Spaces:
Running
Running
ageraustine
commited on
Commit
•
86b5b34
1
Parent(s):
38546ed
mixings with scipy
Browse files
app.py
CHANGED
@@ -4,6 +4,11 @@ import numpy as np
|
|
4 |
import os
|
5 |
import base64
|
6 |
from streamlit import session_state as st_state
|
|
|
|
|
|
|
|
|
|
|
7 |
|
8 |
# Try to get API_URL from environment variables, if not found set to a default value
|
9 |
try:
|
@@ -33,7 +38,15 @@ if 'augmented_audio' not in st_state:
|
|
33 |
|
34 |
# Streamlit app title
|
35 |
st.title("Songlabai")
|
36 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
37 |
uploaded_file = st.file_uploader("Upload Music File", type=["mp3", "wav", "ogg", "flac", "aac"])
|
38 |
|
39 |
genres = [
|
@@ -67,61 +80,107 @@ if st.button("Generate Audio"):
|
|
67 |
sample_rate = response.json()[0]['sample_rate']
|
68 |
st.audio(st_state.audio, format="audio/wav", sample_rate=sample_rate, start_time=0)
|
69 |
|
|
|
|
|
|
|
70 |
# Post-processing options
|
71 |
st.header("Post-processing Options")
|
72 |
-
|
73 |
-
|
74 |
-
|
75 |
-
if
|
76 |
-
|
77 |
-
|
78 |
-
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
83 |
|
84 |
# Reverb
|
85 |
-
st.
|
86 |
-
|
87 |
-
if
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
|
93 |
-
|
94 |
-
|
95 |
-
|
96 |
-
|
97 |
-
|
98 |
-
|
99 |
-
|
100 |
-
|
101 |
-
|
102 |
-
|
103 |
-
st_state.augmented_audio =
|
104 |
-
|
105 |
-
|
106 |
-
|
107 |
-
|
108 |
-
|
109 |
-
|
110 |
-
|
111 |
-
|
112 |
-
|
113 |
-
|
114 |
-
if pitch_shift:
|
115 |
-
# Pitch shift using the Fourier shift method
|
116 |
-
pitch_shift_factor = 2 ** (pitch_semitones / 12)
|
117 |
-
st_state.augmented_audio = scipy.signal.resample(st_state.augmented_audio, int(len(st_state.augmented_audio) / pitch_shift_factor))
|
118 |
-
|
119 |
-
if add_reverb:
|
120 |
-
# Apply reverb using the convolution method
|
121 |
-
reverb_ir = scipy.signal.exponential(reverb_room_scale, reverb_damping, reverb_wet_only)
|
122 |
-
st_state.augmented_audio = scipy.signal.fftconvolve(st_state.augmented_audio, reverb_ir)
|
123 |
-
|
124 |
-
# Play the processed audio
|
125 |
st.audio(st_state.augmented_audio, format="audio/wav", sample_rate=sample_rate, start_time=0)
|
126 |
-
|
127 |
st.link_button("Download/Save", "https://songlabai.com/subcribe/")
|
|
|
4 |
import os
|
5 |
import base64
|
6 |
from streamlit import session_state as st_state
|
7 |
+
import librosa
|
8 |
+
import soundfile as sf
|
9 |
+
from scipy.signal import butter, sosfilt
|
10 |
+
|
11 |
+
|
12 |
|
13 |
# Try to get API_URL from environment variables, if not found set to a default value
|
14 |
try:
|
|
|
38 |
|
39 |
# Streamlit app title
|
40 |
st.title("Songlabai")
|
41 |
+
|
42 |
+
# Initialize session state variables
|
43 |
+
if 'vocal_audio' not in st_state:
|
44 |
+
st_state.vocal_audio = None
|
45 |
+
if 'audio' not in st_state:
|
46 |
+
st_state.audio = None
|
47 |
+
if 'augmented_audio' not in st_state:
|
48 |
+
st_state.augmented_audio = None
|
49 |
+
|
50 |
uploaded_file = st.file_uploader("Upload Music File", type=["mp3", "wav", "ogg", "flac", "aac"])
|
51 |
|
52 |
genres = [
|
|
|
80 |
sample_rate = response.json()[0]['sample_rate']
|
81 |
st.audio(st_state.audio, format="audio/wav", sample_rate=sample_rate, start_time=0)
|
82 |
|
83 |
+
|
84 |
+
vocal_file = st.file_uploader("Upload Vocal File", type=["mp3", "wav", "ogg", "flac", "aac"])
|
85 |
+
|
86 |
# Post-processing options
|
87 |
st.header("Post-processing Options")
|
88 |
+
|
89 |
+
# Mixing
|
90 |
+
mix_vocals = st.checkbox("Mix Vocals")
|
91 |
+
if mix_vocals and st_state.vocal_audio is not None:
|
92 |
+
# Load the vocal audio
|
93 |
+
vocal_audio, _ = librosa.load(vocal_file, sr=sample_rate, mono=False)
|
94 |
+
|
95 |
+
# Adjust the vocal audio length to match the generated audio
|
96 |
+
vocal_audio = librosa.util.fix_length(vocal_audio, len(st_state.audio))
|
97 |
+
|
98 |
+
# Mix the vocal audio with the generated audio
|
99 |
+
st_state.augmented_audio = (st_state.audio + vocal_audio) / 2
|
100 |
+
|
101 |
+
# Mastering
|
102 |
+
st.subheader("Mastering")
|
103 |
+
|
104 |
+
# Volume Balance
|
105 |
+
volume_balance = st.slider("Volume Balance", min_value=-10.0, max_value=10.0, value=0.0, step=0.1)
|
106 |
+
if st.button("Apply Volume Balance"):
|
107 |
+
if st_state.augmented_audio is None:
|
108 |
+
st_state.augmented_audio = st_state.audio
|
109 |
+
st_state.augmented_audio *= 10 ** (volume_balance / 20)
|
110 |
+
|
111 |
+
# Compression
|
112 |
+
compression_ratio = st.slider("Compression Ratio", min_value=1.0, max_value=10.0, value=3.0, step=0.1)
|
113 |
+
if st.button("Apply Compression"):
|
114 |
+
if st_state.augmented_audio is None:
|
115 |
+
st_state.augmented_audio = st_state.audio
|
116 |
+
|
117 |
+
# Apply compression using a simple soft-knee compressor
|
118 |
+
threshold = -20 # dBFS
|
119 |
+
ratio = compression_ratio
|
120 |
+
knee = 10 # dB
|
121 |
+
max_gain = 20 # dB
|
122 |
+
|
123 |
+
def compress(x, threshold, ratio, knee, max_gain):
|
124 |
+
over = np.maximum(x - threshold, 0)
|
125 |
+
gain = over / (over + knee) * (1 - (1 / ratio)) + 1
|
126 |
+
gain = np.maximum(gain, 1 - max_gain)
|
127 |
+
return x * gain
|
128 |
+
|
129 |
+
st_state.augmented_audio = compress(st_state.augmented_audio, threshold, ratio, knee, max_gain)
|
130 |
+
|
131 |
+
# EQ
|
132 |
+
eq_low = st.slider("EQ Low", min_value=-10.0, max_value=10.0, value=0.0, step=0.1)
|
133 |
+
eq_mid = st.slider("EQ Mid", min_value=-10.0, max_value=10.0, value=0.0, step=0.1)
|
134 |
+
eq_high = st.slider("EQ High", min_value=-10.0, max_value=10.0, value=0.0, step=0.1)
|
135 |
+
|
136 |
+
if st.button("Apply EQ"):
|
137 |
+
if st_state.augmented_audio is None:
|
138 |
+
st_state.augmented_audio = st_state.audio
|
139 |
+
|
140 |
+
# Apply a simple 3-band EQ using a butterworth filter
|
141 |
+
nyquist = sample_rate / 2
|
142 |
+
low_cutoff = 200 / nyquist
|
143 |
+
mid_cutoff = 2000 / nyquist
|
144 |
+
high_cutoff = 8000 / nyquist
|
145 |
+
|
146 |
+
low_sos = butter(4, low_cutoff, btype='low', output='sos', analog=False)
|
147 |
+
mid_sos = butter(4, [low_cutoff, mid_cutoff], btype='band', output='sos', analog=False)
|
148 |
+
high_sos = butter(4, high_cutoff, btype='high', output='sos', analog=False)
|
149 |
+
|
150 |
+
st_state.augmented_audio = sosfilt(np.dstack((low_sos, mid_sos, high_sos)),
|
151 |
+
st_state.augmented_audio,
|
152 |
+
np.stack((eq_low, eq_mid, eq_high)))
|
153 |
|
154 |
# Reverb
|
155 |
+
reverb_amount = st.slider("Reverb Amount", min_value=0.0, max_value=1.0, value=0.5, step=0.01)
|
156 |
+
if st.button("Apply Reverb"):
|
157 |
+
if st_state.augmented_audio is None:
|
158 |
+
st_state.augmented_audio = st_state.audio
|
159 |
+
|
160 |
+
# Apply a simple reverb effect using convolution
|
161 |
+
ir_length = int(sample_rate * 2.5) # 2.5 seconds
|
162 |
+
ir = np.zeros(ir_length)
|
163 |
+
ir[0] = 1
|
164 |
+
ir = np.append(ir, np.zeros(len(st_state.augmented_audio) - ir_length))
|
165 |
+
reverb = np.convolve(st_state.augmented_audio, ir, mode='full')[:len(st_state.augmented_audio)]
|
166 |
+
st_state.augmented_audio = st_state.augmented_audio + reverb_amount * reverb
|
167 |
+
|
168 |
+
# Delay
|
169 |
+
delay_amount = st.slider("Delay Amount", min_value=0.0, max_value=1.0, value=0.5, step=0.01)
|
170 |
+
delay_time = st.slider("Delay Time (ms)", min_value=10, max_value=500, value=100, step=10)
|
171 |
+
if st.button("Apply Delay"):
|
172 |
+
if st_state.augmented_audio is None:
|
173 |
+
st_state.augmented_audio = st_state.audio
|
174 |
+
|
175 |
+
# Apply a simple delay effect
|
176 |
+
delay_samples = int(delay_time / 1000 * sample_rate)
|
177 |
+
delay = np.zeros(len(st_state.augmented_audio) + delay_samples)
|
178 |
+
delay[delay_samples:] = st_state.augmented_audio
|
179 |
+
delay[:len(st_state.augmented_audio)] += delay_amount * delay[:-delay_samples]
|
180 |
+
st_state.augmented_audio = delay[:len(st_state.augmented_audio)]
|
181 |
+
|
182 |
+
# Display the final audio
|
183 |
+
if st_state.augmented_audio is not None:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
184 |
st.audio(st_state.augmented_audio, format="audio/wav", sample_rate=sample_rate, start_time=0)
|
185 |
+
|
186 |
st.link_button("Download/Save", "https://songlabai.com/subcribe/")
|