ", DEVICE)
model = torch.load(("model.pth"), map_location=torch.device('cpu'))
model.to(DEVICE)
model.eval()
def identity(video_path):
print(video_path)
# audio = mp.AudioFileClip(x)
# wav_file = x
# audio.write_audiofile(wav_file)
video = mp.VideoFileClip(video_path)
audio = video.audio
wav_file = "tmp.wav"
audio.write_audiofile(wav_file)
print("Wav stored.")
meta = AudioMetaData(-1, -1, -1, -1, "")
sr = config("sr", 48000, int, section="df")
sample, meta = load_audio(wav_file, sr)
len_audio = (meta.num_frames/meta.sample_rate)/60
max_min = 1
if len_audio % max_min < 0.1:
num_chunks = len_audio // max_min
else:
num_chunks = len_audio // max_min + 1
if len_audio and not num_chunks:
num_chunks += 1
print(f"Total length of audio = {len_audio} chunks = {num_chunks}")
estimate = []
split_tensors = torch.tensor_split(sample, int(num_chunks), dim = 1)
for i in range(len(split_tensors)):
enhanced = enhance(model, df, split_tensors[i])
enhanced = enhance(model, df, enhanced)
lim = torch.linspace(0.0, 1.0, int(sr * 0.15)).unsqueeze(0)
lim = torch.cat((lim, torch.ones(1, enhanced.shape[1] - lim.shape[1])), dim=1)
enhanced = enhanced * lim
enhanced = resample(enhanced, sr, meta.sample_rate)
estimate.append(enhanced)
estimate = tuple(estimate)
enhanced = torch.cat(estimate, dim = -1)
sr = meta.sample_rate
save_audio("enhanced_aud.wav", enhanced, sr)
audio = mp.AudioFileClip('enhanced_aud.wav')
video = mp.VideoFileClip(video_path)
final_video = video.set_audio(audio)
final_video.write_videofile("output_video.mp4",
codec='libx264',
audio_codec='aac',
temp_audiofile='temp-audio.m4a',
remove_temp=True
)
return "output_video.mp4"
demo = gr.Interface(
fn=identity,
title="AudioPure - THE BEST AUDIO DENOISER",
description="AudioPure is the only platform you need for removing all kinds of background noise from your videos!!",
examples=[['audiopure_og.mov']],
cache_examples=True,
inputs=gr.Video(label="Input Video", sources="upload"),
outputs=gr.Video(label="Output Video"),
)
demo.launch() |