修复tts未生成就判断结束的问题
This commit is contained in:
朱潮
parent
ce1f101ec7
commit
0ee0252c8a
@ -713,10 +713,11 @@ class OutputProcess:
|
||||
self.playback_cooldown_period = 0.05 # 播放冷却时间（秒）- 防止回声，减少到0.05秒
|
||||
self.playback_completed = False # 播放完成标志
|
||||
self.end_signal_received = False # 结束信号接收标志
|
||||
self.end_signal_time = 0 # 结束信号接收时间
|
||||
|
||||
# 智能缓冲系统 - 借鉴 recorder.py 的智能句子累积策略
|
||||
self.preload_buffer = [] # 预加载缓冲区（保留用于音频块）
|
||||
self.preload_size = 6 # 预加载6个音频块（增加缓冲减少卡顿）
|
||||
self.preload_size = 3 # 预加载3个音频块（减少预加载时间，加快播放启动）
|
||||
|
||||
# 简化的音频系统 - 直接使用预加载缓冲区
|
||||
self.audio_queue_lock = None # 音频队列操作锁（在run方法中初始化）
|
||||
@ -873,6 +874,8 @@ class OutputProcess:
|
||||
if not self.currently_playing:
|
||||
self.currently_playing = True
|
||||
self.last_audio_chunk_time = time.time() # 记录最后播放时间
|
||||
print(f"🎵 播放状态变化: currently_playing = True (开始播放)")
|
||||
print(f"🎵 设置last_audio_chunk_time = {self.last_audio_chunk_time}")
|
||||
|
||||
# 如果是第一次播放，不设置冷却期
|
||||
if chunks_played == 0:
|
||||
@ -917,6 +920,7 @@ class OutputProcess:
|
||||
len(self.preload_buffer) == 0 and
|
||||
time_since_last_chunk > 1.0):
|
||||
self.currently_playing = False
|
||||
print(f"🎵 播放状态变化: currently_playing = False (播放缓冲区和预加载缓冲区都为空)")
|
||||
elif len(self.playback_buffer) == 0 and len(self.preload_buffer) > 0:
|
||||
# 播放缓冲区为空但预加载缓冲区有数据，自动转移数据
|
||||
transfer_count = min(3, len(self.preload_buffer)) # 一次转移3个块
|
||||
@ -937,6 +941,7 @@ class OutputProcess:
|
||||
len(self.preload_buffer) == 0 and
|
||||
time_since_last_chunk > 1.5): # 异常情况下，等待更长时间且确保缓冲区为空
|
||||
self.currently_playing = False
|
||||
print(f"🎵 播放状态变化: currently_playing = False (异常情况处理，缓冲区为空且{time_since_last_chunk:.1f}秒无播放)")
|
||||
else:
|
||||
# 保持播放状态，继续处理
|
||||
pass
|
||||
@ -948,6 +953,7 @@ class OutputProcess:
|
||||
# 线程结束时确保状态正确
|
||||
if self.currently_playing:
|
||||
self.currently_playing = False
|
||||
print(f"🎵 播放状态变化: currently_playing = False (播放工作线程结束)")
|
||||
if playback_stream:
|
||||
try:
|
||||
playback_stream.stop_stream()
|
||||
@ -977,9 +983,10 @@ class OutputProcess:
|
||||
# 1. 处理音频队列（数据接收）
|
||||
self._process_audio_queue()
|
||||
|
||||
# 2. 检查播放状态
|
||||
# 2. 检查播放状态 - 使用增强播放完成检测
|
||||
if self.end_signal_received:
|
||||
self._check_playback_completion()
|
||||
if self._check_enhanced_playback_completion():
|
||||
self._finish_playback()
|
||||
|
||||
# 3. 检查设备健康状态和冷却期 - 防止回声
|
||||
current_time = time.time()
|
||||
@ -1099,20 +1106,16 @@ class OutputProcess:
|
||||
print(f"📥 输出进程收到结束信号")
|
||||
end_signal_received = True
|
||||
self.end_signal_received = True
|
||||
self.end_signal_time = time.time() # 记录收到结束信号的时间
|
||||
|
||||
# 只有在有音频数据或者TTS缓冲区有内容时，才设置all_audio_received
|
||||
# 这样可以避免在没有音频的情况下误判播放完成
|
||||
if (len(self.preload_buffer) > 0 or
|
||||
len(self.playback_buffer) > 0 or
|
||||
len(self.tts_buffer) > 0 or
|
||||
self.tts_task_queue.qsize() > 0):
|
||||
self.all_audio_received = True
|
||||
print(f"📥 设置all_audio_received=True，检测到有待处理的数据")
|
||||
else:
|
||||
# 如果没有任何数据，可能是LLM响应为空或其他问题
|
||||
# 也设置为True，让播放完成检测处理这种情况
|
||||
self.all_audio_received = True
|
||||
print(f"📥 警告：收到结束信号但没有检测到任何音频数据，可能存在问题")
|
||||
# 延迟设置all_audio_received，确保音频真正开始播放后再设置
|
||||
# 暂时设置为False，让播放启动逻辑处理
|
||||
self.all_audio_received = False
|
||||
print(f"📥 收到结束信号，状态变化:")
|
||||
print(f" - end_signal_received: True")
|
||||
print(f" - all_audio_received: False (延迟设置)")
|
||||
print(f" - completion_sent: False")
|
||||
print(f" - playback_completed: False")
|
||||
|
||||
# 重置完成事件标记
|
||||
self.completion_sent = False
|
||||
@ -1206,14 +1209,21 @@ class OutputProcess:
|
||||
|
||||
# 直接添加到预加载缓冲区
|
||||
self.preload_buffer.append(audio_data)
|
||||
print(f"🎵 音频数据已添加到预加载缓冲区，当前大小: {len(self.preload_buffer)}/{self.preload_size}")
|
||||
|
||||
# 检查是否应该开始播放或补充播放缓冲区
|
||||
if not self.is_playing and len(self.preload_buffer) >= self.preload_size:
|
||||
print(f"🎵 预加载缓冲区达到{self.preload_size}个块，开始首次播放...")
|
||||
# 首次启动播放
|
||||
self.playback_buffer.extend(self.preload_buffer)
|
||||
self.preload_buffer.clear()
|
||||
self.is_playing = True
|
||||
self.last_playback_time = 0 # 重置播放时间，避免立即触发冷却期
|
||||
|
||||
# 音频真正开始播放，设置all_audio_received为True
|
||||
if self.end_signal_received and not self.all_audio_received:
|
||||
self.all_audio_received = True
|
||||
print(f"🎵 音频开始播放，设置all_audio_received=True")
|
||||
# 确保播放工作线程知道有数据要播放
|
||||
if not self.currently_playing:
|
||||
# 播放工作线程会自动检测播放缓冲区并开始播放
|
||||
@ -1231,6 +1241,12 @@ class OutputProcess:
|
||||
self.preload_buffer.clear()
|
||||
self.is_playing = True
|
||||
self.last_playback_time = 0
|
||||
|
||||
# 强制转移后设置all_audio_received为True
|
||||
if not self.all_audio_received:
|
||||
self.all_audio_received = True
|
||||
print(f"🎵 强制转移预加载缓冲区后，设置all_audio_received=True")
|
||||
|
||||
print(f"🎵 强制开始播放音频，播放缓冲区大小: {len(self.playback_buffer)}")
|
||||
|
||||
else:
|
||||
@ -1271,6 +1287,11 @@ class OutputProcess:
|
||||
self.preload_buffer.clear()
|
||||
self.is_playing = True
|
||||
self.last_playback_time = 0 # 重置播放时间，避免立即触发冷却期
|
||||
|
||||
# 音频真正开始播放，设置all_audio_received为True
|
||||
if self.end_signal_received and not self.all_audio_received:
|
||||
self.all_audio_received = True
|
||||
print(f"🎵 音频开始播放（最小缓冲区模式），设置all_audio_received=True")
|
||||
print(f"🎵 开始播放音频（最小缓冲区满足）")
|
||||
print(f"🔍 已重置last_playback_time，避免立即触发冷却期")
|
||||
elif self.is_playing and len(self.playback_buffer) < 2 and len(self.preload_buffer) > 0:
|
||||
@ -1356,6 +1377,7 @@ class OutputProcess:
|
||||
self.no_active_playback = False
|
||||
self.end_signal_received = False
|
||||
self.playback_completed = False
|
||||
self.end_signal_time = 0 # 重置结束信号时间
|
||||
|
||||
print("📡 输出进程：已重置所有播放完成状态标志")
|
||||
|
||||
@ -1387,6 +1409,8 @@ class OutputProcess:
|
||||
if not self.end_signal_received:
|
||||
return False
|
||||
|
||||
print(f"🔍 增强播放完成检查开始...")
|
||||
|
||||
# 更新状态变量
|
||||
self.pre_buffer_empty = (len(self.preload_buffer) == 0)
|
||||
self.playback_buffer_empty = (len(self.playback_buffer) == 0)
|
||||
@ -1394,16 +1418,64 @@ class OutputProcess:
|
||||
|
||||
tts_queue_size = self.tts_task_queue.qsize()
|
||||
|
||||
# 计算时间差
|
||||
current_time = time.time()
|
||||
time_since_last_chunk = current_time - self.last_audio_chunk_time
|
||||
|
||||
# 添加时间维度检查
|
||||
if self.last_audio_chunk_time > 0:
|
||||
time_since_last_chunk = time.time() - self.last_audio_chunk_time
|
||||
else:
|
||||
time_since_last_chunk = 0
|
||||
print(f"🔍 增强播放完成检查详情:")
|
||||
print(f" - end_signal_received: {self.end_signal_received}")
|
||||
print(f" - llm_generation_complete: {self.llm_generation_complete}")
|
||||
print(f" - tts_generation_complete: {self.tts_generation_complete}")
|
||||
print(f" - all_audio_received: {self.all_audio_received}")
|
||||
print(f" - tts_queue_size: {tts_queue_size}")
|
||||
print(f" - tts_buffer_size: {len(self.tts_buffer)}")
|
||||
print(f" - playback_buffer: {len(self.playback_buffer)}")
|
||||
print(f" - preload_buffer: {len(self.preload_buffer)}")
|
||||
print(f" - currently_playing: {self.currently_playing}")
|
||||
print(f" - is_playing: {self.is_playing}")
|
||||
print(f" - last_audio_chunk_time: {self.last_audio_chunk_time}")
|
||||
print(f" - time_since_last_chunk: {time_since_last_chunk:.3f}秒")
|
||||
|
||||
# 检查TTS是否正在生成 - 新增条件
|
||||
# TTS正在生成的条件：队列中有任务 或 TTS生成未完成 或 还有待处理的缓冲区内容
|
||||
tts_is_generating = (tts_queue_size > 0 or not self.tts_generation_complete or len(self.tts_buffer) > 0)
|
||||
|
||||
# 检查TTS是否正在生成 - 修复逻辑
|
||||
# 如果TTS生成未完成但队列和缓冲区都为空，可能是状态标记错误，自动修正
|
||||
if not self.tts_generation_complete and tts_queue_size == 0 and len(self.tts_buffer) == 0:
|
||||
print(f"🔧 检测到TTS生成状态异常：tts_generation_complete=False但无数据，自动修正为True")
|
||||
self.tts_generation_complete = True
|
||||
|
||||
# TTS正在生成的条件：队列中有任务 或 还有待处理的缓冲区内容 或 TTS生成未完成
|
||||
tts_is_generating = (tts_queue_size > 0 or len(self.tts_buffer) > 0 or not self.tts_generation_complete)
|
||||
print(f" - tts_is_generating: {tts_is_generating}")
|
||||
print(f" - pre_buffer_empty: {self.pre_buffer_empty}")
|
||||
print(f" - playback_buffer_empty: {self.playback_buffer_empty}")
|
||||
print(f" - no_active_playback: {self.no_active_playback}")
|
||||
|
||||
# 特殊处理1：如果没有任何音频数据，暂时不设置all_audio_received，等待TTS生成
|
||||
# 注意：这里不自动设置all_audio_received，因为可能TTS还在生成中
|
||||
# 等待播放完成检测中的超时机制来处理这种情况
|
||||
|
||||
# 特殊处理2：如果all_audio_received为False但其他条件都满足，强制设置为True
|
||||
print(f"🔍 检查特殊处理条件:")
|
||||
print(f" - not all_audio_received: {not self.all_audio_received}")
|
||||
print(f" - llm_generation_complete: {self.llm_generation_complete}")
|
||||
print(f" - tts_generation_complete: {self.tts_generation_complete}")
|
||||
print(f" - pre_buffer_empty: {self.pre_buffer_empty}")
|
||||
print(f" - playback_buffer_empty: {self.playback_buffer_empty}")
|
||||
print(f" - no_active_playback: {self.no_active_playback}")
|
||||
print(f" - tts_queue_size == 0: {tts_queue_size == 0}")
|
||||
print(f" - not tts_is_generating: {not tts_is_generating}")
|
||||
|
||||
if (not self.all_audio_received and
|
||||
self.llm_generation_complete and
|
||||
self.tts_generation_complete and
|
||||
self.pre_buffer_empty and
|
||||
self.playback_buffer_empty and
|
||||
self.no_active_playback and
|
||||
tts_queue_size == 0 and
|
||||
not tts_is_generating):
|
||||
# 这种情况可能是音频播放开始后没有正确设置all_audio_received
|
||||
print(f"🔧 检测到all_audio_received为False但其他播放条件已满足，强制设置为True")
|
||||
self.all_audio_received = True
|
||||
|
||||
# 检查是否所有条件都满足 - 使用更新的状态变量，添加TTS生成状态检查
|
||||
all_conditions_met = (
|
||||
@ -1417,24 +1489,65 @@ class OutputProcess:
|
||||
not tts_is_generating # 新增：确保TTS不在生成中
|
||||
)
|
||||
|
||||
print(f"🔍 最终条件检查结果:")
|
||||
print(f" - all_conditions_met: {all_conditions_met}")
|
||||
print(f" - 各个条件详情:")
|
||||
print(f" * llm_generation_complete: {self.llm_generation_complete}")
|
||||
print(f" * tts_generation_complete: {self.tts_generation_complete}")
|
||||
print(f" * all_audio_received: {self.all_audio_received}")
|
||||
print(f" * pre_buffer_empty: {self.pre_buffer_empty}")
|
||||
print(f" * playback_buffer_empty: {self.playback_buffer_empty}")
|
||||
print(f" * no_active_playback: {self.no_active_playback}")
|
||||
print(f" * tts_queue_size == 0: {tts_queue_size == 0}")
|
||||
print(f" * not tts_is_generating: {not tts_is_generating}")
|
||||
|
||||
if all_conditions_met:
|
||||
print(f"✅ 所有播放完成条件已满足，进行时间检查...")
|
||||
print(f" - last_audio_chunk_time: {self.last_audio_chunk_time}")
|
||||
print(f" - time_since_last_chunk: {time_since_last_chunk:.3f}秒")
|
||||
print(f" - 需要等待时间: 1.0秒")
|
||||
|
||||
# 额外时间检查：确保音频真正播放完成
|
||||
if self.last_audio_chunk_time > 0 and time_since_last_chunk > 0.3: # 至少0.3秒没有新音频播放
|
||||
if self.last_audio_chunk_time > 0 and time_since_last_chunk > 1.0: # 至少1.0秒没有新音频播放（增加等待时间确保音频完成）
|
||||
print(f"✅ 所有播放完成条件已满足，且{time_since_last_chunk:.2f}秒无新音频，可以结束播放")
|
||||
return True
|
||||
elif self.last_audio_chunk_time > 0 and time_since_last_chunk <= 1.0:
|
||||
print(f"⏳ 所有条件满足但等待时间不足（已等待{time_since_last_chunk:.2f}秒，需要1.0秒）...")
|
||||
return False
|
||||
elif self.last_audio_chunk_time == 0:
|
||||
print(f"🔍 从未开始播放的情况，检查是否有音频数据...")
|
||||
# 如果从未开始播放，检查是否有音频数据
|
||||
if len(self.playback_buffer) == 0 and len(self.preload_buffer) == 0:
|
||||
print(f"🔍 播放缓冲区和预加载缓冲区都为空")
|
||||
# 检查是否有TTS任务在排队
|
||||
if self.tts_task_queue.qsize() == 0 and len(self.tts_buffer) == 0:
|
||||
# 真的没有任何数据，可能是LLM响应为空或TTS失败
|
||||
print(f"⚠️ 从未开始播放且无任何数据，可能LLM响应为空或TTS失败，强制结束")
|
||||
return True
|
||||
print(f"🔍 TTS队列和缓冲区都为空")
|
||||
|
||||
# 检查是否刚刚收到结束信号，需要给TTS一些时间生成音频
|
||||
time_since_end_signal = time.time() - getattr(self, 'end_signal_time', 0)
|
||||
if not hasattr(self, 'end_signal_time'):
|
||||
self.end_signal_time = time.time()
|
||||
|
||||
print(f"🔍 距离收到结束信号: {time_since_end_signal:.2f}秒")
|
||||
|
||||
# 如果距离收到结束信号时间很短（< 5秒），可能是TTS还在生成中
|
||||
if time_since_end_signal < 5.0:
|
||||
print(f"⏳ 刚收到结束信号，等待TTS生成音频（已等待{time_since_end_signal:.1f}秒）...")
|
||||
return False
|
||||
elif time_since_end_signal < 10.0:
|
||||
print(f"⏳ 已等待{time_since_end_signal:.1f}秒，继续等待TTS生成...")
|
||||
return False
|
||||
else:
|
||||
# 等待超过10秒，确实没有音频数据，可能是TTS失败
|
||||
print(f"⚠️ 已等待{time_since_end_signal:.1f}秒仍无音频数据，可能TTS失败，强制结束")
|
||||
return True
|
||||
else:
|
||||
print(f"🔍 还有TTS任务待处理: tts_queue_size={tts_queue_size}, tts_buffer_size={len(self.tts_buffer)}")
|
||||
# 还有TTS任务待处理
|
||||
print(f"⏳ 从未开始播放但还有TTS任务待处理，等待TTS生成...")
|
||||
return False
|
||||
else:
|
||||
print(f"🔍 还有音频数据: playback_buffer={len(self.playback_buffer)}, preload_buffer={len(self.preload_buffer)}")
|
||||
print(f"⏳ 从未开始播放但还有音频数据，等待播放开始...")
|
||||
return False
|
||||
else:
|
||||
@ -1504,53 +1617,8 @@ class OutputProcess:
|
||||
return False
|
||||
|
||||
|
||||
def _check_playback_completion(self):
|
||||
"""简化的播放完成状态检查"""
|
||||
if not self.end_signal_received:
|
||||
return
|
||||
|
||||
tts_queue_size = self.tts_task_queue.qsize()
|
||||
playback_queue_size = len(self.playback_buffer) + len(self.preload_buffer)
|
||||
|
||||
# 调试日志：播放完成状态
|
||||
print(f"🔍 播放完成检查:")
|
||||
print(f" - end_signal_received: {self.end_signal_received}")
|
||||
print(f" - tts_queue_size: {tts_queue_size}")
|
||||
print(f" - playback_buffer: {len(self.playback_buffer)}")
|
||||
print(f" - preload_buffer: {len(self.preload_buffer)}")
|
||||
print(f" - currently_playing: {self.currently_playing}")
|
||||
print(f" - playback_completed: {self.playback_completed}")
|
||||
print(f" - completion_sent: {self.completion_sent}")
|
||||
|
||||
# 检查条件：所有队列为空且没有在播放
|
||||
if tts_queue_size == 0 and playback_queue_size == 0 and not self.currently_playing:
|
||||
print(f"✅ 所有队列已清空且播放器空闲，触发播放完成")
|
||||
print(f"📥 调用 _finish_playback() 前，completion_sent={self.completion_sent}")
|
||||
self._finish_playback()
|
||||
print(f"📥 调用 _finish_playback() 后，completion_sent={self.completion_sent}")
|
||||
# 重置结束信号标志，防止重复触发
|
||||
self.end_signal_received = False
|
||||
print(f"📥 已重置结束信号标志，防止重复触发")
|
||||
return
|
||||
|
||||
# 检查条件：所有队列为空但播放器还在播放（最后一个音频块）
|
||||
if tts_queue_size == 0 and playback_queue_size == 0 and self.currently_playing:
|
||||
print(f"⏳ 等待最后一个音频块播放完成...")
|
||||
time.sleep(0.3)
|
||||
if not self.currently_playing:
|
||||
print(f"✅ 最后一个音频块播放完成，触发播放完成")
|
||||
print(f"📥 调用 _finish_playback() 前，completion_sent={self.completion_sent}")
|
||||
self._finish_playback()
|
||||
print(f"📥 调用 _finish_playback() 后，completion_sent={self.completion_sent}")
|
||||
# 重置结束信号标志，防止重复触发
|
||||
self.end_signal_received = False
|
||||
print(f"📥 已重置结束信号标志，防止重复触发")
|
||||
return
|
||||
else:
|
||||
print(f"⚠️ 最后一个音频块播放超时")
|
||||
# 即使超时也要重置标志，防止重复检测
|
||||
self.end_signal_received = False
|
||||
print(f"📥 已重置结束信号标志（超时情况）")
|
||||
# 注意：简化播放完成检测已移除，统一使用增强播放完成检测 _check_enhanced_playback_completion()
|
||||
|
||||
|
||||
def _cleanup(self):
|
||||
"""清理资源"""
|
||||
|
||||
Loading…
Reference in New Issue
Block a user