From d6303beebb64ea2e6011af710b550e8de824e8d7 Mon Sep 17 00:00:00 2001
From: Markus Hennerbichler <markush@speechmatics.com>
Date: Thu, 29 Feb 2024 13:34:53 +0000
Subject: [PATCH] Add example on how to read and transcribe microphone input
 (#88)

---
 examples/README.md                     |  8 ++-
 examples/transcribe_from_microphone.py | 71 ++++++++++++++++++++++++++
 2 files changed, 77 insertions(+), 2 deletions(-)
 create mode 100644 examples/transcribe_from_microphone.py

diff --git a/examples/README.md b/examples/README.md
index c3c3be0..10ba352 100644
--- a/examples/README.md
+++ b/examples/README.md
@@ -1,6 +1,8 @@
 # Examples
 
-This folder provides some examples of how the Speechmatics python client can be used to build different systems. The current examples include:
+This folder provides some examples of how the Speechmatics python client can be used to build different systems.
+Each of the examples should have a separate README with all the necessary steps to get them up and running.
+The current examples include:
 
 1. [notification_flow](./notification_flow/README.md) (webhooks)
 
@@ -18,4 +20,6 @@ Demonstrates how to run a websocket server that acts as a proxy to a speechmatic
 
 Tools for batch synchronising a local folder of audio into a searchable database of transcriptions.
 
-Each of the examples should have a separate README with all the necessary steps to get them up and running.
+5. [Microphone transcription example](./transcribe_from_microphone.py)
+
+This shows how the `sounddevice` python package can be used to stream audio from a microphone to Speechmatics.
diff --git a/examples/transcribe_from_microphone.py b/examples/transcribe_from_microphone.py
new file mode 100644
index 0000000..471bff2
--- /dev/null
+++ b/examples/transcribe_from_microphone.py
@@ -0,0 +1,71 @@
+import speechmatics
+import speechmatics.models
+import speechmatics.client
+import speechmatics.cli
+import asyncio
+import argparse
+import sys
+import sounddevice as sd
+
+
+class RawInputStreamWrapper:
+    def __init__(self, wrapped: sd.RawInputStream):
+        self.wrapped: sd.RawInputStream = wrapped
+
+    def read(self, frames):
+        return bytes(self.wrapped.read(frames)[0])
+
+
+async def transcribe_from_device(device, speechmatics_client, language: str, max_delay):
+    frame_rate = 44_100
+    with sd.RawInputStream(
+        device=device, channels=1, samplerate=frame_rate, dtype="float32"
+    ) as stream:
+        settings = speechmatics.models.AudioSettings(
+            sample_rate=frame_rate,
+            encoding="pcm_f32" + ("le" if sys.byteorder == "little" else "be"),
+        )
+
+        conf = speechmatics.models.TranscriptionConfig(
+            language=language,
+            operating_point="enhanced",
+            max_delay=max_delay,
+            enable_partials=True,
+            enable_entities=True,
+        )
+        await speechmatics_client.run(RawInputStreamWrapper(stream), conf, settings)
+
+
+def main(args):
+    speechmatics_client = speechmatics.client.WebsocketClient(
+        connection_settings_or_auth_token=args.auth_token
+    )
+    transcripts = speechmatics.cli.Transcripts(text="", json=[])
+    speechmatics.cli.add_printing_handlers(speechmatics_client, transcripts)
+
+    asyncio.run(
+        transcribe_from_device(
+            args.device, speechmatics_client, args.language, args.max_delay
+        )
+    )
+
+
+def int_or_str(text):
+    try:
+        return int(text)
+    except ValueError:
+        return text
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(
+        description="Speechmatics Microphone Realtime Transcription example"
+    )
+    parser.add_argument(
+        "-d", "--device", type=int_or_str, help="input device (numeric ID or substring)"
+    )
+    parser.add_argument("-a", "--auth_token", type=str, required=True)
+    parser.add_argument("-l", "--language", type=str, default="en")
+    parser.add_argument("-m", "--max_delay", type=float, default=2.0)
+
+    main(parser.parse_args())