talkingAvater_bgk / streaming_client.py
oKen38461's picture
ストリーミング関連のフレームレートを25fpsから20fpsに変更し、関連するテストケースを更新しました。これにより、全体のフレーム数計算が一貫性を持つようになりました。
2089ecf
"""
DittoTalkingHead Streaming Client
WebSocketを使用したストリーミングクライアントの実装例
"""
import asyncio
import websockets
import numpy as np
import soundfile as sf
import base64
import json
import cv2
from typing import Optional, Callable
import pyaudio
import threading
import queue
from pathlib import Path
import logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)
class DittoStreamingClient:
"""DittoTalkingHeadストリーミングクライアント"""
def __init__(self, server_url="ws://localhost:8000/ws/generate"):
self.server_url = server_url
self.sample_rate = 16000
self.chunk_duration = 0.2 # 200ms
self.chunk_size = int(self.sample_rate * self.chunk_duration)
self.websocket = None
self.is_connected = False
self.frame_callback: Optional[Callable] = None
self.final_video_callback: Optional[Callable] = None
async def connect(self, source_image_path: str):
"""サーバーに接続してセッションを開始"""
try:
# 画像をBase64エンコード
with open(source_image_path, "rb") as f:
image_b64 = base64.b64encode(f.read()).decode('utf-8')
# WebSocket接続
self.websocket = await websockets.connect(self.server_url)
self.is_connected = True
# 初期設定を送信
await self.websocket.send(json.dumps({
"source_image": image_b64,
"sample_rate": self.sample_rate,
"chunk_duration": self.chunk_duration
}))
# 応答を待つ
response = await self.websocket.recv()
data = json.loads(response)
if data["type"] == "ready":
logger.info(f"Connected to server: {data['message']}")
return True
else:
logger.error(f"Connection failed: {data}")
return False
except Exception as e:
logger.error(f"Connection error: {e}")
self.is_connected = False
raise
async def disconnect(self):
"""接続を切断"""
if self.websocket:
await self.websocket.close()
self.is_connected = False
logger.info("Disconnected from server")
async def stream_audio_file(self, audio_path: str, source_image_path: str):
"""音声ファイルをストリーミング"""
try:
# 接続
await self.connect(source_image_path)
# 音声を読み込み
audio_data, sr = sf.read(audio_path)
if sr != self.sample_rate:
import librosa
audio_data = librosa.resample(
audio_data,
orig_sr=sr,
target_sr=self.sample_rate
)
# フレーム受信タスク
receive_task = asyncio.create_task(self._receive_frames())
# 音声をチャンク単位で送信
total_chunks = 0
for i in range(0, len(audio_data), self.chunk_size):
chunk = audio_data[i:i+self.chunk_size]
if len(chunk) < self.chunk_size:
chunk = np.pad(chunk, (0, self.chunk_size - len(chunk)))
# Float32として送信
await self.websocket.send(chunk.astype(np.float32).tobytes())
total_chunks += 1
# リアルタイムシミュレーション
await asyncio.sleep(self.chunk_duration)
# 進捗表示
progress = (i + self.chunk_size) / len(audio_data) * 100
logger.info(f"Streaming progress: {progress:.1f}%")
# 停止コマンドを送信
await self.websocket.send(json.dumps({"action": "stop"}))
logger.info(f"Sent {total_chunks} audio chunks")
# フレーム受信を待つ
await receive_task
finally:
await self.disconnect()
async def stream_microphone(self, source_image_path: str, duration: Optional[float] = None):
"""マイクからリアルタイムストリーミング"""
try:
# 接続
await self.connect(source_image_path)
# フレーム受信タスク
receive_task = asyncio.create_task(self._receive_frames())
# マイク録音用のキュー
audio_queue = queue.Queue()
stop_event = threading.Event()
# マイク録音スレッド
def record_audio():
p = pyaudio.PyAudio()
stream = p.open(
format=pyaudio.paFloat32,
channels=1,
rate=self.sample_rate,
input=True,
frames_per_buffer=self.chunk_size
)
logger.info("Recording started... Press Ctrl+C to stop")
try:
start_time = asyncio.get_event_loop().time()
while not stop_event.is_set():
if duration and (asyncio.get_event_loop().time() - start_time) > duration:
break
audio_chunk = stream.read(self.chunk_size, exception_on_overflow=False)
audio_queue.put(audio_chunk)
except Exception as e:
logger.error(f"Recording error: {e}")
finally:
stream.stop_stream()
stream.close()
p.terminate()
logger.info("Recording stopped")
# 録音スレッドを開始
record_thread = threading.Thread(target=record_audio)
record_thread.start()
try:
# 音声データを送信
while record_thread.is_alive() or not audio_queue.empty():
try:
audio_chunk = audio_queue.get(timeout=0.1)
audio_array = np.frombuffer(audio_chunk, dtype=np.float32)
await self.websocket.send(audio_array.tobytes())
except queue.Empty:
continue
except KeyboardInterrupt:
logger.info("Stopping recording...")
break
finally:
stop_event.set()
record_thread.join()
# 停止コマンドを送信
await self.websocket.send(json.dumps({"action": "stop"}))
# フレーム受信を待つ
await receive_task
finally:
await self.disconnect()
async def _receive_frames(self):
"""フレームとメッセージを受信"""
frame_count = 0
try:
while True:
message = await self.websocket.recv()
data = json.loads(message)
if data["type"] == "frame":
frame_count += 1
logger.info(f"Received frame {data['frame_id']} (FPS: {data.get('fps', 0)})")
if self.frame_callback:
# フレームをデコード
frame_data = base64.b64decode(data["data"])
nparr = np.frombuffer(frame_data, np.uint8)
frame = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
self.frame_callback(frame, data)
elif data["type"] == "progress":
logger.info(f"Progress: {data['duration_seconds']:.1f}s processed")
elif data["type"] == "processing":
logger.info(f"Server: {data['message']}")
elif data["type"] == "final_video":
logger.info(f"Received final video ({data['size_bytes']} bytes, {data['duration_seconds']:.1f}s)")
if self.final_video_callback:
video_data = base64.b64decode(data["data"])
self.final_video_callback(video_data, data)
break
elif data["type"] == "error":
logger.error(f"Server error: {data['message']}")
break
except websockets.exceptions.ConnectionClosed:
logger.info("Connection closed by server")
except Exception as e:
logger.error(f"Receive error: {e}")
logger.info(f"Total frames received: {frame_count}")
def set_frame_callback(self, callback: Callable):
"""フレーム受信時のコールバックを設定"""
self.frame_callback = callback
def set_final_video_callback(self, callback: Callable):
"""最終動画受信時のコールバックを設定"""
self.final_video_callback = callback
# 使用例とテスト
async def main():
"""使用例"""
client = DittoStreamingClient()
# フレーム表示用のコールバック
def display_frame(frame, metadata):
cv2.imshow("Live Frame", frame)
cv2.waitKey(1)
# 最終動画保存用のコールバック
def save_video(video_data, metadata):
output_path = "output_streaming.mp4"
with open(output_path, "wb") as f:
f.write(video_data)
logger.info(f"Video saved to {output_path}")
client.set_frame_callback(display_frame)
client.set_final_video_callback(save_video)
# テスト画像とサンプル音声のパス
source_image = "example/reference.png"
audio_file = "example/audio.wav"
# ファイルが存在するか確認
if not Path(source_image).exists():
logger.error(f"Source image not found: {source_image}")
return
# 音声ファイルからストリーミング
if Path(audio_file).exists():
logger.info("=== Testing audio file streaming ===")
await client.stream_audio_file(audio_file, source_image)
else:
logger.warning(f"Audio file not found: {audio_file}")
# マイクからストリーミング(5秒間)
# logger.info("\n=== Testing microphone streaming (5 seconds) ===")
# await client.stream_microphone(source_image, duration=5.0)
cv2.destroyAllWindows()
# バッチ処理クライアント
class BatchStreamingClient:
"""複数のリクエストを並列処理するクライアント"""
def __init__(self, server_url="ws://localhost:8000/ws/generate", max_parallel=3):
self.server_url = server_url
self.max_parallel = max_parallel
async def process_batch(self, tasks: list):
"""バッチ処理"""
semaphore = asyncio.Semaphore(self.max_parallel)
async def process_with_limit(task):
async with semaphore:
client = DittoStreamingClient(self.server_url)
await client.stream_audio_file(
task["audio_path"],
task["image_path"]
)
return task["id"]
results = await asyncio.gather(
*[process_with_limit(task) for task in tasks],
return_exceptions=True
)
return results
if __name__ == "__main__":
# 単一クライアントのテスト
asyncio.run(main())
# バッチ処理の例
# batch_client = BatchStreamingClient()
# tasks = [
# {"id": 1, "audio_path": "audio1.wav", "image_path": "image1.png"},
# {"id": 2, "audio_path": "audio2.wav", "image_path": "image2.png"},
# ]
# asyncio.run(batch_client.process_batch(tasks))