Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -5,12 +5,10 @@ import numpy as np
|
|
5 |
import tempfile
|
6 |
import os
|
7 |
import noisereduce as nr
|
8 |
-
import json
|
9 |
import torch
|
10 |
from demucs import pretrained
|
11 |
from demucs.apply import apply_model
|
12 |
import torchaudio
|
13 |
-
from pathlib import Path
|
14 |
import matplotlib.pyplot as plt
|
15 |
from io import BytesIO
|
16 |
from PIL import Image
|
@@ -19,9 +17,8 @@ import datetime
|
|
19 |
import librosa
|
20 |
import warnings
|
21 |
from faster_whisper import WhisperModel
|
22 |
-
from mutagen.mp3 import MP3
|
23 |
-
from mutagen.id3 import ID3, TIT2, TPE1, TALB, TYER
|
24 |
from TTS.api import TTS
|
|
|
25 |
import pickle
|
26 |
|
27 |
# Suppress warnings
|
@@ -58,17 +55,8 @@ def apply_reverb(audio):
|
|
58 |
def apply_pitch_shift(audio, semitones=-2):
|
59 |
new_frame_rate = int(audio.frame_rate * (2 ** (semitones / 12)))
|
60 |
samples = np.array(audio.get_array_of_samples())
|
61 |
-
resampled = np.interp(
|
62 |
-
|
63 |
-
np.arange(len(samples)),
|
64 |
-
samples
|
65 |
-
).astype(np.int16)
|
66 |
-
return AudioSegment(
|
67 |
-
resampled.tobytes(),
|
68 |
-
frame_rate=new_frame_rate,
|
69 |
-
sample_width=audio.sample_width,
|
70 |
-
channels=audio.channels
|
71 |
-
)
|
72 |
|
73 |
def apply_echo(audio, delay_ms=500, decay=0.5):
|
74 |
echo = audio - 10
|
@@ -138,41 +126,17 @@ def match_loudness(audio_path, target_lufs=-14.0):
|
|
138 |
adjusted.export(out_path, format="wav")
|
139 |
return out_path
|
140 |
|
141 |
-
# === AI Mastering Chain – Genre EQ + Loudness Match + Limiting ===
|
142 |
-
def ai_mastering_chain(audio_path, genre="Pop", target_lufs=-14.0):
|
143 |
-
audio = AudioSegment.from_file(audio_path)
|
144 |
-
|
145 |
-
# Apply Genre EQ
|
146 |
-
eq_audio = auto_eq(audio, genre=genre)
|
147 |
-
|
148 |
-
# Convert to numpy for loudness
|
149 |
-
samples, sr = audiosegment_to_array(eq_audio)
|
150 |
-
|
151 |
-
# Apply loudness normalization
|
152 |
-
meter = pyln.Meter(sr)
|
153 |
-
loudness = meter.integrated_loudness(samples.astype(np.float64) / 32768.0)
|
154 |
-
gain_db = target_lufs - loudness
|
155 |
-
final_audio = eq_audio + gain_db
|
156 |
-
|
157 |
-
# Apply final limiting
|
158 |
-
final_audio = apply_limiter(final_audio)
|
159 |
-
|
160 |
-
out_path = os.path.join(tempfile.gettempdir(), "mastered_output.wav")
|
161 |
-
final_audio.export(out_path, format="wav")
|
162 |
-
return out_path
|
163 |
-
|
164 |
# === Auto-EQ per Genre ===
|
165 |
def auto_eq(audio, genre="Pop"):
|
166 |
eq_map = {
|
167 |
-
"Pop": [(200, 500, -3), (2000, 4000, +4)],
|
168 |
-
"EDM": [(60, 250, +6), (8000, 12000, +3)],
|
169 |
-
"Rock": [(1000, 3000, +4), (7000, 10000, -3)],
|
170 |
-
"Hip-Hop": [(20, 100, +6), (7000, 10000, -4)],
|
171 |
-
"Acoustic": [(100, 300, -3), (4000, 8000, +2)],
|
172 |
-
"Metal": [(100, 500, -4), (2000, 5000, +6), (7000, 12000, -3)],
|
173 |
-
"Trap": [(80, 120, +6), (3000, 6000, -4)],
|
174 |
-
"LoFi": [(20, 200, +3), (1000, 3000, -2)]
|
175 |
-
"Default": []
|
176 |
}
|
177 |
|
178 |
from scipy.signal import butter, sosfilt
|
@@ -191,6 +155,23 @@ def auto_eq(audio, genre="Pop"):
|
|
191 |
|
192 |
return array_to_audiosegment(samples.astype(np.int16), sr, channels=audio.channels)
|
193 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
194 |
# === Harmonic Saturation / Exciter ===
|
195 |
def harmonic_saturation(audio, saturation_type="Tube", intensity=0.2):
|
196 |
samples = np.array(audio.get_array_of_samples()).astype(np.float32)
|
@@ -208,81 +189,6 @@ def harmonic_saturation(audio, saturation_type="Tube", intensity=0.2):
|
|
208 |
|
209 |
return array_to_audiosegment(saturated.astype(np.int16), audio.frame_rate, channels=audio.channels)
|
210 |
|
211 |
-
# === Vocal Isolation Helpers ===
|
212 |
-
def load_track_local(path, sample_rate, channels=2):
|
213 |
-
sig, rate = torchaudio.load(path)
|
214 |
-
if rate != sample_rate:
|
215 |
-
sig = torchaudio.functional.resample(sig, rate, sample_rate)
|
216 |
-
if channels == 1:
|
217 |
-
sig = sig.mean(0)
|
218 |
-
return sig
|
219 |
-
|
220 |
-
def save_track(path, wav, sample_rate):
|
221 |
-
path = Path(path)
|
222 |
-
torchaudio.save(str(path), wav, sample_rate)
|
223 |
-
|
224 |
-
def apply_vocal_isolation(audio_path):
|
225 |
-
model = pretrained.get_model(name='htdemucs')
|
226 |
-
wav = load_track_local(audio_path, model.samplerate, channels=2)
|
227 |
-
ref = wav.mean(0)
|
228 |
-
wav -= ref[:, None]
|
229 |
-
sources = apply_model(model, wav[None])[0]
|
230 |
-
wav += ref[:, None]
|
231 |
-
|
232 |
-
vocal_track = sources[3].cpu()
|
233 |
-
out_path = os.path.join(tempfile.gettempdir(), "vocals.wav")
|
234 |
-
save_track(out_path, vocal_track, model.samplerate)
|
235 |
-
return out_path
|
236 |
-
|
237 |
-
# === Stem Splitting (Drums, Bass, Other, Vocals) ===
|
238 |
-
def stem_split(audio_path):
|
239 |
-
model = pretrained.get_model(name='htdemucs')
|
240 |
-
wav = load_track_local(audio_path, model.samplerate, channels=2)
|
241 |
-
sources = apply_model(model, wav[None])[0]
|
242 |
-
|
243 |
-
output_dir = tempfile.mkdtemp()
|
244 |
-
stem_paths = []
|
245 |
-
|
246 |
-
for i, name in enumerate(['drums', 'bass', 'other', 'vocals']):
|
247 |
-
path = os.path.join(output_dir, f"{name}.wav")
|
248 |
-
save_track(path, sources[i].cpu(), model.samplerate)
|
249 |
-
stem_paths.append(gr.File(value=path))
|
250 |
-
|
251 |
-
return stem_paths
|
252 |
-
|
253 |
-
# === Save/Load Project File (.aiproj) ===
|
254 |
-
def save_project(vocals, drums, bass, other, vol_vocals, vol_drums, vol_bass, vol_other):
|
255 |
-
project_data = {
|
256 |
-
"vocals": AudioSegment.from_file(vocals).raw_data,
|
257 |
-
"drums": AudioSegment.from_file(drums).raw_data,
|
258 |
-
"bass": AudioSegment.from_file(bass).raw_data,
|
259 |
-
"other": AudioSegment.from_file(other).raw_data,
|
260 |
-
"volumes": {
|
261 |
-
"vocals": vol_vocals,
|
262 |
-
"drums": vol_drums,
|
263 |
-
"bass": vol_bass,
|
264 |
-
"other": vol_other
|
265 |
-
}
|
266 |
-
}
|
267 |
-
out_path = os.path.join(tempfile.gettempdir(), "mix_session.aiproj")
|
268 |
-
with open(out_path, "wb") as f:
|
269 |
-
pickle.dump(project_data, f)
|
270 |
-
return out_path
|
271 |
-
|
272 |
-
def load_project(project_file):
|
273 |
-
with open(project_file.name, "rb") as f:
|
274 |
-
data = pickle.load(f)
|
275 |
-
return (
|
276 |
-
array_to_audiosegment(data["vocals"], 44100),
|
277 |
-
array_to_audiosegment(data["drums"], 44100),
|
278 |
-
array_to_audiosegment(data["bass"], 44100),
|
279 |
-
array_to_audiosegment(data["other"], 44100),
|
280 |
-
data["volumes"]["vocals"],
|
281 |
-
data["volumes"]["drums"],
|
282 |
-
data["volumes"]["bass"],
|
283 |
-
data["volumes"]["other"]
|
284 |
-
)
|
285 |
-
|
286 |
# === Process Audio Function ===
|
287 |
def process_audio(audio_file, selected_effects, isolate_vocals, preset_name, export_format):
|
288 |
status = "🔊 Loading audio..."
|
@@ -302,7 +208,6 @@ def process_audio(audio_file, selected_effects, isolate_vocals, preset_name, exp
|
|
302 |
"Normalize": apply_normalize,
|
303 |
"Noise Gate": lambda x: apply_noise_gate(x, threshold=-50.0),
|
304 |
"Limiter": lambda x: apply_limiter(x, limit_dB=-1),
|
305 |
-
"Phaser": lambda x: apply_phaser(x),
|
306 |
"Flanger": lambda x: apply_phaser(x, rate=1.2, depth=0.9, mix=0.7),
|
307 |
"Bitcrusher": lambda x: apply_bitcrush(x, bit_depth=8),
|
308 |
"Auto Gain": lambda x: apply_auto_gain(x, target_dB=-20),
|
@@ -340,7 +245,7 @@ def process_audio(audio_file, selected_effects, isolate_vocals, preset_name, exp
|
|
340 |
status = f"❌ Error: {str(e)}"
|
341 |
return None, None, status, "", status
|
342 |
|
343 |
-
# === Waveform
|
344 |
def show_waveform(audio_file):
|
345 |
try:
|
346 |
audio = AudioSegment.from_file(audio_file)
|
@@ -364,7 +269,6 @@ def detect_genre(audio_path):
|
|
364 |
except Exception:
|
365 |
return "Unknown"
|
366 |
|
367 |
-
# === Session Info Export ===
|
368 |
def generate_session_log(audio_path, effects, isolate_vocals, export_format, genre):
|
369 |
log = {
|
370 |
"timestamp": str(datetime.datetime.now()),
|
@@ -396,34 +300,50 @@ preset_choices = {
|
|
396 |
|
397 |
preset_names = list(preset_choices.keys())
|
398 |
|
399 |
-
# ===
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
400 |
with gr.Blocks(title="AI Audio Studio", css="style.css") as demo:
|
401 |
-
gr.
|
|
|
402 |
|
403 |
-
# --- Single File Studio ---
|
404 |
with gr.Tab("🎵 Single File Studio"):
|
405 |
-
gr.
|
406 |
-
|
407 |
-
|
408 |
-
gr.
|
409 |
-
|
410 |
-
gr.
|
411 |
-
gr.Dropdown(choices=
|
412 |
-
gr.
|
413 |
-
|
414 |
-
|
415 |
-
gr.Audio(label="Processed Audio", type="filepath")
|
416 |
-
gr.Image(label="Waveform Preview")
|
417 |
-
gr.Textbox(label="
|
418 |
-
gr.Textbox(label="
|
419 |
-
|
420 |
-
|
421 |
-
|
422 |
-
|
423 |
-
flagging_mode="never",
|
424 |
-
submit_btn="Process Audio",
|
425 |
-
clear_btn=None
|
426 |
-
)
|
427 |
|
428 |
# --- AI Mastering Chain Tab ===
|
429 |
with gr.Tab("🎧 AI Mastering Chain"):
|
@@ -454,6 +374,44 @@ with gr.Blocks(title="AI Audio Studio", css="style.css") as demo:
|
|
454 |
description="Enhance clarity and presence using saturation styles like Tube or Tape."
|
455 |
)
|
456 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
457 |
# --- Remix Mode ---
|
458 |
with gr.Tab("🎛 Remix Mode"):
|
459 |
gr.Interface(
|
@@ -506,37 +464,14 @@ with gr.Blocks(title="AI Audio Studio", css="style.css") as demo:
|
|
506 |
description="Correct vocal pitch automatically"
|
507 |
)
|
508 |
|
509 |
-
# --- Create Karaoke Video from Audio + Lyrics ===
|
510 |
-
with gr.Tab("📹 Create Karaoke Video"):
|
511 |
-
gr.Interface(
|
512 |
-
fn=create_karaoke_video,
|
513 |
-
inputs=[
|
514 |
-
gr.Audio(label="Upload Track", type="filepath"),
|
515 |
-
gr.Textbox(label="Lyrics", lines=10),
|
516 |
-
gr.File(label="Background (Optional)")
|
517 |
-
],
|
518 |
-
outputs=gr.Video(label="Karaoke Video"),
|
519 |
-
title="Make Karaoke Videos from Audio + Lyrics",
|
520 |
-
description="Generate karaoke-style videos with real-time sync."
|
521 |
-
)
|
522 |
-
|
523 |
-
# --- Vocal Doubler / Harmonizer ===
|
524 |
-
with gr.Tab("🎧 Vocal Doubler / Harmonizer"):
|
525 |
-
gr.Interface(
|
526 |
-
fn=vocal_doubler,
|
527 |
-
inputs=gr.Audio(label="Upload Vocal Clip", type="filepath"),
|
528 |
-
outputs=gr.Audio(label="Doubled Output", type="filepath"),
|
529 |
-
title="Add Vocal Doubling / Harmony",
|
530 |
-
description="Enhance vocals with doubling or harmony"
|
531 |
-
)
|
532 |
-
|
533 |
# --- Real-Time Spectrum Analyzer + Live EQ Preview ===
|
534 |
-
with gr.Tab("📊
|
535 |
gr.Interface(
|
536 |
fn=visualize_spectrum,
|
537 |
inputs=gr.Audio(label="Upload Track", type="filepath"),
|
538 |
outputs=gr.Image(label="Spectrum Analysis"),
|
539 |
-
title="
|
|
|
540 |
)
|
541 |
|
542 |
# --- Loudness Graph Tab ===
|
@@ -567,40 +502,28 @@ with gr.Blocks(title="AI Audio Studio", css="style.css") as demo:
|
|
567 |
)
|
568 |
|
569 |
# --- Save/Load Mix Session (.aiproj) ===
|
570 |
-
with gr.Tab("📁 Save/Load
|
571 |
gr.Interface(
|
572 |
fn=save_project,
|
573 |
inputs=[
|
574 |
-
gr.File(label="
|
575 |
-
gr.
|
576 |
-
gr.
|
577 |
-
gr.File(label="Other"),
|
578 |
-
gr.Slider(minimum=-10, maximum=10, value=0, label="Vocals Volume"),
|
579 |
-
gr.Slider(minimum=-10, maximum=10, value=0, label="Drums Volume"),
|
580 |
-
gr.Slider(minimum=-10, maximum=10, value=0, label="Bass Volume"),
|
581 |
-
gr.Slider(minimum=-10, maximum=10, value=0, label="Other Volume")
|
582 |
],
|
583 |
outputs=gr.File(label="Project File (.aiproj)"),
|
584 |
-
title="Save
|
585 |
-
description="Save
|
586 |
)
|
587 |
|
588 |
gr.Interface(
|
589 |
fn=load_project,
|
590 |
inputs=gr.File(label="Upload .aiproj File"),
|
591 |
outputs=[
|
592 |
-
gr.
|
593 |
-
gr.
|
594 |
-
gr.File(label="Bass"),
|
595 |
-
gr.File(label="Other"),
|
596 |
-
gr.Slider(label="Vocals Volume"),
|
597 |
-
gr.Slider(label="Drums Volume"),
|
598 |
-
gr.Slider(label="Bass Volume"),
|
599 |
-
gr.Slider(label="Other Volume")
|
600 |
],
|
601 |
-
title="Resume Last
|
602 |
-
description="Load saved
|
603 |
-
allow_flagging="never"
|
604 |
)
|
605 |
|
606 |
# --- Prompt-Based Editing Tab ===
|
@@ -651,144 +574,4 @@ with gr.Blocks(title="AI Audio Studio", css="style.css") as demo:
|
|
651 |
allow_flagging="never"
|
652 |
)
|
653 |
|
654 |
-
# --- Vocal Pitch Correction – Auto-Tune Style ===
|
655 |
-
def auto_tune_vocal(audio_path, target_key="C"):
|
656 |
-
try:
|
657 |
-
# Placeholder for real-time pitch detection
|
658 |
-
return apply_pitch_shift(AudioSegment.from_file(audio_path), 0.2)
|
659 |
-
except Exception as e:
|
660 |
-
return None
|
661 |
-
|
662 |
-
# --- Create Karaoke Video from Audio + Lyrics ===
|
663 |
-
def create_karaoke_video(audio_path, lyrics, bg_image=None):
|
664 |
-
try:
|
665 |
-
from moviepy.editor import TextClip, CompositeVideoClip, ColorClip, AudioFileClip
|
666 |
-
|
667 |
-
audio = AudioFileClip(audio_path)
|
668 |
-
video = ColorClip(size=(1280, 720), color=(0, 0, 0), duration=audio.duration_seconds)
|
669 |
-
words = [(word.strip(), i * 3, (i+1)*3) for i, word in enumerate(lyrics.split())]
|
670 |
-
|
671 |
-
text_clips = [
|
672 |
-
TextClip(word, fontsize=60, color='white').set_position('center').set_duration(end - start).set_start(start)
|
673 |
-
for word, start, end in words
|
674 |
-
]
|
675 |
-
|
676 |
-
final_video = CompositeVideoClip([video] + text_clips).set_audio(audio)
|
677 |
-
out_path = os.path.join(tempfile.gettempdir(), "karaoke.mp4")
|
678 |
-
final_video.write_videofile(out_path, codec="libx264", audio_codec="aac")
|
679 |
-
return out_path
|
680 |
-
except Exception as e:
|
681 |
-
return f"⚠️ Failed: {str(e)}"
|
682 |
-
|
683 |
-
# --- Vocal Doubler / Harmonizer ===
|
684 |
-
def vocal_doubler(audio):
|
685 |
-
shifted_up = apply_pitch_shift(audio, 0.3)
|
686 |
-
shifted_down = apply_pitch_shift(audio, -0.3)
|
687 |
-
return audio.overlay(shifted_up).overlay(shifted_down)
|
688 |
-
|
689 |
-
# --- AI Suggest Preset Based on Genre ===
|
690 |
-
def suggest_preset_by_genre(audio_path):
|
691 |
-
try:
|
692 |
-
y, sr = torchaudio.load(audio_path)
|
693 |
-
mfccs = librosa.feature.mfcc(y=y.numpy().flatten(), sr=sr, n_mfcc=13).mean(axis=1).reshape(1, -1)
|
694 |
-
return ["Vocal Clarity", "Limiter", "Stereo Expansion"]
|
695 |
-
except Exception:
|
696 |
-
return ["Default"]
|
697 |
-
|
698 |
-
# --- AI Suggest Preset Based on Genre ===
|
699 |
-
with gr.Tab("🧠 AI Suggest Preset"):
|
700 |
-
gr.Interface(
|
701 |
-
fn=suggest_preset_by_genre,
|
702 |
-
inputs=gr.Audio(label="Upload Track", type="filepath"),
|
703 |
-
outputs=gr.Dropdown(choices=preset_names, label="Recommended Preset"),
|
704 |
-
title="Let AI Recommend Best Preset",
|
705 |
-
description="Upload a track and let AI recommend the best preset based on genre."
|
706 |
-
)
|
707 |
-
|
708 |
-
# --- Prompt-Based Editing ===
|
709 |
-
def process_prompt(audio_path, prompt):
|
710 |
-
audio = AudioSegment.from_file(audio_path)
|
711 |
-
|
712 |
-
if "noise" in prompt.lower() or "clean" in prompt.lower():
|
713 |
-
audio = apply_noise_reduction(audio)
|
714 |
-
|
715 |
-
if "normalize" in prompt.lower() or "loud" in prompt.lower():
|
716 |
-
audio = apply_normalize(audio)
|
717 |
-
|
718 |
-
if "bass" in prompt.lower() and ("boost" in prompt.lower()):
|
719 |
-
audio = apply_bass_boost(audio)
|
720 |
-
|
721 |
-
if "treble" in prompt.lower() or "high" in prompt.lower():
|
722 |
-
audio = apply_treble_boost(audio)
|
723 |
-
|
724 |
-
if "echo" in prompt.lower() or "reverb" in prompt.lower():
|
725 |
-
audio = apply_reverb(audio)
|
726 |
-
|
727 |
-
if "pitch" in prompt.lower() and "correct" in prompt.lower():
|
728 |
-
audio = apply_pitch_correction(audio)
|
729 |
-
|
730 |
-
if "harmony" in prompt.lower() or "double" in prompt.lower():
|
731 |
-
audio = apply_harmony(audio)
|
732 |
-
|
733 |
-
out_path = os.path.join(tempfile.gettempdir(), "prompt_output.wav")
|
734 |
-
audio.export(out_path, format="wav")
|
735 |
-
return out_path
|
736 |
-
|
737 |
-
# --- Prompt-Based Editing Tab ===
|
738 |
-
with gr.Tab("🧠 Prompt-Based Editing"):
|
739 |
-
gr.Interface(
|
740 |
-
fn=process_prompt,
|
741 |
-
inputs=[
|
742 |
-
gr.File(label="Upload Audio", type="filepath"),
|
743 |
-
gr.Textbox(label="Describe What You Want", lines=5)
|
744 |
-
],
|
745 |
-
outputs=gr.Audio(label="Edited Output", type="filepath"),
|
746 |
-
title="Type Your Edits – AI Does the Rest",
|
747 |
-
description="Say what you want done and let AI handle it.",
|
748 |
-
allow_flagging="never"
|
749 |
-
)
|
750 |
-
|
751 |
-
# --- Vocal Pitch Correction (Auto-Tune) ===
|
752 |
-
def apply_pitch_correction(audio, target_key="C"):
|
753 |
-
return apply_pitch_shift(audio, 0.2)
|
754 |
-
|
755 |
-
with gr.Tab("🧬 Vocal Pitch Correction"):
|
756 |
-
gr.Interface(
|
757 |
-
fn=auto_tune_vocal,
|
758 |
-
inputs=[
|
759 |
-
gr.File(label="Source Voice Clip"),
|
760 |
-
gr.Textbox(label="Target Key", value="C", lines=1)
|
761 |
-
],
|
762 |
-
outputs=gr.Audio(label="Pitch-Corrected Output", type="filepath"),
|
763 |
-
title="Auto-Tune Style Pitch Correction",
|
764 |
-
description="Correct vocal pitch automatically"
|
765 |
-
)
|
766 |
-
|
767 |
-
# --- Real-Time Spectrum Analyzer + EQ Preview ===
|
768 |
-
def visualize_spectrum(audio_path):
|
769 |
-
y, sr = torchaudio.load(audio_path)
|
770 |
-
y_np = y.numpy().flatten()
|
771 |
-
stft = librosa.stft(y_np)
|
772 |
-
db = librosa.amplitude_to_db(abs(stft))
|
773 |
-
|
774 |
-
plt.figure(figsize=(10, 4))
|
775 |
-
img = librosa.display.specshow(db, sr=sr, x_axis="time", y_axis="hz", cmap="magma")
|
776 |
-
plt.colorbar(img, format="%+2.0f dB")
|
777 |
-
plt.title("Frequency Spectrum")
|
778 |
-
plt.tight_layout()
|
779 |
-
buf = BytesIO()
|
780 |
-
plt.savefig(buf, format="png")
|
781 |
-
plt.close()
|
782 |
-
buf.seek(0)
|
783 |
-
return Image.open(buf)
|
784 |
-
|
785 |
-
with gr.Tab("📊 Frequency Spectrum"):
|
786 |
-
gr.Interface(
|
787 |
-
fn=visualize_spectrum,
|
788 |
-
inputs=gr.Audio(label="Upload Track", type="filepath"),
|
789 |
-
outputs=gr.Image(label="Spectrum Analysis"),
|
790 |
-
title="Real-Time Spectrum Analyzer",
|
791 |
-
description="See the frequency breakdown of your audio"
|
792 |
-
)
|
793 |
-
|
794 |
demo.launch()
|
|
|
5 |
import tempfile
|
6 |
import os
|
7 |
import noisereduce as nr
|
|
|
8 |
import torch
|
9 |
from demucs import pretrained
|
10 |
from demucs.apply import apply_model
|
11 |
import torchaudio
|
|
|
12 |
import matplotlib.pyplot as plt
|
13 |
from io import BytesIO
|
14 |
from PIL import Image
|
|
|
17 |
import librosa
|
18 |
import warnings
|
19 |
from faster_whisper import WhisperModel
|
|
|
|
|
20 |
from TTS.api import TTS
|
21 |
+
import base64
|
22 |
import pickle
|
23 |
|
24 |
# Suppress warnings
|
|
|
55 |
def apply_pitch_shift(audio, semitones=-2):
|
56 |
new_frame_rate = int(audio.frame_rate * (2 ** (semitones / 12)))
|
57 |
samples = np.array(audio.get_array_of_samples())
|
58 |
+
resampled = np.interp(np.arange(0, len(samples), 2 ** (semitones / 12)), np.arange(len(samples)), samples).astype(np.int16)
|
59 |
+
return AudioSegment(resampled.tobytes(), frame_rate=new_frame_rate, sample_width=audio.sample_width, channels=audio.channels)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
60 |
|
61 |
def apply_echo(audio, delay_ms=500, decay=0.5):
|
62 |
echo = audio - 10
|
|
|
126 |
adjusted.export(out_path, format="wav")
|
127 |
return out_path
|
128 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
129 |
# === Auto-EQ per Genre ===
|
130 |
def auto_eq(audio, genre="Pop"):
|
131 |
eq_map = {
|
132 |
+
"Pop": [(200, 500, -3), (2000, 4000, +4)],
|
133 |
+
"EDM": [(60, 250, +6), (8000, 12000, +3)],
|
134 |
+
"Rock": [(1000, 3000, +4), (7000, 10000, -3)],
|
135 |
+
"Hip-Hop": [(20, 100, +6), (7000, 10000, -4)],
|
136 |
+
"Acoustic": [(100, 300, -3), (4000, 8000, +2)],
|
137 |
+
"Metal": [(100, 500, -4), (2000, 5000, +6), (7000, 12000, -3)],
|
138 |
+
"Trap": [(80, 120, +6), (3000, 6000, -4)],
|
139 |
+
"LoFi": [(20, 200, +3), (1000, 3000, -2)]
|
|
|
140 |
}
|
141 |
|
142 |
from scipy.signal import butter, sosfilt
|
|
|
155 |
|
156 |
return array_to_audiosegment(samples.astype(np.int16), sr, channels=audio.channels)
|
157 |
|
158 |
+
# === AI Mastering Chain – Genre EQ + Loudness Match + Limiting ===
|
159 |
+
def ai_mastering_chain(audio_path, genre="Pop", target_lufs=-14.0):
|
160 |
+
audio = AudioSegment.from_file(audio_path)
|
161 |
+
eq_audio = auto_eq(audio, genre=genre)
|
162 |
+
samples, sr = audiosegment_to_array(eq_audio)
|
163 |
+
|
164 |
+
# Apply loudness normalization
|
165 |
+
meter = pyln.Meter(sr)
|
166 |
+
loudness = meter.integrated_loudness(samples.astype(np.float64) / 32768.0)
|
167 |
+
gain_db = target_lufs - loudness
|
168 |
+
final_audio = eq_audio + gain_db
|
169 |
+
final_audio = apply_limiter(final_audio)
|
170 |
+
|
171 |
+
out_path = os.path.join(tempfile.gettempdir(), "mastered_output.wav")
|
172 |
+
final_audio.export(out_path, format="wav")
|
173 |
+
return out_path
|
174 |
+
|
175 |
# === Harmonic Saturation / Exciter ===
|
176 |
def harmonic_saturation(audio, saturation_type="Tube", intensity=0.2):
|
177 |
samples = np.array(audio.get_array_of_samples()).astype(np.float32)
|
|
|
189 |
|
190 |
return array_to_audiosegment(saturated.astype(np.int16), audio.frame_rate, channels=audio.channels)
|
191 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
192 |
# === Process Audio Function ===
|
193 |
def process_audio(audio_file, selected_effects, isolate_vocals, preset_name, export_format):
|
194 |
status = "🔊 Loading audio..."
|
|
|
208 |
"Normalize": apply_normalize,
|
209 |
"Noise Gate": lambda x: apply_noise_gate(x, threshold=-50.0),
|
210 |
"Limiter": lambda x: apply_limiter(x, limit_dB=-1),
|
|
|
211 |
"Flanger": lambda x: apply_phaser(x, rate=1.2, depth=0.9, mix=0.7),
|
212 |
"Bitcrusher": lambda x: apply_bitcrush(x, bit_depth=8),
|
213 |
"Auto Gain": lambda x: apply_auto_gain(x, target_dB=-20),
|
|
|
245 |
status = f"❌ Error: {str(e)}"
|
246 |
return None, None, status, "", status
|
247 |
|
248 |
+
# === Visualize Waveform ===
|
249 |
def show_waveform(audio_file):
|
250 |
try:
|
251 |
audio = AudioSegment.from_file(audio_file)
|
|
|
269 |
except Exception:
|
270 |
return "Unknown"
|
271 |
|
|
|
272 |
def generate_session_log(audio_path, effects, isolate_vocals, export_format, genre):
|
273 |
log = {
|
274 |
"timestamp": str(datetime.datetime.now()),
|
|
|
300 |
|
301 |
preset_names = list(preset_choices.keys())
|
302 |
|
303 |
+
# === Preset Cards Gallery ===
|
304 |
+
def get_preset_cards():
|
305 |
+
card_paths = []
|
306 |
+
for name in preset_names:
|
307 |
+
card_paths.append(f"https://via.placeholder.com/150x100?text={name}")
|
308 |
+
return card_paths
|
309 |
+
|
310 |
+
# === Load Preset by Name ===
|
311 |
+
def load_preset_by_card(name_index):
|
312 |
+
name = preset_names[name_index]
|
313 |
+
return name, preset_choices[name]
|
314 |
+
|
315 |
+
# === Logo Embedding (Base64 or file) ===
|
316 |
+
def get_logo():
|
317 |
+
try:
|
318 |
+
with open("logo.png", "rb") as img_file:
|
319 |
+
return "data:image/png;base64," + base64.b64encode(img_file.read()).decode()
|
320 |
+
except FileNotFoundError:
|
321 |
+
return "https://via.placeholder.com/400x100?text=AI+Audio+Studio"
|
322 |
+
|
323 |
+
# === Main UI ===
|
324 |
with gr.Blocks(title="AI Audio Studio", css="style.css") as demo:
|
325 |
+
gr.HTML(f'<div class="studio-header"><img src="{get_logo()}" width="400" /></div>')
|
326 |
+
gr.Markdown("### Upload, edit, export — powered by AI!")
|
327 |
|
|
|
328 |
with gr.Tab("🎵 Single File Studio"):
|
329 |
+
with gr.Row():
|
330 |
+
with gr.Column(min_width=300):
|
331 |
+
input_audio = gr.Audio(label="Upload Audio", type="filepath")
|
332 |
+
effect_checkbox = gr.CheckboxGroup(choices=preset_choices.get("Default", []),
|
333 |
+
label="Apply Effects in Order")
|
334 |
+
preset_dropdown = gr.Dropdown(choices=preset_names, label="Select Preset", value=preset_names[0])
|
335 |
+
export_format = gr.Dropdown(choices=["MP3", "WAV"], label="Export Format", value="MP3")
|
336 |
+
isolate_vocals = gr.Checkbox(label="Isolate Vocals After Effects")
|
337 |
+
submit_btn = gr.Button("Process Audio")
|
338 |
+
with gr.Column(min_width=300):
|
339 |
+
output_audio = gr.Audio(label="Processed Audio", type="filepath")
|
340 |
+
waveform_img = gr.Image(label="Waveform Preview")
|
341 |
+
genre_out = gr.Textbox(label="Detected Genre")
|
342 |
+
status_box = gr.Textbox(label="Status", value="✅ Ready", lines=1)
|
343 |
+
|
344 |
+
submit_btn.click(fn=process_audio, inputs=[
|
345 |
+
input_audio, effect_checkbox, isolate_vocals, preset_dropdown, export_format
|
346 |
+
], outputs=[output_audio, waveform_img, _, genre_out, status_box])
|
|
|
|
|
|
|
|
|
347 |
|
348 |
# --- AI Mastering Chain Tab ===
|
349 |
with gr.Tab("🎧 AI Mastering Chain"):
|
|
|
374 |
description="Enhance clarity and presence using saturation styles like Tube or Tape."
|
375 |
)
|
376 |
|
377 |
+
# --- Preset Cards Gallery ===
|
378 |
+
with gr.Tab("🎛 Preset Gallery"):
|
379 |
+
gr.Markdown("### Select a preset visually")
|
380 |
+
|
381 |
+
preset_images = [
|
382 |
+
("https://via.placeholder.com/150x100?text=Pop", "Pop"),
|
383 |
+
("https://via.placeholder.com/150x100?text=EDM", "EDM"),
|
384 |
+
("https://via.placeholder.com/150x100?text=Rock", "Rock"),
|
385 |
+
("https://via.placeholder.com/150x100?text=Hip-Hop", "Hip-Hop"),
|
386 |
+
("https://via.placeholder.com/150x100?text=Acoustic", "Acoustic"),
|
387 |
+
("https://via.placeholder.com/150x100?text=Tube+Saturation", "Tube"),
|
388 |
+
("https://via.placeholder.com/150x100?text=Stage+Mode", "Stage Mode"),
|
389 |
+
("https://via.placeholder.com/150x100?text=Vocal+Distortion", "Vocal Distortion")
|
390 |
+
]
|
391 |
+
|
392 |
+
preset_gallery = gr.Gallery(value=preset_images, label="Preset Cards", columns=4, height="auto")
|
393 |
+
preset_name_out = gr.Dropdown(choices=preset_names, label="Selected Preset")
|
394 |
+
preset_effects_out = gr.CheckboxGroup(choices=[e for e in preset_choices["Default"]], label="Effects")
|
395 |
+
|
396 |
+
def select_preset(evt: gr.SelectData):
|
397 |
+
selected = evt.index
|
398 |
+
name = preset_names[selected % len(preset_names)]
|
399 |
+
effects = preset_choices.get(name, [])
|
400 |
+
return name, effects
|
401 |
+
|
402 |
+
preset_gallery.select(fn=select_preset, inputs=[], outputs=[preset_name_out, preset_effects_out])
|
403 |
+
|
404 |
+
# --- Vocal Doubler / Harmonizer ===
|
405 |
+
with gr.Tab("🎧 Vocal Doubler / Harmonizer"):
|
406 |
+
gr.Interface(
|
407 |
+
fn=lambda x: apply_harmony(x),
|
408 |
+
inputs=gr.Audio(label="Upload Vocal Clip", type="filepath"),
|
409 |
+
outputs=gr.Audio(label="Doubled Output", type="filepath"),
|
410 |
+
title="Add Vocal Doubling / Harmony",
|
411 |
+
description="Enhance vocals with doubling or harmony",
|
412 |
+
allow_flagging="never"
|
413 |
+
)
|
414 |
+
|
415 |
# --- Remix Mode ---
|
416 |
with gr.Tab("🎛 Remix Mode"):
|
417 |
gr.Interface(
|
|
|
464 |
description="Correct vocal pitch automatically"
|
465 |
)
|
466 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
467 |
# --- Real-Time Spectrum Analyzer + Live EQ Preview ===
|
468 |
+
with gr.Tab("📊 Frequency Spectrum"):
|
469 |
gr.Interface(
|
470 |
fn=visualize_spectrum,
|
471 |
inputs=gr.Audio(label="Upload Track", type="filepath"),
|
472 |
outputs=gr.Image(label="Spectrum Analysis"),
|
473 |
+
title="Real-Time Spectrum Analyzer",
|
474 |
+
description="See the frequency breakdown of your audio"
|
475 |
)
|
476 |
|
477 |
# --- Loudness Graph Tab ===
|
|
|
502 |
)
|
503 |
|
504 |
# --- Save/Load Mix Session (.aiproj) ===
|
505 |
+
with gr.Tab("📁 Save/Load Project"):
|
506 |
gr.Interface(
|
507 |
fn=save_project,
|
508 |
inputs=[
|
509 |
+
gr.File(label="Original Audio"),
|
510 |
+
gr.Dropdown(choices=preset_names, label="Used Preset", value=preset_names[0]),
|
511 |
+
gr.CheckboxGroup(choices=[e for e in preset_choices.get("Default", [])], label="Applied Effects")
|
|
|
|
|
|
|
|
|
|
|
512 |
],
|
513 |
outputs=gr.File(label="Project File (.aiproj)"),
|
514 |
+
title="Save Everything Together",
|
515 |
+
description="Save your session, effects, and settings in one file to reuse later."
|
516 |
)
|
517 |
|
518 |
gr.Interface(
|
519 |
fn=load_project,
|
520 |
inputs=gr.File(label="Upload .aiproj File"),
|
521 |
outputs=[
|
522 |
+
gr.Dropdown(choices=preset_names, label="Loaded Preset"),
|
523 |
+
gr.CheckboxGroup(choices=[e for e in preset_choices.get("Default", [])], label="Loaded Effects")
|
|
|
|
|
|
|
|
|
|
|
|
|
524 |
],
|
525 |
+
title="Resume Last Project",
|
526 |
+
description="Load your saved session"
|
|
|
527 |
)
|
528 |
|
529 |
# --- Prompt-Based Editing Tab ===
|
|
|
574 |
allow_flagging="never"
|
575 |
)
|
576 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
577 |
demo.launch()
|