livekit-plugins-elevenlabs 0.7.0.dev7__py3-none-any.whl → 0.7.2__py3-none-any.whl

livekit/plugins/elevenlabs/__init__.py

@@ -28,13 +28,12 @@ __all__ = [
 
 from livekit.agents import Plugin
 
+from .log import logger
+
 
 class ElevenLabsPlugin(Plugin):
     def __init__(self):
-        super().__init__(__name__, __version__, __package__)
-
-    def download_files(self):
-        pass
+        super().__init__(__name__, __version__, __package__, logger)
 
 
 Plugin.register_plugin(ElevenLabsPlugin())

livekit/plugins/elevenlabs/models.py

@@ -5,6 +5,7 @@ TTSModels = Literal[
     "eleven_multilingual_v1",
     "eleven_multilingual_v2",
     "eleven_turbo_v2",
+    "eleven_turbo_v2_5",
 ]
 
 TTSEncoding = Literal[

livekit/plugins/elevenlabs/tts.py

@@ -93,7 +93,7 @@ class TTS(tts.TTS):
         self,
         *,
         voice: Voice = DEFAULT_VOICE,
-        model_id: TTSModels = "eleven_turbo_v2",
+        model_id: TTSModels = "eleven_turbo_v2_5",
         api_key: str | None = None,
         base_url: str | None = None,
         encoding: TTSEncoding = "mp3_22050_32",
@@ -101,9 +101,7 @@ class TTS(tts.TTS):
         word_tokenizer: tokenize.WordTokenizer = tokenize.basic.WordTokenizer(
             ignore_punctuation=False  # punctuation can help for intonation
         ),
-        # default value of 11labs is [120, 160, 250, 290], but we want faster responses by default
-        # (range is 50-500)
-        chunk_length_schedule: list[int] = [80, 120, 200, 260],
+        chunk_length_schedule: list[int] = [80, 120, 200, 260],  # range is [50, 500]
         http_session: aiohttp.ClientSession | None = None,
     ) -> None:
         super().__init__(
@@ -158,6 +156,8 @@ class ChunkedStream(tts.ChunkedStream):
     ) -> None:
         super().__init__()
         self._text, self._opts, self._session = text, opts, session
+        if _encoding_from_format(self._opts.encoding) == "mp3":
+            self._mp3_decoder = utils.codecs.Mp3StreamDecoder()
 
     @utils.log_exceptions(logger=logger)
     async def _main_task(self) -> None:
@@ -183,21 +183,39 @@ class ChunkedStream(tts.ChunkedStream):
             headers={AUTHORIZATION_HEADER: self._opts.api_key},
             json=data,
         ) as resp:
-            async for bytes_data, _ in resp.content.iter_chunks():
-                for frame in bstream.write(bytes_data):
+            if not resp.content_type.startswith("audio/"):
+                content = await resp.text()
+                logger.error("11labs returned non-audio data: %s", content)
+                return
+            encoding = _encoding_from_format(self._opts.encoding)
+            if encoding == "mp3":
+                async for bytes_data, _ in resp.content.iter_chunks():
+                    for frame in self._mp3_decoder.decode_chunk(bytes_data):
+                        self._event_ch.send_nowait(
+                            tts.SynthesizedAudio(
+                                request_id=request_id,
+                                segment_id=segment_id,
+                                frame=frame,
+                            )
+                        )
+            else:
+                async for bytes_data, _ in resp.content.iter_chunks():
+                    for frame in bstream.write(bytes_data):
+                        self._event_ch.send_nowait(
+                            tts.SynthesizedAudio(
+                                request_id=request_id,
+                                segment_id=segment_id,
+                                frame=frame,
+                            )
+                        )
+
+                for frame in bstream.flush():
                     self._event_ch.send_nowait(
                         tts.SynthesizedAudio(
                             request_id=request_id, segment_id=segment_id, frame=frame
                         )
                     )
 
-            for frame in bstream.flush():
-                self._event_ch.send_nowait(
-                    tts.SynthesizedAudio(
-                        request_id=request_id, segment_id=segment_id, frame=frame
-                    )
-                )
-
 
 class SynthesizeStream(tts.SynthesizeStream):
     """Streamed API using websockets"""
@@ -208,8 +226,7 @@ class SynthesizeStream(tts.SynthesizeStream):
         opts: _TTSOptions,
     ):
         super().__init__()
-        self._opts = opts
-        self._session = session
+        self._opts, self._session = opts, session
         self._mp3_decoder = utils.codecs.Mp3StreamDecoder()
 
     @utils.log_exceptions(logger=logger)
@@ -222,31 +239,39 @@ class SynthesizeStream(tts.SynthesizeStream):
             word_stream = None
             async for input in self._input_ch:
                 if isinstance(input, str):
-                    if not word_stream:
+                    if word_stream is None:
+                        # new segment (after flush for e.g)
                         word_stream = self._opts.word_tokenizer.stream()
                         self._segments_ch.send_nowait(word_stream)
 
                     word_stream.push_text(input)
                 elif isinstance(input, self._FlushSentinel):
-                    word_stream.end_input()
+                    if word_stream is not None:
+                        word_stream.end_input()
+
                     word_stream = None
 
             self._segments_ch.close()
 
+        @utils.log_exceptions(logger=logger)
         async def _run():
             async for word_stream in self._segments_ch:
                 await self._run_ws(word_stream)
 
-        await asyncio.gather(_tokenize_input(), _run(), return_exceptions=True)
+        tasks = [
+            asyncio.create_task(_tokenize_input()),
+            asyncio.create_task(_run()),
+        ]
+        try:
+            await asyncio.gather(*tasks)
+        finally:
+            await utils.aio.gracefully_cancel(*tasks)
 
     async def _run_ws(
         self,
         word_stream: tokenize.WordStream,
-        max_retry: int = 1,
+        max_retry: int = 3,
     ) -> None:
-        request_id = utils.shortuuid()
-        segment_id = utils.shortuuid()
-
         ws_conn: aiohttp.ClientWebSocketResponse | None = None
         for try_i in range(max_retry):
             retry_delay = 5
@@ -268,6 +293,10 @@ class SynthesizeStream(tts.SynthesizeStream):
         if ws_conn is None:
             raise Exception(f"failed to connect to 11labs after {max_retry} retries")
 
+        request_id = utils.shortuuid()
+        segment_id = utils.shortuuid()
+
+        # 11labs protocol expects the first message to be an "init msg"
         init_pkt = dict(
             text=" ",
             try_trigger_generation=True,
@@ -291,7 +320,6 @@ class SynthesizeStream(tts.SynthesizeStream):
                     text=f"{data.token} ",  # must always end with a space
                     try_trigger_generation=False,
                 )
-                print(data_pkt)
                 await ws_conn.send_str(json.dumps(data_pkt))
 
             # no more token, mark eos
@@ -300,6 +328,8 @@ class SynthesizeStream(tts.SynthesizeStream):
             eos_sent = True
 
         async def recv_task():
+            nonlocal eos_sent
+
             while True:
                 msg = await ws_conn.receive()
                 if msg.type in (
@@ -378,11 +408,11 @@ def _synthesize_url(opts: _TTSOptions) -> str:
     base_url = opts.base_url
     voice_id = opts.voice.id
     model_id = opts.model_id
-    sample_rate = _sample_rate_from_format(opts.encoding)
+    output_format = opts.encoding
     latency = opts.streaming_latency
     return (
         f"{base_url}/text-to-speech/{voice_id}/stream?"
-        f"model_id={model_id}&output_format=pcm_{sample_rate}&optimize_streaming_latency={latency}"
+        f"model_id={model_id}&output_format={output_format}&optimize_streaming_latency={latency}"
     )
 
 

livekit/plugins/elevenlabs/version.py

@@ -12,4 +12,4 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-__version__ = "0.7.0-dev.7"
+__version__ = "0.7.2"

{livekit_plugins_elevenlabs-0.7.0.dev7.dist-info → livekit_plugins_elevenlabs-0.7.2.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: livekit-plugins-elevenlabs
-Version: 0.7.0.dev7
+Version: 0.7.2
 Summary: Agent Framework plugin for voice synthesis with ElevenLabs' API.
 Home-page: https://github.com/livekit/agents
 License: Apache-2.0

livekit_plugins_elevenlabs-0.7.2.dist-info/RECORD

@@ -0,0 +1,10 @@
+livekit/plugins/elevenlabs/__init__.py,sha256=cYRVIPXkRvB3-jK9bKZ9rYiMBACytWlCSq6yoZXaSgA,1080
+livekit/plugins/elevenlabs/log.py,sha256=hIuXqDsEB5GBa7rQY3z4Uqi1oCqc_lRmCHZEmXz0LHw,73
+livekit/plugins/elevenlabs/models.py,sha256=ddBUlDT4707f64WDJASR0B60X0yQ-LRHK1ZpTuBJXK8,387
+livekit/plugins/elevenlabs/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+livekit/plugins/elevenlabs/tts.py,sha256=g5w__bq5OdtZDjjclw3zYq4mAPMpkVgWPqlhkb_qpBg,14320
+livekit/plugins/elevenlabs/version.py,sha256=wNTnO8L3jrMdUjS-xAEFoMTKPaPYiFY9Kxnvzm4hTBc,600
+livekit_plugins_elevenlabs-0.7.2.dist-info/METADATA,sha256=WdOaTQBGsLgrjKQIM2_pgXLyPUqzBfTml14OFRv2qLQ,1311
+livekit_plugins_elevenlabs-0.7.2.dist-info/WHEEL,sha256=R0nc6qTxuoLk7ShA2_Y-UWkN8ZdfDBG2B6Eqpz2WXbs,91
+livekit_plugins_elevenlabs-0.7.2.dist-info/top_level.txt,sha256=OoDok3xUmXbZRvOrfvvXB-Juu4DX79dlq188E19YHoo,8
+livekit_plugins_elevenlabs-0.7.2.dist-info/RECORD,,

{livekit_plugins_elevenlabs-0.7.0.dev7.dist-info → livekit_plugins_elevenlabs-0.7.2.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: setuptools (71.1.0)
+Generator: setuptools (72.1.0)
 Root-Is-Purelib: true
 Tag: py3-none-any
 

livekit_plugins_elevenlabs-0.7.0.dev7.dist-info/RECORD

@@ -1,10 +0,0 @@
-livekit/plugins/elevenlabs/__init__.py,sha256=ez1ybDPt7GfKAKgPkxZFRB7Vyd-_i-0hfUMI79GQ5w4,1091
-livekit/plugins/elevenlabs/log.py,sha256=hIuXqDsEB5GBa7rQY3z4Uqi1oCqc_lRmCHZEmXz0LHw,73
-livekit/plugins/elevenlabs/models.py,sha256=8jTchztgpiTokHEaWUK8PPxWWfvm5SMrOGsJpzxbYAw,362
-livekit/plugins/elevenlabs/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-livekit/plugins/elevenlabs/tts.py,sha256=HpaHJQysUhThDdlYDHpQxroo9L2_m6G6QBAaNXs04K4,13032
-livekit/plugins/elevenlabs/version.py,sha256=z7YbosRr6jiTE2IIvHSRWSl4-yyS21CiHE5WD547wJo,606
-livekit_plugins_elevenlabs-0.7.0.dev7.dist-info/METADATA,sha256=7yOB_7fjkuBAXd6RRWzO6n8FnERRNUT3FP1uhn3-JtY,1316
-livekit_plugins_elevenlabs-0.7.0.dev7.dist-info/WHEEL,sha256=Wyh-_nZ0DJYolHNn1_hMa4lM7uDedD_RGVwbmTjyItk,91
-livekit_plugins_elevenlabs-0.7.0.dev7.dist-info/top_level.txt,sha256=OoDok3xUmXbZRvOrfvvXB-Juu4DX79dlq188E19YHoo,8
-livekit_plugins_elevenlabs-0.7.0.dev7.dist-info/RECORD,,