12.26

services/a2f_api/text_to_blendshapes_service.py

@@ -6,14 +6,26 @@ import queue
 import threading
 from datetime import datetime
 from tts_service import TTSService
+from edge_tts_service import EdgeTTSService
 from a2f_service import A2FService
 from blend_shape_parser import BlendShapeParser
 
 class TextToBlendShapesService:
     DEFAULT_SPLIT_PUNCTUATIONS = '。！？；!?;,，'
 
-    def __init__(self, lang='zh-CN', a2f_url="192.168.1.39:52000"):
-        self.tts = TTSService(lang=lang)
+    def __init__(self, lang='zh-CN', a2f_url="192.168.1.39:52000", tts_provider='edge-tts'):
+        """
+        初始化服务
+        :param lang: 语言
+        :param a2f_url: A2F服务地址
+        :param tts_provider: TTS提供商 ('pyttsx3' 或 'edge-tts')
+        """
+        # 根据选择初始化TTS服务
+        if tts_provider == 'edge-tts':
+            self.tts = EdgeTTSService(lang=lang)
+        else:
+            self.tts = TTSService(lang=lang)
+
         self.a2f = A2FService(a2f_url=a2f_url)
         self.parser = BlendShapeParser()
 
@@ -67,7 +79,18 @@ class TextToBlendShapesService:
             yield {'type': 'error', 'message': '文本为空'}
             return
 
-        yield {'type': 'status', 'stage': 'split', 'sentences': len(sentences), 'message': f'已拆分为 {len(sentences)} 个句子'}
+        yield {
+            'type': 'status',
+            'stage': 'split',
+            'sentences': len(sentences),
+            'sentence_texts': sentences,  # 发送句子文本列表
+            'message': f'已拆分为 {len(sentences)} 个句子'
+        }
+
+        # 打印句子列表用于调试
+        print(f"[调试] 发送给前端的句子列表:")
+        for i, s in enumerate(sentences):
+            print(f"  [{i}] {s}")
 
         # 使用队列来收集处理完成的句子
         result_queue = queue.Queue()
@@ -126,6 +149,7 @@ class TextToBlendShapesService:
                 is_continuation = self.is_continuation[next_index] if next_index < len(self.is_continuation) else False
 
                 print(f"[主线程] 正在推送句子 {next_index} 的 {len(frames)} 帧 {'(连续)' if is_continuation else ''}")
+                print(f"[调试] 句子 {next_index} 对应文本: {sentences[next_index] if next_index < len(sentences) else 'N/A'}")
 
                 # 如果不是连续句子，重置累计时间
                 if not is_continuation and next_index > 0:
@@ -135,7 +159,6 @@ class TextToBlendShapesService:
                     # 调整时间码：从累计时间开始
                     frame['timeCode'] = cumulative_time + frame['timeCode']
                     frame['sentenceIndex'] = next_index
-                    frame['isContinuation'] = is_continuation
                     total_frames += 1
                     yield {'type': 'frame', 'frame': frame}
 
@@ -157,6 +180,7 @@ class TextToBlendShapesService:
         start_time = time.time()
 
         print(f"[线程 {index}] 开始处理: {sentence[:30]}...")
+        print(f"[调试] 线程 {index} 实际处理的完整文本: [{sentence}] (长度: {len(sentence)}字)")
         _, audio_path = self._prepare_output_paths(output_dir, suffix=f's{index:03d}')
 
         print(f"[线程 {index}] TTS 开始...")
@@ -166,7 +190,7 @@ class TextToBlendShapesService:
         print(f"[线程 {index}] TTS 完成，耗时 {tts_time:.2f}秒，A2F 开始...")
 
         a2f_start = time.time()
-        csv_path = self.a2f.audio_to_csv(audio_path)
+        csv_path, temp_dir = self.a2f.audio_to_csv(audio_path)  # 接收临时目录路径
         a2f_time = time.time() - a2f_start
         print(f"[线程 {index}] A2F 完成，耗时 {a2f_time:.2f}秒，解析中...")
 
@@ -174,6 +198,14 @@ class TextToBlendShapesService:
         frames = list(self.parser.iter_csv_to_blend_shapes(csv_path))
         parse_time = time.time() - parse_start
 
+        # 解析完成后清理临时目录
+        import shutil
+        try:
+            shutil.rmtree(temp_dir, ignore_errors=True)
+            print(f"[线程 {index}] 已清理临时目录: {temp_dir}")
+        except Exception as e:
+            print(f"[线程 {index}] 清理临时目录失败: {e}")
+
         total_time = time.time() - start_time
         print(f"[线程 {index}] 完成！生成了 {len(frames)} 帧 | 总耗时: {total_time:.2f}秒 (TTS: {tts_time:.2f}s, A2F: {a2f_time:.2f}s, 解析: {parse_time:.2f}s)")
 
@@ -239,12 +271,15 @@ class TextToBlendShapesService:
             length = len(first)
             parts = []
 
-            if length <= 12:
-                # 12字以内分两部分
+            if length <= 8:
+                # 8字以下不拆分
+                parts = [first]
+            elif length <= 12:
+                # 8-12字分两部分
                 mid = length // 2
                 parts = [first[:mid], first[mid:]]
             else:
-                # 12字之后：前6字，再6字，剩下的
+                # 12字以上：前6字，再6字，剩下的
                 parts = [first[:6], first[6:12], first[12:]]
 
             # 替换第一句为多个小句