tee342 commited on
Commit
ba1c3bd
ยท
verified ยท
1 Parent(s): a415232

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +53 -101
app.py CHANGED
@@ -23,42 +23,11 @@ import base64
23
  import pickle
24
  import json
25
  import soundfile as sf
26
- import wave
27
- import struct
28
 
29
  # Suppress warnings
30
  warnings.filterwarnings("ignore")
31
 
32
- # === Global Effect Map ===
33
- effect_map = {
34
- "Noise Reduction": "apply_noise_reduction",
35
- "Compress Dynamic Range": "apply_compression",
36
- "Add Reverb": "apply_reverb",
37
- "Pitch Shift": "apply_pitch_shift",
38
- "Echo": "apply_echo",
39
- "Stereo Widening": "apply_stereo_widen",
40
- "Bass Boost": "apply_bass_boost",
41
- "Treble Boost": "apply_treble_boost",
42
- "Normalize": "apply_normalize",
43
- "Limiter": "apply_limiter",
44
- "Auto Gain": "apply_auto_gain",
45
- "Vocal Distortion": "apply_vocal_distortion",
46
- "Stage Mode": "apply_stage_mode"
47
- }
48
-
49
- # === Helper Functions ===
50
- def audiosegment_to_array(audio):
51
- return np.array(audio.get_array_of_samples()), audio.frame_rate
52
-
53
- def array_to_audiosegment(samples, frame_rate, channels=1):
54
- return AudioSegment(
55
- samples.tobytes(),
56
- frame_rate=int(frame_rate),
57
- sample_width=samples.dtype.itemsize,
58
- channels=channels
59
- )
60
-
61
- # === Effect Functions ===
62
  def apply_normalize(audio):
63
  return audio.normalize()
64
 
@@ -124,6 +93,18 @@ def apply_bitcrush(audio, bit_depth=8):
124
  downsampled = np.round(samples / (32768 / max_val)).astype(np.int16)
125
  return array_to_audiosegment(downsampled, audio.frame_rate // 2, channels=audio.channels)
126
 
 
 
 
 
 
 
 
 
 
 
 
 
127
  # === Loudness Matching (EBU R128) ===
128
  try:
129
  import pyloudnorm as pyln
@@ -185,7 +166,7 @@ def auto_eq(audio, genre="Pop"):
185
 
186
  return array_to_audiosegment(samples.astype(np.int16), sr, channels=audio.channels)
187
 
188
- # === Vocal Isolation Helpers ===
189
  def load_track_local(path, sample_rate, channels=2):
190
  sig, rate = torchaudio.load(path)
191
  if rate != sample_rate:
@@ -198,6 +179,7 @@ def save_track(path, wav, sample_rate):
198
  path = Path(path)
199
  torchaudio.save(str(path), wav, sample_rate)
200
 
 
201
  def apply_vocal_isolation(audio_path):
202
  model = pretrained.get_model(name='htdemucs')
203
  wav = load_track_local(audio_path, model.samplerate, channels=2)
@@ -211,6 +193,22 @@ def apply_vocal_isolation(audio_path):
211
  save_track(out_path, vocal_track, model.samplerate)
212
  return out_path
213
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
214
  # === Process Audio Function โ€“ Fully Featured ===
215
  def process_audio(audio_file, selected_effects, isolate_vocals, preset_name, export_format):
216
  status = "๐Ÿ”Š Loading audio..."
@@ -256,7 +254,7 @@ def process_audio(audio_file, selected_effects, isolate_vocals, preset_name, exp
256
 
257
  waveform_image = show_waveform(output_path)
258
  genre = detect_genre(output_path)
259
- session_log = generate_session_log(audio_file, selected_effects, isolate_vocals, export_format, genre)
260
 
261
  status = "๐ŸŽ‰ Done!"
262
  return output_path, waveform_image, session_log, genre, status, history
@@ -437,7 +435,7 @@ def live_effect_preview(stream, selected_effect):
437
  except Exception as e:
438
  return None
439
 
440
- # === Voice Command Parsing (No Extra Libraries Needed) ===
441
  def voice_to_effect(command):
442
  command = command.lower()
443
  if "noise reduction" in command or "remove noise" in command:
@@ -488,13 +486,17 @@ def harmonic_saturation(audio, saturation_type="Tube", intensity=0.2):
488
 
489
  return array_to_audiosegment(saturated.astype(np.int16), audio.frame_rate, channels=audio.channels)
490
 
491
- # --- Multi-track Mixing Console โ€“ Now Working ===
492
- def mix_tracks(track1_path, track2_path, vol1, vol2):
493
- seg1 = AudioSegment.from_file(track1_path) * vol1
494
- seg2 = AudioSegment.from_file(track2_path) * vol2
495
- mixed = seg1.overlay(seg2)
496
- out_path = os.path.join(tempfile.gettempdir(), "mixed_output.wav")
497
- mixed.export(out_path, format="wav")
 
 
 
 
498
  return out_path
499
 
500
  # === Save/Load Mix Session (.aiproj) โ€“ Added Back ===
@@ -518,30 +520,9 @@ def load_project(project_file):
518
  def process_prompt(audio, prompt):
519
  return apply_noise_reduction(audio)
520
 
521
- # === A/B Compare Two Tracks ===
522
- def compare_ab(track1_path, track2_path):
523
- return track1_path, track2_path
524
-
525
- # === Loop Section ===
526
- def loop_section(audio_path, start_ms, end_ms, loops=2):
527
- audio = AudioSegment.from_file(audio_path)
528
- section = audio[start_ms:end_ms]
529
- looped = section * loops
530
- out_path = os.path.join(tempfile.gettempdir(), "looped_output.wav")
531
- looped.export(out_path, format="wav")
532
- return out_path
533
-
534
- # === Share Effect Chain ===
535
- def share_effect_chain(selected_effects):
536
- return json.dumps(selected_effects)
537
-
538
- def load_effect_chain(chain_json):
539
- return json.loads(chain_json)
540
-
541
  # === Vocal Formant Correction (AI Voice Tuning) ===
542
  def formant_correct(audio, shift=1.0):
543
  samples, sr = audiosegment_to_array(audio)
544
- # Simulate formant shifting using librosa
545
  corrected = librosa.effects.pitch_shift(samples, sr=sr, n_steps=shift)
546
  return array_to_audiosegment(corrected.astype(np.int16), sr, channels=audio.channels)
547
 
@@ -733,7 +714,7 @@ with gr.Blocks(css="""
733
  output_other = gr.Audio(label="Processed Other", type="filepath")
734
  btn_other.click(fn=apply_effects_to_stem, inputs=[other_path, other_effects], outputs=output_other)
735
 
736
- # --- AI Remastering Tab โ€“ New Feature (Without SpeechBrain) ===
737
  with gr.Tab("๐Ÿ”ฎ AI Remastering"):
738
  gr.Interface(
739
  fn=ai_remaster,
@@ -975,8 +956,7 @@ with gr.Blocks(css="""
975
  gr.Audio(label="Version B", type="filepath")
976
  ],
977
  title="Compare Two Versions",
978
- description="Hear two mixes side-by-side",
979
- allow_flagging="never"
980
  )
981
 
982
  # --- Loop Section ===
@@ -991,7 +971,7 @@ with gr.Blocks(css="""
991
  ],
992
  outputs=gr.Audio(label="Looped Output", type="filepath"),
993
  title="Repeat a Section",
994
- description="Useful for editing a specific part"
995
  )
996
 
997
  # --- Share Effect Chain ===
@@ -1001,7 +981,7 @@ with gr.Blocks(css="""
1001
  inputs=gr.CheckboxGroup(choices=preset_choices["Default"]),
1002
  outputs=gr.Textbox(label="Share Code", lines=2),
1003
  title="Copy/Paste Effect Chain",
1004
- description="Share your setup with others via link/code"
1005
  )
1006
 
1007
  with gr.Tab("๐Ÿ“ฅ Load Shared Chain"):
@@ -1019,14 +999,14 @@ with gr.Blocks(css="""
1019
  ### Keyboard Controls
1020
  - `Ctrl + Z`: Undo last effect
1021
  - `Ctrl + Y`: Redo
1022
- - `Spacebar`: Play/Stop playback
1023
- - `Ctrl + S`: Save current session
1024
  - `Ctrl + O`: Open session
1025
  - `Ctrl + C`: Copy effect chain
1026
  - `Ctrl + V`: Paste effect chain
1027
  - `Ctrl + Shift + L`: Loop selection
1028
- - `Ctrl + Shift + P`: Export to PDF report
1029
- - `Ctrl + Shift + M`: Export full mix ZIP
1030
  """)
1031
 
1032
  # --- Vocal Formant Correction ===
@@ -1039,35 +1019,7 @@ with gr.Blocks(css="""
1039
  ],
1040
  outputs=gr.Audio(label="Natural-Sounding Vocal", type="filepath"),
1041
  title="Preserve Vocal Quality During Pitch Shift",
1042
- description="Make pitch-shifted vocals sound more human"
1043
  )
1044
 
1045
- # --- Effect History โ€“ Undo/Redo Stack ===
1046
- effect_history = []
1047
-
1048
- def undo_last_effect(history):
1049
- if len(history) > 1:
1050
- return history[-2]
1051
- return history[0]
1052
-
1053
- def redo_last_effect(history):
1054
- return history[-1]
1055
-
1056
- with gr.Tab("๐Ÿ”„ Undo/Redo"):
1057
- history_state = gr.State([])
1058
- with gr.Row():
1059
- input_audio_undo = gr.Audio(label="Upload Track", type="filepath")
1060
- effect_checkbox_undo = gr.CheckboxGroup(choices=preset_choices["Default"], label="Apply Effects")
1061
- process_btn_undo = gr.Button("Apply")
1062
- undo_btn = gr.Button("Undo Last")
1063
- redo_btn = gr.Button("Redo Last")
1064
- output_audio_undo = gr.Audio(label="Processed Audio", type="filepath")
1065
-
1066
- process_btn_undo.click(fn=lambda a, e: process_audio(a, e, False, "Default", "WAV")[0],
1067
- inputs=[input_audio_undo, effect_checkbox_undo],
1068
- outputs=[output_audio_undo, history_state])
1069
-
1070
- undo_btn.click(fn=undo_last_effect, inputs=history_state, outputs=output_audio_undo)
1071
- redo_btn.click(fn=redo_last_effect, inputs=history_state, outputs=output_audio_undo)
1072
-
1073
  demo.launch()
 
23
  import pickle
24
  import json
25
  import soundfile as sf
 
 
26
 
27
  # Suppress warnings
28
  warnings.filterwarnings("ignore")
29
 
30
+ # === Effect Map (Global) ===
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
31
  def apply_normalize(audio):
32
  return audio.normalize()
33
 
 
93
  downsampled = np.round(samples / (32768 / max_val)).astype(np.int16)
94
  return array_to_audiosegment(downsampled, audio.frame_rate // 2, channels=audio.channels)
95
 
96
+ # === Helper Functions ===
97
+ def audiosegment_to_array(audio):
98
+ return np.array(audio.get_array_of_samples()), audio.frame_rate
99
+
100
+ def array_to_audiosegment(samples, frame_rate, channels=1):
101
+ return AudioSegment(
102
+ samples.tobytes(),
103
+ frame_rate=int(frame_rate),
104
+ sample_width=samples.dtype.itemsize,
105
+ channels=channels
106
+ )
107
+
108
  # === Loudness Matching (EBU R128) ===
109
  try:
110
  import pyloudnorm as pyln
 
166
 
167
  return array_to_audiosegment(samples.astype(np.int16), sr, channels=audio.channels)
168
 
169
+ # === Load Track Helpers ===
170
  def load_track_local(path, sample_rate, channels=2):
171
  sig, rate = torchaudio.load(path)
172
  if rate != sample_rate:
 
179
  path = Path(path)
180
  torchaudio.save(str(path), wav, sample_rate)
181
 
182
+ # === Vocal Isolation Helpers ===
183
  def apply_vocal_isolation(audio_path):
184
  model = pretrained.get_model(name='htdemucs')
185
  wav = load_track_local(audio_path, model.samplerate, channels=2)
 
193
  save_track(out_path, vocal_track, model.samplerate)
194
  return out_path
195
 
196
+ # === Stem Splitting Function (Now Added Back!) ===
197
+ def stem_split(audio_path):
198
+ model = pretrained.get_model(name='htdemucs')
199
+ wav = load_track_local(audio_path, model.samplerate, channels=2)
200
+ sources = apply_model(model, wav[None])[0]
201
+
202
+ output_dir = tempfile.mkdtemp()
203
+ stem_paths = []
204
+
205
+ for i, name in enumerate(['drums', 'bass', 'other', 'vocals']):
206
+ path = os.path.join(output_dir, f"{name}.wav")
207
+ save_track(path, sources[i].cpu(), model.samplerate)
208
+ stem_paths.append(gr.File(value=path))
209
+
210
+ return stem_paths
211
+
212
  # === Process Audio Function โ€“ Fully Featured ===
213
  def process_audio(audio_file, selected_effects, isolate_vocals, preset_name, export_format):
214
  status = "๐Ÿ”Š Loading audio..."
 
254
 
255
  waveform_image = show_waveform(output_path)
256
  genre = detect_genre(output_path)
257
+ session_log = generate_session_log(audio_file, effects_to_apply, isolate_vocals, export_format, genre)
258
 
259
  status = "๐ŸŽ‰ Done!"
260
  return output_path, waveform_image, session_log, genre, status, history
 
435
  except Exception as e:
436
  return None
437
 
438
+ # === Voice Command Parsing ===
439
  def voice_to_effect(command):
440
  command = command.lower()
441
  if "noise reduction" in command or "remove noise" in command:
 
486
 
487
  return array_to_audiosegment(saturated.astype(np.int16), audio.frame_rate, channels=audio.channels)
488
 
489
+ # === A/B Compare Two Tracks ===
490
+ def compare_ab(track1_path, track2_path):
491
+ return track1_path, track2_path
492
+
493
+ # === Loop Section ===
494
+ def loop_section(audio_path, start_ms, end_ms, loops=2):
495
+ audio = AudioSegment.from_file(audio_path)
496
+ section = audio[start_ms:end_ms]
497
+ looped = section * loops
498
+ out_path = os.path.join(tempfile.gettempdir(), "looped_output.wav")
499
+ looped.export(out_path, format="wav")
500
  return out_path
501
 
502
  # === Save/Load Mix Session (.aiproj) โ€“ Added Back ===
 
520
  def process_prompt(audio, prompt):
521
  return apply_noise_reduction(audio)
522
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
523
  # === Vocal Formant Correction (AI Voice Tuning) ===
524
  def formant_correct(audio, shift=1.0):
525
  samples, sr = audiosegment_to_array(audio)
 
526
  corrected = librosa.effects.pitch_shift(samples, sr=sr, n_steps=shift)
527
  return array_to_audiosegment(corrected.astype(np.int16), sr, channels=audio.channels)
528
 
 
714
  output_other = gr.Audio(label="Processed Other", type="filepath")
715
  btn_other.click(fn=apply_effects_to_stem, inputs=[other_path, other_effects], outputs=output_other)
716
 
717
+ # --- AI Remastering Tab โ€“ New Feature (Fixed) ===
718
  with gr.Tab("๐Ÿ”ฎ AI Remastering"):
719
  gr.Interface(
720
  fn=ai_remaster,
 
956
  gr.Audio(label="Version B", type="filepath")
957
  ],
958
  title="Compare Two Versions",
959
+ description="Hear two mixes side-by-side"
 
960
  )
961
 
962
  # --- Loop Section ===
 
971
  ],
972
  outputs=gr.Audio(label="Looped Output", type="filepath"),
973
  title="Repeat a Section",
974
+ description="Useful for editing specific parts"
975
  )
976
 
977
  # --- Share Effect Chain ===
 
981
  inputs=gr.CheckboxGroup(choices=preset_choices["Default"]),
982
  outputs=gr.Textbox(label="Share Code", lines=2),
983
  title="Copy/Paste Effect Chain",
984
+ description="Share your setup via link/code"
985
  )
986
 
987
  with gr.Tab("๐Ÿ“ฅ Load Shared Chain"):
 
999
  ### Keyboard Controls
1000
  - `Ctrl + Z`: Undo last effect
1001
  - `Ctrl + Y`: Redo
1002
+ - `Spacebar`: Play/Stop
1003
+ - `Ctrl + S`: Save session
1004
  - `Ctrl + O`: Open session
1005
  - `Ctrl + C`: Copy effect chain
1006
  - `Ctrl + V`: Paste effect chain
1007
  - `Ctrl + Shift + L`: Loop selection
1008
+ - `Ctrl + Shift + P`: Export PDF report
1009
+ - `Ctrl + Shift + M`: Export full mix ZIP
1010
  """)
1011
 
1012
  # --- Vocal Formant Correction ===
 
1019
  ],
1020
  outputs=gr.Audio(label="Natural-Sounding Vocal", type="filepath"),
1021
  title="Preserve Vocal Quality During Pitch Shift",
1022
+ description="Make pitch-shifted vocals sound more natural"
1023
  )
1024
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1025
  demo.launch()