Better

requirements.txt

@@ -5,4 +5,5 @@ soundfile==0.12.1
 sphn==0.1.3
 torch==2.2.0
 websockets>=11.0.3, <13.1
-huggingface-hub==0.24.6
+huggingface-hub==0.24.6
+gradio

server_opus.py

@@ -16,6 +16,7 @@
 from websockets.server import serve
 
 from huggingface_hub import hf_hub_download
+from gradio import networking
 
 import msh
 
@@ -169,13 +170,18 @@ async def opus_loop():
                     log("info", f"frame handled in {1000 * (time.time() - be):.1f}ms")
 
         async def send_loop():
+            buf = []
             while True:
                 if close:
                     return
                 await asyncio.sleep(0.001)
                 msg = opus_writer.read_bytes()
                 if len(msg) > 0:
-                    await websocket.send(b"\x01" + msg)
+                    if msg[:4] == b'OggS':
+                        if buf:
+                            await websocket.send(b"\x01" + b''.join(buf))
+                            buf.clear()
+                    buf.append(msg)
 
 
         log("info", "accepted connection")
@@ -185,6 +191,7 @@ async def send_loop():
             opus_reader = sphn.OpusStreamReader(self.ec.sample_rate)
             self.ec.reset_streaming()
             self.lm_gen.reset_streaming()
+            await websocket.send(b'\x00')
             await asyncio.gather(opus_loop(), recv_loop(), send_loop())
         log("info", "done with connection")
 
@@ -193,6 +200,8 @@ async def main():
     state = ServerState()
     log("info", "warming up the model")
     state.warmup()
+    tunnel = networking.setup_tunnel('127.0.0.1', args.port, 'testlapin', None)
+    print("Tunnel", tunnel)
     log("info", f"listening to ws://{args.host}:{args.port}")
     async with serve(state.handle_conn, args.host, args.port):
         await asyncio.Future()  # run forever