wyoming-piper 1.5.3__py3-none-any.whl → 1.6.3__py3-none-any.whl

wyoming_piper/__init__.py

@@ -1,4 +1,5 @@
 """Wyoming server for piper."""
+
 from importlib.metadata import version
 
 __version__ = version("wyoming_piper")

wyoming_piper/__main__.py

@@ -60,6 +60,11 @@ async def main() -> None:
         default=1,
         help="Maximum number of piper process to run simultaneously (default: 1)",
     )
+    parser.add_argument(
+        "--streaming",
+        action="store_true",
+        help="Enable audio streaming on sentence boundaries",
+    )
     #
     parser.add_argument(
         "--update-voices",
@@ -113,12 +118,14 @@ async def main() -> None:
                     voice_info.get("espeak", {}).get("voice", voice_name.split("_")[0]),
                 )
             ],
-            speakers=[
-                TtsVoiceSpeaker(name=speaker_name)
-                for speaker_name in voice_info["speaker_id_map"]
-            ]
-            if voice_info.get("speaker_id_map")
-            else None,
+            speakers=(
+                [
+                    TtsVoiceSpeaker(name=speaker_name)
+                    for speaker_name in voice_info["speaker_id_map"]
+                ]
+                if voice_info.get("speaker_id_map")
+                else None
+            ),
         )
         for voice_name, voice_info in voices_info.items()
         if not voice_info.get("_is_alias", False)
@@ -180,6 +187,7 @@ async def main() -> None:
                 installed=True,
                 voices=sorted(voices, key=lambda v: v.name),
                 version=__version__,
+                supports_synthesize_streaming=args.streaming,
             )
         ],
     )

wyoming_piper/download.py

@@ -1,4 +1,5 @@
 """Utility for downloading Piper voices."""
+
 import json
 import logging
 import shutil
@@ -10,7 +11,7 @@ from urllib.request import urlopen
 
 from .file_hash import get_file_hash
 
-URL_FORMAT = "https://huggingface.co/rhasspy/piper-voices/resolve/v1.0.0/{file}"
+URL_FORMAT = "https://huggingface.co/rhasspy/piper-voices/resolve/main/{file}"
 
 _DIR = Path(__file__).parent
 _LOGGER = logging.getLogger(__name__)
@@ -47,20 +48,21 @@ def get_voices(
         except Exception:
             _LOGGER.exception("Failed to update voices list")
 
+    voices_embedded = _DIR / "voices.json"
+    _LOGGER.debug("Loading %s", voices_embedded)
+    with open(voices_embedded, "r", encoding="utf-8") as voices_file:
+        voices = json.load(voices_file)
+
     # Prefer downloaded file to embedded
     if voices_download.exists():
         try:
             _LOGGER.debug("Loading %s", voices_download)
             with open(voices_download, "r", encoding="utf-8") as voices_file:
-                return json.load(voices_file)
+                voices.update(json.load(voices_file))
         except Exception:
             _LOGGER.exception("Failed to load %s", voices_download)
 
-    # Fall back to embedded
-    voices_embedded = _DIR / "voices.json"
-    _LOGGER.debug("Loading %s", voices_embedded)
-    with open(voices_embedded, "r", encoding="utf-8") as voices_file:
-        return json.load(voices_file)
+    return voices
 
 
 def ensure_voice_exists(

wyoming_piper/handler.py

@@ -1,4 +1,5 @@
 """Event handler for clients of the server."""
+
 import argparse
 import json
 import logging
@@ -12,9 +13,16 @@ from wyoming.error import Error
 from wyoming.event import Event
 from wyoming.info import Describe, Info
 from wyoming.server import AsyncEventHandler
-from wyoming.tts import Synthesize
+from wyoming.tts import (
+    Synthesize,
+    SynthesizeChunk,
+    SynthesizeStart,
+    SynthesizeStop,
+    SynthesizeStopped,
+)
 
 from .process import PiperProcessManager
+from .sentence_boundary import SentenceBoundaryDetector, remove_asterisks
 
 _LOGGER = logging.getLogger(__name__)
 
@@ -33,6 +41,9 @@ class PiperEventHandler(AsyncEventHandler):
         self.cli_args = cli_args
         self.wyoming_info_event = wyoming_info.event()
         self.process_manager = process_manager
+        self.sbd = SentenceBoundaryDetector()
+        self.is_streaming: Optional[bool] = None
+        self._synthesize: Optional[Synthesize] = None
 
     async def handle_event(self, event: Event) -> bool:
         if Describe.is_type(event.type):
@@ -40,20 +51,67 @@ class PiperEventHandler(AsyncEventHandler):
             _LOGGER.debug("Sent info")
             return True
 
-        if not Synthesize.is_type(event.type):
-            _LOGGER.warning("Unexpected event: %s", event)
-            return True
-
         try:
-            return await self._handle_event(event)
+            if Synthesize.is_type(event.type):
+                if self.is_streaming:
+                    # Ignore since this is only sent for compatibility reasons.
+                    # For streaming, we expect:
+                    # [synthesize-start] -> [synthesize-chunk]+ -> [synthesize]? -> [synthesize-stop]
+                    return True
+
+                # Sent outside a stream, so we must process it
+                synthesize = Synthesize.from_event(event)
+                synthesize.text = remove_asterisks(synthesize.text)
+                return await self._handle_synthesize(synthesize)
+
+            if not self.cli_args.streaming:
+                # Streaming is not enabled
+                return True
+
+            if SynthesizeStart.is_type(event.type):
+                # Start of a stream
+                stream_start = SynthesizeStart.from_event(event)
+                self.is_streaming = True
+                self.sbd = SentenceBoundaryDetector()
+                self._synthesize = Synthesize(text="", voice=stream_start.voice)
+                _LOGGER.debug("Text stream started: voice=%s", stream_start.voice)
+                return True
+
+            if SynthesizeChunk.is_type(event.type):
+                assert self._synthesize is not None
+                stream_chunk = SynthesizeChunk.from_event(event)
+                for sentence in self.sbd.add_chunk(stream_chunk.text):
+                    _LOGGER.debug("Synthesizing stream sentence: %s", sentence)
+                    self._synthesize.text = sentence
+                    await self._handle_synthesize(self._synthesize)
+
+                return True
+
+            if SynthesizeStop.is_type(event.type):
+                assert self._synthesize is not None
+                self._synthesize.text = self.sbd.finish()
+                if self._synthesize.text:
+                    # Final audio chunk(s)
+                    await self._handle_synthesize(self._synthesize)
+
+                # End of audio
+                await self.write_event(SynthesizeStopped().event())
+
+                _LOGGER.debug("Text stream stopped")
+                return True
+
+            if not Synthesize.is_type(event.type):
+                return True
+
+            synthesize = Synthesize.from_event(event)
+            return await self._handle_synthesize(synthesize)
         except Exception as err:
             await self.write_event(
                 Error(text=str(err), code=err.__class__.__name__).event()
             )
             raise err
 
-    async def _handle_event(self, event: Event) -> bool:
-        synthesize = Synthesize.from_event(event)
+    async def _handle_synthesize(self, synthesize: Synthesize) -> bool:
         _LOGGER.debug(synthesize)
 
         raw_text = synthesize.text

wyoming_piper/sentence_boundary.py

@@ -0,0 +1,58 @@
+"""Guess the sentence boundaries in text."""
+
+from collections.abc import Iterable
+
+import regex as re
+
+SENTENCE_END = r"[.!?…]|[。！？]|[؟]|[।॥]"
+ABBREVIATION_RE = re.compile(r"\b\p{L}{1,3}\.$", re.UNICODE)
+
+SENTENCE_BOUNDARY_RE = re.compile(
+    rf"(.*?(?:{SENTENCE_END}+))(?=\s+[\p{{Lu}}\p{{Lt}}\p{{Lo}}]|(?:\s+\d+\.\s+))",
+    re.DOTALL,
+)
+WORD_ASTERISKS = re.compile(r"\*+([^\*]+)\*+")
+LINE_ASTERICKS = re.compile(r"(?<=^|\n)\s*\*+")
+
+
+class SentenceBoundaryDetector:
+    def __init__(self) -> None:
+        self.remaining_text = ""
+        self.current_sentence = ""
+
+    def add_chunk(self, chunk: str) -> Iterable[str]:
+        self.remaining_text += chunk
+        while self.remaining_text:
+            match = SENTENCE_BOUNDARY_RE.search(self.remaining_text)
+            if not match:
+                break
+
+            match_text = match.group(0)
+
+            if not self.current_sentence:
+                self.current_sentence = match_text
+            elif ABBREVIATION_RE.search(self.current_sentence[-5:]):
+                self.current_sentence += match_text
+            else:
+                yield remove_asterisks(self.current_sentence.strip())
+                self.current_sentence = match_text
+
+            if not ABBREVIATION_RE.search(self.current_sentence[-5:]):
+                yield remove_asterisks(self.current_sentence.strip())
+                self.current_sentence = ""
+
+            self.remaining_text = self.remaining_text[match.end() :]
+
+    def finish(self) -> str:
+        text = (self.current_sentence + self.remaining_text).strip()
+        self.remaining_text = ""
+        self.current_sentence = ""
+
+        return remove_asterisks(text)
+
+
+def remove_asterisks(text: str) -> str:
+    """Remove *asterisks* surrounding **words**"""
+    text = WORD_ASTERISKS.sub(r"\1", text)
+    text = LINE_ASTERICKS.sub("", text)
+    return text

wyoming_piper/voices.json

@@ -4013,6 +4013,36 @@
         },
         "aliases": []
     },
+    "es_AR-daniela-high": {
+        "key": "es_AR-daniela-high",
+        "name": "daniela",
+        "language": {
+            "code": "es_AR",
+            "family": "es",
+            "region": "AR",
+            "name_native": "Español",
+            "name_english": "Spanish",
+            "country_english": "Argentina"
+        },
+        "quality": "high",
+        "num_speakers": 1,
+        "speaker_id_map": {},
+        "files": {
+            "es/es_AR/daniela/high/es_AR-daniela-high.onnx": {
+                "size_bytes": 114199011,
+                "md5_digest": "e373fb657c93877dbc438badeadff4cb"
+            },
+            "es/es_AR/daniela/high/es_AR-daniela-high.onnx.json": {
+                "size_bytes": 7248,
+                "md5_digest": "b68d699a1779d98ee94c0cd9a08ba95b"
+            },
+            "es/es_AR/daniela/high/MODEL_CARD": {
+                "size_bytes": 353,
+                "md5_digest": "d940f7c6667a11ea165ff83ef7a86669"
+            }
+        },
+        "aliases": []
+    },
     "es_ES-carlfm-x_low": {
         "key": "es_ES-carlfm-x_low",
         "name": "carlfm",
@@ -4791,6 +4821,66 @@
         },
         "aliases": []
     },
+    "hi_IN-pratham-medium": {
+        "key": "hi_IN-pratham-medium",
+        "name": "pratham",
+        "language": {
+            "code": "hi_IN",
+            "family": "hi",
+            "region": "IN",
+            "name_native": "हिन्दी",
+            "name_english": "Hindi",
+            "country_english": "India"
+        },
+        "quality": "medium",
+        "num_speakers": 1,
+        "speaker_id_map": {},
+        "files": {
+            "hi/hi_IN/pratham/medium/hi_IN-pratham-medium.onnx": {
+                "size_bytes": 63516050,
+                "md5_digest": "f1e5a629a9e533a7155910530109eb86"
+            },
+            "hi/hi_IN/pratham/medium/hi_IN-pratham-medium.onnx.json": {
+                "size_bytes": 4970,
+                "md5_digest": "ab655cdad90f1939f00d1b3aaf440e99"
+            },
+            "hi/hi_IN/pratham/medium/MODEL_CARD": {
+                "size_bytes": 304,
+                "md5_digest": "eccc77c3d857a883e85c52cb17653c1f"
+            }
+        },
+        "aliases": []
+    },
+    "hi_IN-priyamvada-medium": {
+        "key": "hi_IN-priyamvada-medium",
+        "name": "priyamvada",
+        "language": {
+            "code": "hi_IN",
+            "family": "hi",
+            "region": "IN",
+            "name_native": "हिन्दी",
+            "name_english": "Hindi",
+            "country_english": "India"
+        },
+        "quality": "medium",
+        "num_speakers": 1,
+        "speaker_id_map": {},
+        "files": {
+            "hi/hi_IN/priyamvada/medium/hi_IN-priyamvada-medium.onnx": {
+                "size_bytes": 63516050,
+                "md5_digest": "7d5e20c2d1e72de8ed772f222e679626"
+            },
+            "hi/hi_IN/priyamvada/medium/hi_IN-priyamvada-medium.onnx.json": {
+                "size_bytes": 4973,
+                "md5_digest": "599ca4dc5d421a9c66692433f618e080"
+            },
+            "hi/hi_IN/priyamvada/medium/MODEL_CARD": {
+                "size_bytes": 307,
+                "md5_digest": "e2b745cf97087be0a97c7f10215bfa70"
+            }
+        },
+        "aliases": []
+    },
     "hu_HU-anna-medium": {
         "key": "hu_HU-anna-medium",
         "name": "anna",
@@ -5264,6 +5354,96 @@
         },
         "aliases": []
     },
+    "ml_IN-arjun-medium": {
+        "key": "ml_IN-arjun-medium",
+        "name": "arjun",
+        "language": {
+            "code": "ml_IN",
+            "family": "ml",
+            "region": "IN",
+            "name_native": "മലയാളം",
+            "name_english": "Malayalam",
+            "country_english": "India"
+        },
+        "quality": "medium",
+        "num_speakers": 1,
+        "speaker_id_map": {},
+        "files": {
+            "ml/ml_IN/arjun/medium/ml_IN-arjun-medium.onnx": {
+                "size_bytes": 62950044,
+                "md5_digest": "4f20109c108aa80f46df85ab9cda5daa"
+            },
+            "ml/ml_IN/arjun/medium/ml_IN-arjun-medium.onnx.json": {
+                "size_bytes": 5044,
+                "md5_digest": "18454efcad5bb889ee4cb1cb035993f5"
+            },
+            "ml/ml_IN/arjun/medium/MODEL_CARD": {
+                "size_bytes": 308,
+                "md5_digest": "eb0037396b93f38ffbc093dba42cf8be"
+            }
+        },
+        "aliases": []
+    },
+    "ml_IN-meera-medium": {
+        "key": "ml_IN-meera-medium",
+        "name": "meera",
+        "language": {
+            "code": "ml_IN",
+            "family": "ml",
+            "region": "IN",
+            "name_native": "മലയാളം",
+            "name_english": "Malayalam",
+            "country_english": "India"
+        },
+        "quality": "medium",
+        "num_speakers": 1,
+        "speaker_id_map": {},
+        "files": {
+            "ml/ml_IN/meera/medium/ml_IN-meera-medium.onnx": {
+                "size_bytes": 62950044,
+                "md5_digest": "3eb7b05d25c1551f7a7cec1e1c153b1f"
+            },
+            "ml/ml_IN/meera/medium/ml_IN-meera-medium.onnx.json": {
+                "size_bytes": 5045,
+                "md5_digest": "64c2eee28c62e3a072ccc2ea3d812748"
+            },
+            "ml/ml_IN/meera/medium/MODEL_CARD": {
+                "size_bytes": 308,
+                "md5_digest": "626a0c1d744d0ce6ffca95665c526939"
+            }
+        },
+        "aliases": []
+    },
+    "ne_NP-chitwan-medium": {
+        "key": "ne_NP-chitwan-medium",
+        "name": "chitwan",
+        "language": {
+            "code": "ne_NP",
+            "family": "ne",
+            "region": "NP",
+            "name_native": "नेपाली",
+            "name_english": "Nepali",
+            "country_english": "Nepal"
+        },
+        "quality": "medium",
+        "num_speakers": 1,
+        "speaker_id_map": {},
+        "files": {
+            "ne/ne_NP/chitwan/medium/ne_NP-chitwan-medium.onnx": {
+                "size_bytes": 62950044,
+                "md5_digest": "74cdb5b32816c366af74b55ed7494e25"
+            },
+            "ne/ne_NP/chitwan/medium/ne_NP-chitwan-medium.onnx.json": {
+                "size_bytes": 5043,
+                "md5_digest": "7ddc520e24a862016f6d22cde4b70ac1"
+            },
+            "ne/ne_NP/chitwan/medium/MODEL_CARD": {
+                "size_bytes": 275,
+                "md5_digest": "dcd35df097f84ace498d7afeb5b1a29c"
+            }
+        },
+        "aliases": []
+    },
     "ne_NP-google-medium": {
         "key": "ne_NP-google-medium",
         "name": "google",

{wyoming_piper-1.5.3.dist-info → wyoming_piper-1.6.3.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: wyoming-piper
-Version: 1.5.3
+Version: 1.6.3
 Summary: Wyoming Server for Piper
 Author-email: Michael Hansen <mike@rhasspy.org>
 License: MIT
@@ -14,10 +14,12 @@ Classifier: Programming Language :: Python :: 3.9
 Classifier: Programming Language :: Python :: 3.10
 Classifier: Programming Language :: Python :: 3.11
 Classifier: Programming Language :: Python :: 3.12
-Requires-Python: <3.13,>=3.8.1
+Classifier: Programming Language :: Python :: 3.13
+Requires-Python: >=3.8
 Description-Content-Type: text/markdown
 License-File: LICENSE.md
-Requires-Dist: wyoming>=1.5.3
+Requires-Dist: wyoming<1.8,>=1.7.2
+Requires-Dist: regex==2024.11.6
 Provides-Extra: dev
 Requires-Dist: black==22.12.0; extra == "dev"
 Requires-Dist: flake8==6.0.0; extra == "dev"

wyoming_piper-1.6.3.dist-info/RECORD

@@ -0,0 +1,15 @@
+wyoming_piper/__init__.py,sha256=z1dsCtGazHHufHQpoVgNtMObt25qYBSOM85o7xgbIJA,139
+wyoming_piper/__main__.py,sha256=SuJ6XY6zy68N8L-N_n_EIWK0vpZwbYQXW5vlSC8BpW8,7445
+wyoming_piper/const.py,sha256=04sCdtJ2QGuF1BQGkOuQW10og61PgH3fCnPhaYu-YoU,1015
+wyoming_piper/download.py,sha256=UpczxHWqLkcOblHmrwgBHSR6wG1LR-hZ4V6QSsrghns,6185
+wyoming_piper/file_hash.py,sha256=HMuwrgEIg-bCOXHG0wE3vtjrqGD7QaA_UNfvBMXeUcY,1107
+wyoming_piper/handler.py,sha256=WVpmnRVYmsd3DrLcMfBOsuo_J1HJ0h00-HwQ1iZEToo,7360
+wyoming_piper/process.py,sha256=L_qqxQcQawrC940fwlv4u6KM9KjCq6N6ym-OADSZcrM,5794
+wyoming_piper/sentence_boundary.py,sha256=pHVo92_weusnVLRVicnS0-Tst_eR-pMrnRrGL96HxC8,1875
+wyoming_piper/voices.json,sha256=elUT3cM0Wlgo8N8E5nhMbMSCPB8zU4SY2XGKwe-T2ys,209108
+wyoming_piper-1.6.3.dist-info/licenses/LICENSE.md,sha256=E3RtUJ105V6iJl--8gS7fNv4SoMVsCB-mIMmy1Q4cCg,1071
+wyoming_piper-1.6.3.dist-info/METADATA,sha256=sNP4bue0pO2mFBb3xUnXfgibofOvaCFuJqN7Hik3fmQ,2543
+wyoming_piper-1.6.3.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+wyoming_piper-1.6.3.dist-info/entry_points.txt,sha256=n2UgsOCQitQ5Itr20aITTWZLL2dAtaVKn5pdecXdDHE,61
+wyoming_piper-1.6.3.dist-info/top_level.txt,sha256=t7U7-u1sK_4xy_qbTJhxQRbxle3cLQfPq2oVLezHVNU,14
+wyoming_piper-1.6.3.dist-info/RECORD,,

{wyoming_piper-1.5.3.dist-info → wyoming_piper-1.6.3.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: setuptools (80.7.0)
+Generator: setuptools (80.9.0)
 Root-Is-Purelib: true
 Tag: py3-none-any
 

wyoming_piper-1.5.3.dist-info/RECORD

@@ -1,14 +0,0 @@
-wyoming_piper/__init__.py,sha256=jq4RWRoEFQNm73oe9FNeVxdHTbu-MdElAnW1fbodESo,138
-wyoming_piper/__main__.py,sha256=knqXEoX_oxJr69oKitAILZKioKsn14HSrTpa3lCT-nA,7186
-wyoming_piper/const.py,sha256=04sCdtJ2QGuF1BQGkOuQW10og61PgH3fCnPhaYu-YoU,1015
-wyoming_piper/download.py,sha256=n3oKXN0fLk7Dn7x4wnR17b3QzXilFgF-QSxtN0QE10s,6185
-wyoming_piper/file_hash.py,sha256=HMuwrgEIg-bCOXHG0wE3vtjrqGD7QaA_UNfvBMXeUcY,1107
-wyoming_piper/handler.py,sha256=0bcllkoR6iNsseWQ1Y-Uj4sMV_az6mn4ivS8AAo4RHw,4936
-wyoming_piper/process.py,sha256=L_qqxQcQawrC940fwlv4u6KM9KjCq6N6ym-OADSZcrM,5794
-wyoming_piper/voices.json,sha256=3uEoyiBYbQlJnFYUEKnFHYK0Wq40sbaHiD9AsgRWKzc,202976
-wyoming_piper-1.5.3.dist-info/licenses/LICENSE.md,sha256=E3RtUJ105V6iJl--8gS7fNv4SoMVsCB-mIMmy1Q4cCg,1071
-wyoming_piper-1.5.3.dist-info/METADATA,sha256=9fYLRK0GOJCnaf9FBwlV1zoYUoDtyb0h66tOilKf2YU,2463
-wyoming_piper-1.5.3.dist-info/WHEEL,sha256=A8Eltl-h0W-qZDVezsLjjslosEH_pdYC2lQ0JcbgCzs,91
-wyoming_piper-1.5.3.dist-info/entry_points.txt,sha256=n2UgsOCQitQ5Itr20aITTWZLL2dAtaVKn5pdecXdDHE,61
-wyoming_piper-1.5.3.dist-info/top_level.txt,sha256=t7U7-u1sK_4xy_qbTJhxQRbxle3cLQfPq2oVLezHVNU,14
-wyoming_piper-1.5.3.dist-info/RECORD,,