Add speaker ID function

examples/rt/async/speaker_id/README.md

@@ -0,0 +1,15 @@
+# Live Real-Time Speaker ID Example
+
+This example demonstrates how to use the Speechmatics Python SDK to perform speaker ID in real-time.
+
+The SDK requires an API key to be set as an environment variable before it can be used. You can obtain an API key by signing up for a Speechmatics account at https://portal.speechmatics.com/dashboard
+
+## Prerequisites
+
+- Install Speechmatics RT SDK: `pip install speechmatics-rt`
+- Export Speechmatics API key: `export SPEECHMATICS_API_KEY=YOUR-API-KEY`
+
+## Usage
+
+- Generate speaker IDs: `python generate.py` - this will generate a `speakers.json` file
+- Transcribe audio: `python transcribe.py` - this will use the `speakers.json` file to perform speaker ID on a conversation

examples/rt/async/speaker_id/generate.py

@@ -0,0 +1,56 @@
+import asyncio
+import logging
+import os
+import json
+
+from speechmatics.rt import (
+    AsyncClient,
+    OperatingPoint,
+    TranscriptionConfig,
+    ServerMessageType,
+)
+
+
+logging.basicConfig(level=logging.INFO)
+
+
+speakers: list[dict] = []
+
+
+async def generate_ids(voice_file: str) -> None:
+    """Run async transcription example."""
+
+    transcription_config = TranscriptionConfig(
+        operating_point=OperatingPoint.ENHANCED,
+        diarization="speaker",
+    )
+
+    # Initialize client with API key from environment
+    async with AsyncClient() as client:
+        try:
+            @client.on(ServerMessageType.SPEAKERS_RESULT)
+            def handle_speakers_result(msg):
+                new_speakers = msg.get('speakers', [])
+                new_speakers[0]["label"] = voice_file
+                speakers.append(new_speakers[0])
+
+            # Transcribe audio file
+            with open(os.path.join(voices_folder, voice_file), "rb") as audio_file:
+                await client.transcribe(
+                    audio_file,
+                    transcription_config=transcription_config,
+                    get_speakers=True,
+                )
+        except Exception as e:
+            print(f"Transcription error: {e}")
+
+
+if __name__ == "__main__":
+    voices_folder = "./examples/rt/async/speaker_id/voices"
+    voice_files = [f for f in os.listdir(voices_folder) if os.path.isfile(os.path.join(voices_folder, f))]
+
+    for voice_file in voice_files:
+        asyncio.run(generate_ids(voice_file))
+
+    with open('./speakers.json', 'w') as f:
+        json.dump(speakers, f)

examples/rt/async/speaker_id/transcribe.py

@@ -0,0 +1,51 @@
+import asyncio
+import logging
+import json
+
+from speechmatics.rt import SpeakerIdentifier
+from speechmatics.rt import SpeakerDiarizationConfig
+from speechmatics.rt import (
+    AsyncClient,
+    OperatingPoint,
+    TranscriptionConfig,
+    ServerMessageType
+)
+
+
+logging.basicConfig(level=logging.INFO)
+
+
+async def main() -> None:
+    """Run async transcription example."""
+
+    with open('./speakers.json') as f:
+        speaker_identifiers = [SpeakerIdentifier(**s) for s in json.load(f)]
+
+    transcription_config = TranscriptionConfig(
+        operating_point=OperatingPoint.ENHANCED,
+        diarization="speaker",
+        max_delay=4,
+        speaker_diarization_config=SpeakerDiarizationConfig(
+            speakers=speaker_identifiers,
+        )
+    )
+
+    # Initialize client with API key from environment
+    async with AsyncClient() as client:
+        try:
+            @client.on(ServerMessageType.ADD_TRANSCRIPT)
+            def handle_finals(msg):
+                print(f"Final: {msg['metadata']['speaker']} {msg['metadata']['transcript']}")
+
+            # Transcribe audio file
+            with open("./examples/conversation.wav", "rb") as audio_file:
+                await client.transcribe(
+                    audio_file,
+                    transcription_config=transcription_config,
+                )
+        except Exception as e:
+            print(f"Transcription error: {e}")
+
+
+if __name__ == "__main__":
+    asyncio.run(main())

sdk/batch/README.md

@@ -25,6 +25,8 @@ pip install speechmatics-batch
 
 ### Quick Start
 
+To run transcription, you'll need an audio file. You can find an example file [here](https://github.com/speechmatics/speechmatics-python-sdk/blob/main/examples/example.wav).
+
 ```python
 import asyncio
 from speechmatics.batch import AsyncClient

sdk/rt/README.md

@@ -24,6 +24,8 @@ pip install speechmatics-rt
 ```
 ## Quick Start
 
+To run transcription, you'll need an audio file. You can find an example file [here](https://github.com/speechmatics/speechmatics-python-sdk/blob/main/examples/example.wav).
+
 ```python
 import asyncio
 from speechmatics.rt import AsyncClient, ServerMessageType

sdk/rt/speechmatics/rt/_async_client.py

@@ -171,6 +171,7 @@ async def transcribe(
         audio_events_config: Optional[AudioEventsConfig] = None,
         ws_headers: Optional[dict] = None,
         timeout: Optional[float] = None,
+        get_speakers: Optional[bool] = False,
     ) -> None:
         """
         Transcribe a single audio stream in real-time.
@@ -193,6 +194,7 @@ async def transcribe(
             ws_headers: Additional headers to include in the WebSocket handshake.
             timeout: Maximum time in seconds to wait for transcription completion.
                     Default None.
+            get_speakers: Send a speaker identifier event at the end of the session.
 
         Raises:
             AudioError: If source is invalid or cannot be read.
@@ -233,6 +235,9 @@ async def transcribe(
             ws_headers=ws_headers,
         )
 
+        if get_speakers:
+            await self.send_message({"message": "GetSpeakers", "final": True})
+
         try:
             await asyncio.wait_for(
                 self._audio_producer(source, audio_format.chunk_size),

sdk/rt/speechmatics/rt/_base_client.py

@@ -18,7 +18,9 @@
 from ._models import AudioEventsConfig
 from ._models import AudioFormat
 from ._models import ConnectionConfig
+from ._models import ServerMessageType
 from ._models import SessionInfo
+from ._models import SpeakerIdentifier
 from ._models import TranscriptionConfig
 from ._models import TranslationConfig
 from ._transport import Transport
@@ -149,6 +151,36 @@ async def send_message(self, message: dict[str, Any]) -> None:
             self._closed_evt.set()
             raise
 
+    async def get_speakers(self, final=False) -> list[SpeakerIdentifier]:
+        """
+        Get the list of speakers in the current session.
+        This method returns as soon as a SPEAKERS_RESULT message is received.
+        Multiple requests to the method may therefore cause a race condition in which the same
+        SPEAKERS_RESULT message is received by multiple requests. This should not cause any issues,
+        but will result in redundant SPEAKERS_RESULT events.
+
+        Args:
+            final: Whether to wait to the end of the session to return speaker IDs (default: False)
+
+        Returns:
+            List of SpeakerIdentifier objects
+        """
+        try:
+            await self.send_message({"message": "GetSpeakers", "final": final})
+            speaker_evt = asyncio.Event()
+            speaker_identifiers: list[SpeakerIdentifier] = []
+            self.once(
+                ServerMessageType.SPEAKERS_RESULT,
+                lambda msg: (speaker_identifiers.extend(msg.get("speakers", [])), speaker_evt.set()),
+            )
+            await speaker_evt.wait()
+            return speaker_identifiers
+        except asyncio.TimeoutError:
+            raise TransportError("Timeout waiting for SPEAKERS_RESULT")
+        except Exception:
+            self._closed_evt.set()
+            raise
+
     async def _recv_loop(self) -> None:
         """
         Background task that continuously receives and dispatches server messages.