...

lib/__init__.py

@@ -7,8 +7,9 @@ from .models import (
 
 from .conf import (
     FULL_DOCKER, NATIVE, audiobooks_cli_dir, audiobooks_gradio_dir,
-    audiobooks_host_dir, debug_mode, default_audio_proc_format, default_device,
-    default_gpu_wiki, default_output_format, device_list, ebook_formats,
+    audiobooks_host_dir, debug_mode, default_audio_proc_samplerate, 
+    default_audio_proc_format, default_device, default_gpu_wiki, 
+    default_output_format, device_list, ebook_formats,
     ebooks_dir, interface_component_options, interface_concurrency_limit,
     interface_host, interface_port, interface_shared_tmp_expire,
     max_python_version, min_python_version, models_dir, os,
@@ -36,9 +37,9 @@ __all__ = [
 
     # from conf
     "FULL_DOCKER", "NATIVE", "audiobooks_cli_dir", "audiobooks_gradio_dir",
-    "audiobooks_host_dir", "debug_mode", "default_audio_proc_format",
-    "default_device", "default_gpu_wiki", "default_output_format",
-    "device_list", "ebook_formats", "ebooks_dir",
+    "audiobooks_host_dir", "debug_mode", "default_audio_proc_samplerate",
+    "default_audio_proc_format", "default_device", "default_gpu_wiki",
+    "default_output_format", "device_list", "ebook_formats", "ebooks_dir",
     "interface_component_options", "interface_concurrency_limit",
     "interface_host", "interface_port", "interface_shared_tmp_expire",
     "max_python_version", "min_python_version", "models_dir", "os",

lib/classes/tts_engines/coqui.py

@@ -306,7 +306,7 @@ class Coqui:
                                 new_voice_path = re.sub(r'([\\/])eng([\\/])', rf'\1{lang_dir}\2', voice_path)
                                 proc_voice_path = new_voice_path.replace('.wav', '_temp.wav')
                                 torchaudio.save(proc_voice_path, audio_tensor, default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate'], format='wav')
-                                if normalize_audio(proc_voice_path, new_voice_path, 24000):
+                                if normalize_audio(proc_voice_path, new_voice_path, default_audio_proc_samplerate):
                                     del audio_data, sourceTensor, audio_tensor  
                                     if self.session['tts_engine'] != TTS_ENGINES['XTTSv2']:
                                         del tts

lib/classes/voice_extractor.py

@@ -10,7 +10,7 @@ from io import BytesIO
 from pydub import AudioSegment, silence
 from pydub.silence import detect_silence
 
-from lib.conf import voice_formats
+from lib.conf import voice_formats, default_audio_proc_samplerate
 from lib.models import TTS_ENGINES, models
 from lib.classes.background_detector import BackgroundDetector
 
@@ -204,8 +204,9 @@ class VoiceExtractor:
             raise ValueError(error)
 
     def _normalize_audio(self):
-        try:                 
-            process_file = os.path.join(self.session['voice_dir'], f'{self.voice_name}.wav')
+        try:
+            rate = default_audio_proc_samplerate
+            process_file = os.path.join(self.session['voice_dir'], f'{self.voice_name}_proc.wav')
             ffmpeg_cmd = [shutil.which('ffmpeg'), '-hide_banner', '-nostats', '-i', self.voice_track]
             filter_complex = (
                 'agate=threshold=-25dB:ratio=1.4:attack=10:release=250,'
@@ -226,34 +227,33 @@ class VoiceExtractor:
                 '-y', process_file
             ]
             error = None
-            for rate in ['16000', '24000']:
-                ffmpeg_cmd[-3] = rate
-                output_file = re.sub(r'\.wav$', f'_{rate}.wav', process_file)
-                ffmpeg_cmd[-1] = output_file
-                try:
-                    process = subprocess.Popen(
-                        ffmpeg_cmd,
-                        env={},
-                        stdout=subprocess.PIPE,
-                        stderr=subprocess.STDOUT,
-                        text=True,
-                        universal_newlines=True,
-                        encoding='utf-8'
-                    )
-                    for line in process.stdout:
-                        print(line, end='')  # Print each line of stdout
-                    process.wait()
-                    if process.returncode != 0:
-                        error = f'_normalize_audio(): process.returncode: {process.returncode}'
-                        break
-                    elif not os.path.exists(output_file) or os.path.getsize(output_file) == 0:
-                        error = f'_normalize_audio() error: {output_file} was not created or is empty.'
-                        break
-                    else:
-                        self.final_files.append(output_file)
-                except subprocess.CalledProcessError as e:
-                    error = f'_normalize_audio() ffmpeg.Error: {e.stderr.decode()}'
+            ffmpeg_cmd[-3] = rate
+            output_file = re.sub(r'_proc\.wav$', f'.wav', process_file)
+            ffmpeg_cmd[-1] = output_file
+            try:
+                process = subprocess.Popen(
+                    ffmpeg_cmd,
+                    env={},
+                    stdout=subprocess.PIPE,
+                    stderr=subprocess.STDOUT,
+                    text=True,
+                    universal_newlines=True,
+                    encoding='utf-8'
+                )
+                for line in process.stdout:
+                    print(line, end='')  # Print each line of stdout
+                process.wait()
+                if process.returncode != 0:
+                    error = f'_normalize_audio(): process.returncode: {process.returncode}'
                     break
+                elif not os.path.exists(output_file) or os.path.getsize(output_file) == 0:
+                    error = f'_normalize_audio() error: {output_file} was not created or is empty.'
+                    break
+                else:
+                    self.final_files.append(output_file)
+            except subprocess.CalledProcessError as e:
+                error = f'_normalize_audio() ffmpeg.Error: {e.stderr.decode()}'
+                break
             shutil.rmtree(self.demucs_dir, ignore_errors=True)
             if os.path.exists(process_file):
                 os.remove(process_file)

lib/conf.py

@@ -71,6 +71,7 @@ audiobooks_cli_dir = os.path.abspath(os.path.join('audiobooks','cli'))
 ebook_formats = ['.epub', '.mobi', '.azw3', '.fb2', '.lrf', '.rb', '.snb', '.tcr', '.pdf', '.txt', '.rtf', '.doc', '.docx', '.html', '.odt', '.azw'] # Add or remove the format you accept as input
 voice_formats = ['.mp4', '.m4b', '.m4a', '.mp3', '.wav', '.aac', '.flac', '.alac', '.ogg', '.aiff', '.aif', '.wma', '.dsd', '.opus', '.pcmu', '.pcma', '.gsm'] # Add or remove the format you accept as input
 output_formats = ['aac', 'flac', 'mp3', 'm4b', 'm4a', 'mp4', 'mov', 'ogg', 'wav', 'webm']
+default_audio_proc_samplerate = '24000'
 default_audio_proc_format = 'flac' # or 'mp3', 'aac', 'm4a', 'm4b', 'amr', '3gp', 'alac'. 'wav' format is ok but limited to process files < 4GB
 default_output_format = 'm4b'