diff --git a/examples/foundational/websocket-server/index.html b/examples/foundational/websocket-server/index.html
index 4a8bb5999..f94ab688c 100644
--- a/examples/foundational/websocket-server/index.html
+++ b/examples/foundational/websocket-server/index.html
@@ -4,6 +4,7 @@
 <head>
     <meta charset="UTF-8">
     <meta name="viewport" content="width=device-width, initial-scale=1.0">
+    <script src="//cdn.jsdelivr.net/npm/protobufjs@7.X.X/dist/protobuf.min.js"></script>
     <title>WebSocket Audio Stream</title>
 </head>
 
@@ -14,12 +15,10 @@ <h1>WebSocket Audio Stream</h1>
     <script>
         let audioContext;
         let microphoneStream;
-        let mediaRecorder;
-        let refresher;
         let audioCtx;
 
 
-        function startAudio() {
+        async function startAudio() {
             audioCtx = new (window.AudioContext || window.webkitAudioContext)();
 
             if (!navigator.mediaDevices || !navigator.mediaDevices.getUserMedia) {
@@ -27,52 +26,40 @@ <h1>WebSocket Audio Stream</h1>
                 return;
             }
 
-            navigator.mediaDevices.getUserMedia({ audio: true })
-                .then(function (stream) {
-                    console.log("creating audio context")
-                    audioContext = new AudioContext(options = { sampleRate: 16000 });
-                    microphoneStream = audioContext.createMediaStreamSource(stream);
-                    mediaRecorder = new MediaRecorder(stream);
-                    const captureNode = audioContext.createScriptProcessor(8192, 1, 1);
-                    captureNode.onaudioprocess = function (e) {
-                        console.log("on audio process")
-                        debugger;
-                        var left = e.inputBuffer.getChannelData(0);
-                        ws.send(left.buffer);
-                    };
-                    /*
-                    mediaRecorder.ondataavailable = function (event) {
-                        console.log("ondataavailable")
-                        if (event.data.size > 0 && ws && ws.readyState === ws.OPEN) {
-                            console.log("Sending audio blob:", event.data.size, 'bytes')
-                            debugger
-                            ws.send(event.data);
-                        }
-                    };
-                    */
-                    mediaRecorder.start();
-
-                    refresher = setInterval(function () {
-                        console.log("sending audio")
-                        mediaRecorder.requestData();
-                    }, 1000);
-
-                })
-                .catch(function (error) {
-                    console.error('Error accessing microphone:', error);
-                });
+            // Get the microphone stream.
+            microphoneStream = await navigator.mediaDevices.getUserMedia({ audio: true });
+
+            // Create an AudioContext.
+            const context = new AudioContext();
+
+            // Create a ScriptProcessorNode.
+            const scriptProcessor = context.createScriptProcessor(8192, 1, 1);
+
+            // Connect the microphone stream to the ScriptProcessorNode.
+            const source = context.createMediaStreamSource(microphoneStream);
+            source.connect(scriptProcessor);
+
+            // Connect the ScriptProcessorNode to the destination.
+            scriptProcessor.connect(context.destination);
+
+            scriptProcessor.onaudioprocess = (event) => {
+                const rawLeftChannelData = event.inputBuffer.getChannelData(0);
+                scaledToInt = []
+                for (var i = 0; i < rawLeftChannelData.length; i++) {
+                    // Convert each item from -1.0-1.0 to a 16-bit signed integer
+                    scaledToInt[i] = (rawLeftChannelData[i] * 32767 + 32767) % 65535;
+                }
+                ws.send(scaledToInt);
+            };
+
 
             initWebSocket();
         }
 
         function stopAudio() {
-            if (mediaRecorder && mediaRecorder.state === 'recording') {
-                clearInterval(refresher);
-                mediaRecorder.stop();
-                microphoneStream.disconnect();
-                ws.close();
-                ws = undefined;
-            }
+            microphoneStream.disconnect();
+            ws.close();
+            ws = undefined;
         }
 
         function initWebSocket() {
diff --git a/examples/foundational/websocket-server/sample.py b/examples/foundational/websocket-server/sample.py
index 5f0d1fb61..f73aabde9 100644
--- a/examples/foundational/websocket-server/sample.py
+++ b/examples/foundational/websocket-server/sample.py
@@ -2,20 +2,21 @@
 import aiohttp
 import logging
 import os
+import wave
 
-from dailyai.pipeline.frames import TextFrame
+from dailyai.pipeline.frame_processor import FrameProcessor
+from dailyai.pipeline.frames import AudioFrame, TextFrame
 from dailyai.pipeline.pipeline import Pipeline
 from dailyai.services.elevenlabs_ai_service import ElevenLabsTTSService
 from dailyai.services.websocket_transport_service import WebsocketTransportService
-
-from runner import configure
+from dailyai.services.whisper_ai_services import WhisperSTTService
 
 logging.basicConfig(format=f"%(levelno)s %(asctime)s %(message)s")
-logger = logging.getLogger("websockets.server")
+logger = logging.getLogger("dailyai")
 logger.setLevel(logging.DEBUG)
 
 
-async def main(room_url):
+async def main():
     async with aiohttp.ClientSession() as session:
         transport = WebsocketTransportService(
             mic_enabled=True, speaker_enabled=True)
@@ -26,7 +27,17 @@ async def main(room_url):
             voice_id=os.getenv("ELEVENLABS_VOICE_ID"),
         )
 
-        pipeline = Pipeline([tts])
+        class WriteToWav(FrameProcessor):
+            async def process_frame(self, frame):
+                if isinstance(frame, AudioFrame):
+                    with wave.open("output.wav", "wb") as f:
+                        f.setnchannels(1)
+                        f.setsampwidth(2)
+                        f.setframerate(16000)
+                        f.writeframes(frame.data)
+                yield frame
+
+        pipeline = Pipeline([WriteToWav(), WhisperSTTService(), tts])
 
         @transport.on_connection
         async def queue_frame():
@@ -37,5 +48,4 @@ async def queue_frame():
 
 
 if __name__ == "__main__":
-    (url, token) = configure()
-    asyncio.run(main(url))
+    asyncio.run(main())
diff --git a/output.wav b/output.wav
new file mode 100644
index 000000000..d26b87bdb
Binary files /dev/null and b/output.wav differ
diff --git a/src/dailyai/pipeline/frames.proto b/src/dailyai/pipeline/frames.proto
new file mode 100644
index 000000000..7ecea6d25
--- /dev/null
+++ b/src/dailyai/pipeline/frames.proto
@@ -0,0 +1,25 @@
+syntax = "proto3";
+
+package dailyai_proto;
+
+message TextFrame {
+    string text = 1;
+}
+
+message AudioFrame {
+    bytes audio = 1;
+}
+
+message TranscriptionFrame {
+    string text = 1;
+    string participant_id = 2;
+    string timestamp = 3;
+}
+
+message Frame {
+    oneof frame {
+        TextFrame text = 1;
+        AudioFrame audio = 2;
+        TranscriptionFrame transcription = 3;
+    }
+}
diff --git a/src/dailyai/pipeline/frames.py b/src/dailyai/pipeline/frames.py
index 6c39fb2c7..57694cb92 100644
--- a/src/dailyai/pipeline/frames.py
+++ b/src/dailyai/pipeline/frames.py
@@ -2,12 +2,19 @@
 from typing import Any, List
 
 from dailyai.services.openai_llm_context import OpenAILLMContext
+import dailyai.pipeline.protobufs.frames_pb2 as frame_protos
 
 
 class Frame:
     def __str__(self):
         return f"{self.__class__.__name__}"
 
+    def to_proto(self):
+        raise NotImplementedError
+
+    @staticmethod
+    def from_proto(proto):
+        raise NotImplementedError
 
 class ControlFrame(Frame):
     # Control frames should contain no instance data, so
@@ -61,12 +68,30 @@ class LLMResponseEndFrame(ControlFrame):
 @dataclass()
 class AudioFrame(Frame):
     """A chunk of audio. Will be played by the transport if the transport's mic
-    has been enabled."""
+    has been enabled.
+
+    >>> str(AudioFrame(data=b'1234567890'))
+    'AudioFrame, size: 10 B'
+
+    >>> AudioFrame.from_proto(AudioFrame(data=b'1234567890').to_proto())
+    AudioFrame(data=b'1234567890')
+    """
     data: bytes
 
     def __str__(self):
         return f"{self.__class__.__name__}, size: {len(self.data)} B"
 
+    def to_proto(self) -> frame_protos.Frame:
+        frame = frame_protos.Frame()
+        frame.audio.audio = self.data
+        return frame
+
+    @staticmethod
+    def from_proto(proto: frame_protos.Frame):
+        if proto.WhichOneof("frame") != "audio":
+            raise ValueError("Proto does not contain an audio frame")
+        return AudioFrame(data=proto.audio.audio)
+
 
 @dataclass()
 class ImageFrame(Frame):
@@ -93,12 +118,25 @@ def __str__(self):
 @dataclass()
 class TextFrame(Frame):
     """A chunk of text. Emitted by LLM services, consumed by TTS services, can
-    be used to send text through pipelines."""
+    be used to send text through pipelines.
+
+    >>> str(TextFrame.from_proto(TextFrame(text='hello world').to_proto()))
+    'TextFrame: "hello world"'
+    """
     text: str
 
     def __str__(self):
         return f'{self.__class__.__name__}: "{self.text}"'
 
+    def to_proto(self) -> frame_protos.Frame:
+        proto_frame = frame_protos.Frame()
+        proto_frame.text.text = self.text
+        return proto_frame
+
+    @staticmethod
+    def from_proto(proto: frame_protos.TextFrame):
+        return TextFrame(text=proto.text.text)
+
 
 @dataclass()
 class TranscriptionQueueFrame(TextFrame):
@@ -107,6 +145,21 @@ class TranscriptionQueueFrame(TextFrame):
     participantId: str
     timestamp: str
 
+    def to_proto(self) -> frame_protos.Frame:
+        frame = frame_protos.Frame()
+        frame.transcription.text = self.text
+        frame.transcription.participant_id = self.participantId
+        frame.transcription.timestamp = self.timestamp
+        return frame
+
+    @staticmethod
+    def from_proto(proto: frame_protos.Frame):
+        return TranscriptionQueueFrame(
+            text=proto.transcription.text,
+            participantId=proto.transcription.participant_id,
+            timestamp=proto.transcription.timestamp
+        )
+
 
 @dataclass()
 class LLMMessagesQueueFrame(Frame):
@@ -179,3 +232,21 @@ class LLMFunctionCallFrame(Frame):
     """Emitted when the LLM has received an entire function call completion."""
     function_name: str
     arguments: str
+
+
+if __name__ == "__main__":
+    audio_frame = AudioFrame(data=b'1234567890')
+    print(audio_frame)
+    print(audio_frame.to_proto().SerializeToString())
+    print(AudioFrame.from_proto(audio_frame.to_proto()))
+
+    text_frame = TextFrame(text="Hello there!")
+    print(text_frame)
+    print(text_frame.to_proto().SerializeToString())
+    print(TextFrame.from_proto(text_frame.to_proto()))
+
+    transcripton_frame = TranscriptionQueueFrame(
+        text="Hello there!", participantId="123", timestamp="2021-01-01")
+    print(transcripton_frame)
+    print(transcripton_frame.to_proto().SerializeToString())
+    print(TranscriptionQueueFrame.from_proto(transcripton_frame.to_proto()))
diff --git a/src/dailyai/pipeline/protobufs/frames_pb2.py b/src/dailyai/pipeline/protobufs/frames_pb2.py
new file mode 100644
index 000000000..fb4c2301c
--- /dev/null
+++ b/src/dailyai/pipeline/protobufs/frames_pb2.py
@@ -0,0 +1,32 @@
+# -*- coding: utf-8 -*-
+# Generated by the protocol buffer compiler.  DO NOT EDIT!
+# source: frames.proto
+# Protobuf Python Version: 4.25.3
+"""Generated protocol buffer code."""
+from google.protobuf import descriptor as _descriptor
+from google.protobuf import descriptor_pool as _descriptor_pool
+from google.protobuf import symbol_database as _symbol_database
+from google.protobuf.internal import builder as _builder
+# @@protoc_insertion_point(imports)
+
+_sym_db = _symbol_database.Default()
+
+
+
+
+DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n\x0c\x66rames.proto\x12\rdailyai_proto\"\x19\n\tTextFrame\x12\x0c\n\x04text\x18\x01 \x01(\t\"\x1b\n\nAudioFrame\x12\r\n\x05\x61udio\x18\x01 \x01(\x0c\"M\n\x12TranscriptionFrame\x12\x0c\n\x04text\x18\x01 \x01(\t\x12\x16\n\x0eparticipant_id\x18\x02 \x01(\t\x12\x11\n\ttimestamp\x18\x03 \x01(\t\"\xa2\x01\n\x05\x46rame\x12(\n\x04text\x18\x01 \x01(\x0b\x32\x18.dailyai_proto.TextFrameH\x00\x12*\n\x05\x61udio\x18\x02 \x01(\x0b\x32\x19.dailyai_proto.AudioFrameH\x00\x12:\n\rtranscription\x18\x03 \x01(\x0b\x32!.dailyai_proto.TranscriptionFrameH\x00\x42\x07\n\x05\x66rameb\x06proto3')
+
+_globals = globals()
+_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
+_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'frames_pb2', _globals)
+if _descriptor._USE_C_DESCRIPTORS == False:
+  DESCRIPTOR._options = None
+  _globals['_TEXTFRAME']._serialized_start=31
+  _globals['_TEXTFRAME']._serialized_end=56
+  _globals['_AUDIOFRAME']._serialized_start=58
+  _globals['_AUDIOFRAME']._serialized_end=85
+  _globals['_TRANSCRIPTIONFRAME']._serialized_start=87
+  _globals['_TRANSCRIPTIONFRAME']._serialized_end=164
+  _globals['_FRAME']._serialized_start=167
+  _globals['_FRAME']._serialized_end=329
+# @@protoc_insertion_point(module_scope)
diff --git a/src/dailyai/services/local_stt_service.py b/src/dailyai/services/local_stt_service.py
index 727bafbc6..98d37dc10 100644
--- a/src/dailyai/services/local_stt_service.py
+++ b/src/dailyai/services/local_stt_service.py
@@ -42,6 +42,7 @@ def _new_wave(self):
     async def process_frame(self, frame: Frame) -> AsyncGenerator[Frame, None]:
         """Processes a frame of audio data, either buffering or transcribing it."""
         if not isinstance(frame, AudioFrame):
+            yield frame
             return
 
         data = frame.data
diff --git a/src/dailyai/services/websocket_transport_service.py b/src/dailyai/services/websocket_transport_service.py
index e67e579e6..7e4ab1f3c 100644
--- a/src/dailyai/services/websocket_transport_service.py
+++ b/src/dailyai/services/websocket_transport_service.py
@@ -1,6 +1,7 @@
 import asyncio
 from concurrent.futures import Future
 import queue
+import wave
 import websockets
 
 from dailyai.services.base_transport_service import BaseTransportService
@@ -36,7 +37,6 @@ def write_frame_to_mic(self, frame: bytes):
     def read_audio_frames(self, desired_frame_count):
         audio_bytes = bytes()
         while not self._audio_queue.empty() and len(audio_bytes) < desired_frame_count:
-            print("getting audio out of queue")
             audio_bytes += self._audio_queue.get()
 
         return audio_bytes
@@ -52,7 +52,11 @@ async def _websocket_handler(self, websocket: websockets.WebSocketServerProtocol
 
         self._websocket = websocket
         async for message in websocket:
-            print("got message")
+            with wave.open("output.wav", "wb") as f:
+                f.setnchannels(1)
+                f.setsampwidth(2)
+                f.setframerate(16000)
+                f.writeframes(message)
             self._audio_queue.put(message)
 
     async def _start_server(self) -> None:
@@ -68,7 +72,6 @@ async def _start_server(self) -> None:
             print(f"Error in start server: {e}")
 
     async def _close_server(self):
-        print("_close_server")
         if self._websocket:
             await self._websocket.close()
 
diff --git a/src/dailyai/services/whisper_ai_services.py b/src/dailyai/services/whisper_ai_services.py
index cc657e6cf..2ed1c3563 100644
--- a/src/dailyai/services/whisper_ai_services.py
+++ b/src/dailyai/services/whisper_ai_services.py
@@ -52,4 +52,5 @@ async def run_stt(self, audio: BinaryIO) -> str:
         res: str = ""
         for segment in segments:
             res += f"{segment.text} "
+            print("Transcription: ", segment.text)
         return res
