标准化响度

api.py

@@ -2,6 +2,7 @@ import asyncio
 import aiohttp
 import io
 import os
+from utils import normalize_audio_loudness
 
 BASE_URL = os.getenv("BASE_URL")
 AUDIO_URL = os.getenv("AUDIO_URL")
@@ -16,6 +17,7 @@ async def generate_api(voice_ids, text):
                     audio_data = await response.read()
                     # print(type(audio_data))
                     # 创建一个字节流对象
+                    audio_data = normalize_audio_loudness(audio_data)
                     return audio_data
                 else:
                     print(response)

i18n/emotion.json

@@ -1,46 +1,122 @@
 {
     "zh": {
-        "正常": "正常",
-        "生气": "生气",
-        "悲伤": "悲伤",
-        "惊讶": "惊讶",
-        "害怕": "害怕",
-        "厌恶": "厌恶",
-        "开心": "开心",
-        "失望": "失望",
-        "紧张": "紧张"
+      "正常": "正常",
+      "生气": "生气",
+      "悲伤": "悲伤",
+      "惊讶": "惊讶",
+      "害怕": "害怕",
+      "厌恶": "厌恶",
+      "开心": "开心",
+      "失望": "失望",
+      "紧张": "紧张",
+      "不屑": "不屑",
+      "不甘": "不甘",
+      "愉悦": "愉悦",
+      "戏谑": "戏谑",
+      "疑惑": "疑惑",
+      "尴尬": "尴尬",
+      "伤心": "伤心",
+      "困倦": "困倦",
+      "无奈": "无奈",
+      "警惕": "警惕",
+      "坚决": "坚决",
+      "委屈": "委屈",
+      "害羞": "害羞",
+      "傲娇": "傲娇",
+      "女主": "女主",
+      "平淡": "平淡",
+      "疑问": "疑问",
+      "男主": "男主",
+      "警惕 ": "警惕 "
     },
     "en": {
-        "正常": "normal",
-        "生气": "angry",
-        "悲伤": "sad",
-        "惊讶": "surprise",
-        "害怕": "fear",
-        "厌恶": "disgust",
-        "开心": "happy",
-        "失望": "disappointment",
-        "紧张": "nervous"
+      "正常": "normal",
+      "生气": "angry",
+      "悲伤": "sad",
+      "惊讶": "surprise",
+      "害怕": "fear",
+      "厌恶": "disgust",
+      "开心": "happy",
+      "失望": "disappointment",
+      "紧张": "nervous",
+      "女主": "female",
+      "平淡": "indifferent",
+      "疑问": "question",
+      "戏谑": "playful",
+      "困倦": "sleepy",
+      "男主": "male",
+      "疑惑": "puzzled",
+      "无奈": "helpless",
+      "不甘": "unwilling",
+      "傲娇": "tsundere",
+      "警惕": "vigilant",
+      "害羞": "shy",
+      "警惕 ": "vigilant ",
+      "尴尬": "embarrassed",
+      "坚决": "determined",
+      "不屑": "disdainful",
+      "愉悦": "joyful",
+      "委屈": "aggrieved",
+      "伤心": "heartbroken"
     },
     "ja": {
-        "正常": "正常",
-        "生气": "怒り",
-        "悲伤": "悲しい",
-        "惊讶": "驚き",
-        "害怕": "恐れ",
-        "厌恶": "嫌悪",
-        "开心": "嬉しい",
-        "失望": "失望",
-        "紧张": "緊張"
+      "正常": "正常",
+      "生气": "怒り",
+      "悲伤": "悲しい",
+      "惊讶": "驚き",
+      "害怕": "恐れ",
+      "厌恶": "嫌悪",
+      "开心": "嬉しい",
+      "失望": "失望",
+      "紧张": "緊張",
+      "女主": "ヒロイン",
+      "平淡": "平淡",
+      "疑问": "疑問",
+      "戏谑": "からかい",
+      "困倦": "眠い",
+      "男主": "主人公",
+      "疑惑": "困惑",
+      "无奈": "仕方ない",
+      "不甘": "悔しい",
+      "傲娇": "ツンデレ",
+      "警惕": "警戒",
+      "害羞": "恥ずかしい",
+      "警惕 ": "警戒 ",
+      "尴尬": "気まずい",
+      "坚决": "断固とした",
+      "不屑": "軽蔑",
+      "愉悦": "喜び",
+      "委屈": "悔しい",
+      "伤心": "悲しい"
     },
     "ko": {
-        "正常": "보통",
-        "生气": "화남",
-        "悲伤": "슬픔",
-        "惊讶": "놀람",
-        "害怕": "두려움",
-        "厌恶": "혐오",
-        "开心": "행복",
-        "失望": "실망",
-        "紧张": "긴장"
+      "正常": "보통",
+      "生气": "화남",
+      "悲伤": "슬픔",
+      "惊讶": "놀람",
+      "害怕": "두려움",
+      "厌恶": "혐오",
+      "开心": "행복",
+      "失望": "실망",
+      "紧张": "긴장",
+      "女主": "여주인공",
+      "平淡": "평범함",
+      "疑问": "의문",
+      "戏谑": "농담",
+      "困倦": "피곤함",
+      "男主": "남주인공",
+      "疑惑": "의혹",
+      "无奈": "어쩔 수 없음",
+      "不甘": "불만",
+      "傲娇": "츤데레",
+      "警惕": "경계",
+      "害羞": "부끄러움",
+      "警惕 ": "경계 ",
+      "尴尬": "당황",
+      "坚决": "단호함",
+      "不屑": "경멸",
+      "愉悦": "즐거움",
+      "委屈": "억울함",
+      "伤心": "상심"
     }
-}
+  }

requirements.txt

@@ -1,4 +1,7 @@
 gradio
 aiohttp
 pypinyin
-gradio-i18n==0.0.10
+gradio-i18n==0.0.10
+pyloudnorm
+soundfile
+pydub

test.py

@@ -0,0 +1,8 @@
+from main import demo
+
+if __name__ == "__main__":
+    demo.launch(
+        server_port=8080,
+        show_api=False,
+        share=True
+    )

utils.py

@@ -0,0 +1,41 @@
+import io
+import os
+import pickle
+
+import soundfile as sf
+import numpy as np
+from pydub import AudioSegment
+from pyloudnorm import Meter
+
+os.chdir(os.path.dirname(os.path.abspath(__file__)))
+
+def normalize_audio_loudness(data: bytes, target_loudness: float = -23.0) -> bytes:
+    audio = AudioSegment.from_file(io.BytesIO(data), format='mp3')    
+    meter = Meter(audio.frame_rate)  # 创建一个响度计量器
+    sr = audio.frame_rate
+    samples = audio.get_array_of_samples()
+    audio = np.array(samples, dtype=np.float64)
+
+    # 测量积分响度
+    loudness = meter.integrated_loudness(audio)
+
+    # 计算增益
+    gain_db = target_loudness - loudness
+    gain_linear = 10 ** (gain_db / 20.0)
+
+    # 应用增益
+    balanced_audio = audio * gain_linear
+
+    # 应用软限幅以防止削波
+    balanced_audio = np.tanh(balanced_audio)
+
+    # 将numpy数组转换回bytes
+    balanced_audio = (balanced_audio * 32767).astype(np.int16)
+    byte_io = io.BytesIO()
+    sf.write(byte_io, balanced_audio, sr, format='mp3')
+    normalized_audio_bytes = byte_io.getvalue()
+
+    return normalized_audio_bytes
+
+if __name__ == "__main__":
+    normalize_audio_loudness()