添加分段合成

webui.py

@@ -24,6 +24,7 @@ from text import cleaned_text_to_sequence, get_bert
 from text.cleaner import clean_text
 import gradio as gr
 import webbrowser
+import numpy as np
 
 net_g = None
 
@@ -103,22 +104,17 @@ def infer(text, sdp_ratio, noise_scale, noise_scale_w, length_scale, sid, langua
         return audio
 
 
-def tts_fn(
+def tts_fn(text, speaker, sdp_ratio, noise_scale, noise_scale_w, length_scale):
-    text, speaker, sdp_ratio, noise_scale, noise_scale_w, length_scale, language
+    slices = text.split("|")
-):
+    audio_list = []
     with torch.no_grad():
-        audio = infer(
+        for slice in slices:
-            text,
+            audio = infer(slice, sdp_ratio=sdp_ratio, noise_scale=noise_scale, noise_scale_w=noise_scale_w, length_scale=length_scale, sid=speaker)
-            sdp_ratio=sdp_ratio,
+            audio_list.append(audio)
-            noise_scale=noise_scale,
+            silence = np.zeros(hps.data.sampling_rate)  # 生成1秒的静音
-            noise_scale_w=noise_scale_w,
+            audio_list.append(silence)  # 将静音添加到列表中
-            length_scale=length_scale,
+    audio_concat = np.concatenate(audio_list)
-            sid=speaker,
+    return "Success", (hps.data.sampling_rate, audio_concat)
-            language=language,
-        )
-        torch.cuda.empty_cache()
-    return "Success", (hps.data.sampling_rate, audio)
-
 
 if __name__ == "__main__":
     parser = argparse.ArgumentParser()