update speed

Artrajz · Apr 23, 2024 · 0414037 · 0414037
1 parent 6a71703
commit 0414037
Show file tree

Hide file tree

Showing 4 changed files with 50 additions and 22 deletions.
diff --git a/gpt_sovits/gpt_sovits.py b/gpt_sovits/gpt_sovits.py
@@ -449,18 +449,13 @@ def audio_postprocess(self, audio: List[torch.Tensor], sr: int, batch_index_list
 
         try:
             if speed_factor != 1.0:
-                audio = self.speed_change(audio, speed_factor=speed_factor, sr=int(sr))
+                from .utils import speed_change
+                audio = speed_change(audio, speed_factor=speed_factor, sr=int(sr))
         except Exception as e:
             logging.error(f"Failed to change speed of audio: \n{e}")
 
         return audio
 
-    def speed_change(self, input_audio: np.ndarray, speed_factor: float, sr: int):
-        # 变速处理
-        processed_audio = librosa.effects.time_stretch(input_audio, rate=speed_factor)
-
-        return processed_audio
-
     def infer(self, text, lang, reference_audio, reference_audio_sr, prompt_text, prompt_lang, top_k, top_p,
               temperature, batch_size: int = 5, batch_threshold: float = 0.75, split_bucket: bool = True,
               return_fragment: bool = False, speed_factor: float = 1.0, seed: int = -1,

diff --git a/gpt_sovits/utils.py b/gpt_sovits/utils.py
@@ -1,12 +1,11 @@
-"""
-改用librosa。
-
-（不是ffmpeg不好用，而是少装一个软件更有性价比）
-"""
+import logging
+import subprocess
 
 import librosa
 import numpy as np
 
+import config
+
 
 class DictToAttrRecursive(dict):
     def __init__(self, input_dict):
@@ -65,3 +64,37 @@ def __delattr__(self, item):
 #         raise RuntimeError(f"Failed to load audio: {e}")
 #
 #     return np.frombuffer(out, np.float32).flatten()
+
+def speed_change(input_audio: np.ndarray, speed_factor: float, sr: int):
+    """这里修改为使用原数据float32变速，GPT-SoVITS官方为int16"""
+
+    raw_audio = input_audio.tobytes()
+
+    ffmpeg_command = [
+        # f'{config.abs_path}/bin/ffmpeg',
+        'ffmpeg',
+        '-f', 'f32le',
+        '-ar', str(sr),
+        '-ac', '1',
+        '-i', 'pipe:',
+        '-filter:a', f'atempo={speed_factor}',
+        '-f', 'f32le',
+        '-acodec', 'pcm_f32le',
+        'pipe:'
+    ]
+    try:
+        ffmpeg_process = subprocess.Popen(
+            ffmpeg_command,
+            stdin=subprocess.PIPE,
+            stdout=subprocess.PIPE,
+            stderr=subprocess.PIPE
+        )
+        out, info = ffmpeg_process.communicate(input=raw_audio)
+    except Exception as e:
+        logging.error(e)
+        logging.error(info)
+
+    # 将管道输出解码为 NumPy 数组
+    processed_audio = np.frombuffer(out, np.float32)
+
+    return processed_audio
diff --git a/tts_app/static/js/index.js b/tts_app/static/js/index.js
@@ -114,7 +114,7 @@ function getLink() {
         temperature = document.getElementById('input_temperature4').value;
         batch_size = document.getElementById('input_batch_size4').value;
         seed = document.getElementById('input_seed4').value;
-        // speed = document.getElementById('input_speed4').value;
+        speed = document.getElementById('input_speed4').value;
         url += "/voice/gpt-sovits?id=" + id;
 
     } else {
@@ -317,7 +317,7 @@ function setAudioSourceByPost() {
         top_p = $("#input_top_p4").val();
         temperature = $("#input_temperature4").val();
         batch_size = $("#input_batch_size4").val();
-        // speed = $("#input_speed4").val();
+        speed = $("#input_speed4").val();
         seed = $("#input_seed4").val();
 
     }

diff --git a/tts_app/templates/pages/gpt_sovits.html b/tts_app/templates/pages/gpt_sovits.html
@@ -93,14 +93,14 @@
                        placeholder="5"/>
             </div>
 
-            {#            <div class="col-md-2 mb-3">#}
-            {#                <label for="speed" data-toggle="tooltip" data-placement="top"#}
-            {#                       title="">speed</label>#}
-            {#                <input type="text" class="form-control speed" id="input_speed4"#}
-            {#                       oninput="updateLink()"#}
-            {#                       value=""#}
-            {#                       placeholder="1.0"/>#}
-            {#            </div>#}
+            <div class="col-md-2 mb-3">
+                <label for="speed" data-toggle="tooltip" data-placement="top"
+                       title="调节语速，speed越大语速越快">speed</label>
+                <input type="text" class="form-control speed" id="input_speed4"
+                       oninput="updateLink()"
+                       value=""
+                       placeholder="1.0"/>
+            </div>
 
         </div>
         <div class="row">