Merge pull request #2552 from Packeting1/multi_lang_tts

feat: 为阿里百炼流式TTS支持多语言音色

main/xiaozhi-server/core/providers/asr/fun_local.py

@@ -8,8 +8,9 @@ import asyncio
 
 from config.logger import setup_logging
 from typing import Optional, Tuple, List
+from core.providers.asr.base import ASRProviderBase
+from core.providers.asr.utils import lang_tag_filter
 from funasr import AutoModel
-from funasr.utils.postprocess_utils import rich_transcription_postprocess
 from core.providers.asr.base import ASRProviderBase
 from core.providers.asr.dto.dto import InterfaceType
 
@@ -102,7 +103,9 @@ class ASRProvider(ASRProviderBase):
                     use_itn=True,
                     batch_size_s=60,
                 )
-                text = await asyncio.to_thread(rich_transcription_postprocess, result[0]["text"])
+                # text = await asyncio.to_thread(rich_transcription_postprocess, result[0]["text"])
+                # 使用lang_tag_filter处理识别结果
+                text = await asyncio.to_thread(lang_tag_filter, result[0]["text"])
                 logger.bind(tag=TAG).debug(
                     f"语音识别耗时: {time.time() - start_time:.3f}s | 结果: {text}"
                 )

main/xiaozhi-server/core/providers/asr/fun_server.py

@@ -1,12 +1,12 @@
 from typing import Optional, Tuple, List
 from core.providers.asr.base import ASRProviderBase
+from core.providers.asr.utils import lang_tag_filter
 from core.providers.asr.dto.dto import InterfaceType
 import ssl
 import json
 import websockets
 from config.logger import setup_logging
 import asyncio
-import re
 
 TAG = __name__
 logger = setup_logging()
@@ -151,9 +151,13 @@ class ASRProvider(ASRProviderBase):
 
                 # Get the result from the receive task
                 result = receive_task.result()
-                match = re.match(r"<\|(.*?)\|><\|(.*?)\|><\|(.*?)\|>(.*)", result)
-                if match:
-                    result = match.group(4).strip()
+                
+                # match = re.match(r"<\|(.*?)\|><\|(.*?)\|><\|(.*?)\|>(.*)", result)
+                # if match:
+                #     result = match.group(4).strip()
+
+                # Handle language tags
+                result = lang_tag_filter(result)
                 return (
                     result,
                     file_path,

main/xiaozhi-server/core/providers/asr/utils.py

@@ -0,0 +1,43 @@
+import re
+from config.logger import setup_logging
+
+TAG = __name__
+logger = setup_logging()
+
+
+def lang_tag_filter(text):
+    """
+    过滤函数：只保留语言标签，移除其他所有标签
+    
+    用于FunASR识别结果的处理，保留语言标签（如<|zh|>、<|en|>等），
+    但移除其他所有格式的标签（如时间戳、情感标签等）
+    
+    Args:
+        text: ASR识别的原始文本，可能包含多种标签
+        
+    Returns:
+        str: 处理后的文本，只保留语言标签（如果存在）
+        
+    Examples:
+        >>> lang_tag_filter("<|zh|><|emotion:happy|>你好")
+        '<|zh|>你好'
+        >>> lang_tag_filter("<|en|>hello world")
+        '<|en|>hello world'
+    """
+    # 定义语言标签模式
+    lang_pattern = r"<\|(zh|en|yue|ja|ko|nospeech)\|>"
+    lang_tags = re.findall(lang_pattern, text)
+
+    # 移除所有 < | ... | > 格式的标签
+    clean_text = re.sub(r"<\|.*?\|>", "", text)
+
+    # 在开头添加语言标签（如果存在）
+    if lang_tags:
+        if len(lang_tags) > 1:
+            logger.bind(tag=TAG).warning(
+                f"检测到多个语言标签: {lang_tags}，仅使用第一个: {lang_tags[0]}"
+            )
+        clean_text = f"<|{lang_tags[0]}|>{clean_text}"
+
+    return clean_text.strip()
+