Open-source speech recognition toolkit for training, inference, streaming ASR, VAD, punctuation, speaker diarization pipelines, and OpenAI-compatible/MCP serving.
Python asraudiochineseemotion-recognitionfunasrmcp-servermultilingual-asropenai-compatible-apiparaformerpunctuationpytorchreal-time-asrspeaker-diarizationspeech-recognitionspeech-to-textstreaming-asrtranscriptionvllmvoice-activity-detectionwhisper-alternative
websocket
游雁e2154e459963
diff --git a/runtime/python/websocket/funasr_client_api.py b/runtime/python/websocket/funasr_client_api.pyindex d0992ff5b..093ed5186 100644--- a/runtime/python/websocket/funasr_client_api.py+++ b/runtime/python/websocket/funasr_client_api.py@@ -5,7 +5,7 @@2022-2023 by [email protected]'''# pip install websocket-client-import ssl+import sslfrom websocket import ABNFfrom websocket import create_connectionfrom queue import Queue@@ -20,7 +20,13 @@ class Funasr_websocket_recognizer():python asr recognizer lib'''- def __init__(self, host="127.0.0.1", port="30035", is_ssl=True,chunk_size="5, 10, 5",chunk_interval=10,mode="offline",wav_name="default"):+ def __init__(self, host="127.0.0.1",+ port="30035",+ is_ssl=True,+ chunk_size="0, 10, 5",+ chunk_interval=10,+ mode="offline",+ wav_name="default"):'''host: server host ipport: server port@@ -43,17 +49,21 @@ class Funasr_websocket_recognizer():self.msg_queue = Queue() # used for recognized result textprint("connect to url",uri)- self.websocket=create_connection(uri,ssl=ssl_context,sslopt=ssl_opt)+ self.websocket=create_connection(uri, ssl=ssl_context, sslopt=ssl_opt)- self.thread_msg = threading.Thread(target=Funasr_websocket_recognizer.thread_rec_msg,args=(self,))+ self.thread_msg = threading.Thread(target=Funasr_websocket_recognizer.thread_rec_msg, args=(self,))self.thread_msg.start()chunk_size = [int(x) for x in chunk_size.split(",")]- stride = int(60 * chunk_size[1]/ chunk_interval / 1000 * 16000 * 2)+ stride = int(60 * chunk_size[1] / chunk_interval / 1000 * 16000 * 2)chunk_num = (len(audio_bytes) - 1) // stride + 1- message = json.dumps({"mode": args.mode, "chunk_size": args.chunk_size, "encoder_chunk_look_back": 4,- "decoder_chunk_look_back": 1, "chunk_interval": args.chunk_interval,- "wav_name": wav_name, "is_speaking": True})+ message = json.dumps({"mode": mode,+ "chunk_size": chunk_size,+ "encoder_chunk_look_back": 4,+ "decoder_chunk_look_back": 1,+ "chunk_interval": chunk_interval,+ "wav_name": wav_name,+ "is_speaking": True})self.websocket.send(message)@@ -68,7 +78,7 @@ class Funasr_websocket_recognizer():try:while(True):msg=self.websocket.recv()- if msg is None or len(msg)==0:+ if msg is None or len(msg) == 0:continuemsg = json.loads(msg)@@ -77,7 +87,7 @@ class Funasr_websocket_recognizer():print("client closed")# feed data to asr engine, wait_time means waiting for result until time out- def feed_chunk(self, chunk,wait_time=0.01):+ def feed_chunk(self, chunk, wait_time=0.01):try:self.websocket.send(chunk, ABNF.OPCODE_BINARY)# loop to check if there is a message, timeout in 0.01s@@ -89,6 +99,7 @@ class Funasr_websocket_recognizer():return msgexcept:return ""+def close(self,timeout=1):message = json.dumps({"is_speaking": False})self.websocket.send(message)@@ -103,9 +114,10 @@ class Funasr_websocket_recognizer():return msgif __name__ == '__main__':+print('example for Funasr_websocket_recognizer')import wave- wav_path="asr_example.wav"+ wav_path = "/Users/zhifu/Downloads/modelscope_models/speech_seaco_paraformer_large_asr_nat-zh-cn-16k-common-vocab8404-pytorch/example/asr_example.wav"with wave.open(wav_path, "rb") as wav_file:params = wav_file.getparams()frames = wav_file.readframes(wav_file.getnframes())@@ -115,20 +127,24 @@ if __name__ == '__main__':stride = int(60 * 10 / 10 / 1000 * 16000 * 2)chunk_num = (len(audio_bytes) - 1) // stride + 1# create an recognizer- rcg=Funasr_websocket_recognizer(host="127.0.0.1",port="30035",is_ssl=True,mode="2pass")+ rcg = Funasr_websocket_recognizer(host="127.0.0.1",+ port="10095",+ is_ssl=True,+ mode="2pass",+ chunk_size="0,10,5")# loop to send chunkfor i in range(chunk_num):beg = i * stridedata = audio_bytes[beg:beg + stride]- text=rcg.feed_chunk(data,wait_time=0.02)+ text = rcg.feed_chunk(data,wait_time=0.02)if len(text)>0:print("text",text)time.sleep(0.05)# get last message- text=rcg.close(timeout=3)+ text = rcg.close(timeout=3)print("text",text)diff --git a/runtime/python/websocket/funasr_wss_server.py b/runtime/python/websocket/funasr_wss_server.pyindex 7db7b0ac7..fa82ea525 100644--- a/runtime/python/websocket/funasr_wss_server.py+++ b/runtime/python/websocket/funasr_wss_server.py@@ -152,7 +152,7 @@ async def ws_serve(websocket, path):frames_asr = []frames_asr_online = []global websocket_users- await clear_websocket()+ # await clear_websocket()websocket_users.add(websocket)websocket.status_dict_asr = {}websocket.status_dict_asr_online = {"cache": {}, "is_final": False}