livekit-plugins-elevenlabs 0.1.2__py3-none-any.whl → 0.1.4__py3-none-any.whl

livekit/plugins/elevenlabs/tts.py

@@ -63,6 +63,8 @@ class TTSOptions:
     voice: Voice
     model_id: TTSModels
     base_url: str
+    sample_rate: int
+    latency: int
 
 
 class TTS(tts.TTS):
@@ -73,6 +75,8 @@ class TTS(tts.TTS):
         model_id: TTSModels = "eleven_multilingual_v2",
         api_key: Optional[str] = None,
         base_url: Optional[str] = None,
+        sample_rate: int = 24000,
+        latency: int = 2,
     ) -> None:
         super().__init__(streaming_supported=True)
         api_key = api_key or os.environ.get("ELEVEN_API_KEY")
@@ -85,6 +89,8 @@ class TTS(tts.TTS):
             model_id=model_id,
             api_key=api_key,
             base_url=base_url or API_BASE_URL_V1,
+            sample_rate=sample_rate,
+            latency=latency,
         )
 
     async def list_voices(self) -> List[Voice]:
@@ -134,11 +140,9 @@ class SynthesizeStream(tts.SynthesizeStream):
         self,
         session: aiohttp.ClientSession,
         config: TTSOptions,
-        latency: int = 2,  # [1-4] the higher the more optimized for streaming latency
     ):
         self._config = config
         self._session = session
-        self._latency = latency
 
         self._queue = asyncio.Queue[str]()
         self._event_queue = asyncio.Queue[tts.SynthesisEvent]()
@@ -157,7 +161,7 @@ class SynthesizeStream(tts.SynthesizeStream):
         base_url = self._config.base_url
         voice_id = self._config.voice.id
         model_id = self._config.model_id
-        return f"{base_url}/text-to-speech/{voice_id}/stream-input?model_id={model_id}&output_format=pcm_44100&optimize_streaming_latency={self._latency}"
+        return f"{base_url}/text-to-speech/{voice_id}/stream-input?model_id={model_id}&output_format=pcm_{self._config.sample_rate}&optimize_streaming_latency={self._config.latency}"
 
     def push_text(self, token: str) -> None:
         if self._closed:
@@ -180,6 +184,7 @@ class SynthesizeStream(tts.SynthesizeStream):
         retry_count = 0
         listen_task: Optional[asyncio.Task] = None
         ws: Optional[aiohttp.ClientWebSocketResponse] = None
+        retry_text_queue: asyncio.Queue[str] = asyncio.Queue()
         while True:
             try:
                 ws = await self._try_connect()
@@ -190,7 +195,13 @@ class SynthesizeStream(tts.SynthesizeStream):
                 # forward queued text to 11labs
                 started = False
                 while not ws.closed:
-                    text = await self._queue.get()
+                    text = None
+                    if retry_text_queue.empty():
+                        text = await retry_text_queue.get()
+                        retry_text_queue.task_done()
+                    else:
+                        text = await self._queue.get()
+
                     if not started:
                         self._event_queue.put_nowait(
                             tts.SynthesisEvent(type=tts.SynthesisEventType.STARTED)
@@ -200,7 +211,19 @@ class SynthesizeStream(tts.SynthesizeStream):
                         text=text,
                         try_trigger_generation=True,
                     )
-                    await ws.send_str(json.dumps(text_packet))
+
+                    # This case can happen in normal operation because 11labs will not
+                    # keep connections open indefinitely if we are not sending data.
+                    try:
+                        await ws.send_str(json.dumps(text_packet))
+                    except Exception:
+                        await retry_text_queue.put(text)
+                        break
+
+                    # We call self._queue.task_done() even if we are retrying the text because
+                    # all text has gone through self._queue. An exception may have short-circuited
+                    # out of the loop so task_done() will not have already been called on text that
+                    # is being retried.
                     self._queue.task_done()
                     if text == STREAM_EOS:
                         await listen_task
@@ -265,7 +288,7 @@ class SynthesizeStream(tts.SynthesizeStream):
                 data = base64.b64decode(msg["audio"])
                 audio_frame = rtc.AudioFrame(
                     data=data,
-                    sample_rate=44100,
+                    sample_rate=self._config.sample_rate,
                     num_channels=1,
                     samples_per_channel=len(data) // 2,
                 )
@@ -277,6 +300,8 @@ class SynthesizeStream(tts.SynthesizeStream):
                 )
             elif msg.get("isFinal"):
                 break
+            else:
+                logging.error(f"Unhandled message from ElevenLabs: {msg}")
 
     async def flush(self) -> None:
         self._queue.put_nowait(self._text + " ")

livekit/plugins/elevenlabs/version.py

@@ -12,4 +12,4 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-__version__ = "0.1.2"
+__version__ = "0.1.4"

{livekit_plugins_elevenlabs-0.1.2.dist-info → livekit_plugins_elevenlabs-0.1.4.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: livekit-plugins-elevenlabs
-Version: 0.1.2
+Version: 0.1.4
 Summary: Agent Framework plugin for voice synthesis with ElevenLabs' API.
 Home-page: https://github.com/livekit/agents
 License: Apache-2.0

livekit_plugins_elevenlabs-0.1.4.dist-info/RECORD

@@ -0,0 +1,8 @@
+livekit/plugins/elevenlabs/__init__.py,sha256=-FQ-hnTqqbaVFa0sEu8CwInVp9vzkt-nRWkd34ruFFk,977
+livekit/plugins/elevenlabs/models.py,sha256=g46mCMMHP3x3qtHmybHHMcid1UwmjKCcF0T4IWjMjWE,163
+livekit/plugins/elevenlabs/tts.py,sha256=K3dg8En-GX6-pKxioSeEP1jTsIOuPaDAfgYvC_itE4k,11110
+livekit/plugins/elevenlabs/version.py,sha256=JisuVeJTYHFXsPWF9pf3j_4J8bsnqlKqN-xXzq_6vhI,600
+livekit_plugins_elevenlabs-0.1.4.dist-info/METADATA,sha256=wZTATw_9t4d1x-UsahhpwFITfX_J6c1kxvaU5Izq8kM,1360
+livekit_plugins_elevenlabs-0.1.4.dist-info/WHEEL,sha256=oiQVh_5PnQM0E3gPdiz09WCNmwiHDMaGer_elqB3coM,92
+livekit_plugins_elevenlabs-0.1.4.dist-info/top_level.txt,sha256=OoDok3xUmXbZRvOrfvvXB-Juu4DX79dlq188E19YHoo,8
+livekit_plugins_elevenlabs-0.1.4.dist-info/RECORD,,

livekit_plugins_elevenlabs-0.1.2.dist-info/RECORD

@@ -1,8 +0,0 @@
-livekit/plugins/elevenlabs/__init__.py,sha256=-FQ-hnTqqbaVFa0sEu8CwInVp9vzkt-nRWkd34ruFFk,977
-livekit/plugins/elevenlabs/models.py,sha256=g46mCMMHP3x3qtHmybHHMcid1UwmjKCcF0T4IWjMjWE,163
-livekit/plugins/elevenlabs/tts.py,sha256=kGFh5yCdAxss97wf-Z3mfQtxs7V8wXJmmnKlsOGTe30,9975
-livekit/plugins/elevenlabs/version.py,sha256=S3xxF-H96nScSv_7l7IUvROJ0avu9oz5Gm6j673md_Y,600
-livekit_plugins_elevenlabs-0.1.2.dist-info/METADATA,sha256=1R3mr6tpE6KWEyzseGy80VOifmjSdcv9vasuIKDk-Hs,1360
-livekit_plugins_elevenlabs-0.1.2.dist-info/WHEEL,sha256=oiQVh_5PnQM0E3gPdiz09WCNmwiHDMaGer_elqB3coM,92
-livekit_plugins_elevenlabs-0.1.2.dist-info/top_level.txt,sha256=OoDok3xUmXbZRvOrfvvXB-Juu4DX79dlq188E19YHoo,8
-livekit_plugins_elevenlabs-0.1.2.dist-info/RECORD,,