Update edge-tts to version 4.0.11

* Add pylint check for lint.sh
* Change overlapping default from 5second to 1second for SubMaker and util.py
* Default to WordBoundary for edge-playback (from SentenceBoundary)
* Drop SentenceBoundary support (never works properly and too many hacks)
  - No longer actually supported by Azure's official SDK for a few years
    already
* Stop attempting to correct broken offsets sent back to us by Azure
  - Fixes never work properly because AI voice sometimes takes odd pauses
    at the start and sometimes doesn't do so. It's never predictable and
    cannot be fixed on the library's end.
  - Solution is for Microsoft to fix the integer overflow bug they are facing
    in the {Word,Sentence}Boundary offsets. It doesn't affect us until we reach
    30min long TTS anyway.
* Have edge-tts --list-voices use the configured HTTP proxy
* More misc changes and fixes

lint.sh

@@ -1,2 +1,3 @@
 find src examples -name '*.py' | xargs black
 find src examples -name '*.py' | xargs isort
+find src examples -name '*.py' | xargs pylint

setup.cfg

@@ -1,6 +1,6 @@
 [metadata]
 name = edge-tts
-version = 4.0.10
+version = 4.0.11
 author = rany
 author_email = ranygh@riseup.net
 description = Microsoft Edge's TTS

src/edge_playback/__init__.py

@@ -28,7 +28,7 @@ def main():
             with subprocess.Popen(
                 [
                     "edge-tts",
-                    "--boundary-type=2",
+                    "--boundary-type=1",
                     f"--write-media={media.name}",
                     f"--write-subtitles={subtitle.name}",
                 ]

src/edge_tts/communicate.py

@@ -55,8 +55,8 @@ def remove_incompatible_characters(string):
 
     string = list(string)
 
-    for idx in range(len(string)):  # pylint: disable=consider-using-enumerate
+    for idx, char in enumerate(string):
-        code = ord(string[idx])
+        code = ord(char)
         if (0 <= code <= 8) or (11 <= code <= 12) or (14 <= code <= 31):
             string[idx] = " "
 
@@ -193,7 +193,7 @@ def ssml_headers_plus_data(request_id, timestamp, ssml):
     )
 
 
-class Communicate:  # pylint: disable=too-few-public-methods
+class Communicate:
     """
     Class for communicating with the service.
     """
@@ -215,7 +215,7 @@ class Communicate:  # pylint: disable=too-few-public-methods
         volume="+0%",
         customspeak=False,
         proxy=None,
-    ):  # pylint: disable=too-many-arguments, too-many-locals
+    ):
         """
         Runs the Communicate class.
 
@@ -234,14 +234,14 @@ class Communicate:  # pylint: disable=too-few-public-methods
         """
 
         word_boundary = False
-        sentence_boundary = False
 
         if boundary_type > 0:
             word_boundary = True
         if boundary_type > 1:
-            sentence_boundary = True
+            raise ValueError(
+                "Invalid boundary type. SentenceBoundary is no longer supported."
+            )
 
-        sentence_boundary = str(sentence_boundary).lower()
         word_boundary = str(word_boundary).lower()
 
         if not customspeak:
@@ -262,12 +262,8 @@ class Communicate:  # pylint: disable=too-few-public-methods
             if isinstance(messages, str):
                 messages = [messages]
 
-
         # Variables for the loop
         download = False
-        current_subtitle = ""
-        first_offset = None
-        last_offset = None
         async with aiohttp.ClientSession(trust_env=True) as session:
             async with session.ws_connect(
                 f"{WSS_URL}&ConnectionId={connect_id()}",
@@ -304,7 +300,7 @@ class Communicate:  # pylint: disable=too-few-public-methods
                         "Content-Type:application/json; charset=utf-8\r\n"
                         "Path:speech.config\r\n\r\n"
                         '{"context":{"synthesis":{"audio":{"metadataoptions":{'
-                        f'"sentenceBoundaryEnabled":{sentence_boundary},'
+                        f'"sentenceBoundaryEnabled":false,'
                         f'"wordBoundaryEnabled":{word_boundary}}},"outputFormat":"{codec}"'
                         "}}}}\r\n"
                     )
@@ -326,12 +322,6 @@ class Communicate:  # pylint: disable=too-few-public-methods
 
                     # Begin listening for the response.
                     async for received in websocket:
-                        if received.type in (
-                            aiohttp.WSMsgType.CLOSED,
-                            aiohttp.WSMsgType.ERROR,
-                        ):
-                            break
-
                         if received.type == aiohttp.WSMsgType.TEXT:
                             parameters, data = get_headers_and_data(received.data)
                             if (
@@ -358,12 +348,15 @@ class Communicate:  # pylint: disable=too-few-public-methods
                                     metadata_duration = metadata["Metadata"][0]["Data"][
                                         "Duration"
                                     ]
-                                except KeyError:
+                                except KeyError as exception:
-                                    metadata_duration = 0
+                                    raise ValueError(
+                                        "The metadata doesn't contain a Duration field. "
+                                        + "This usually happens when SentenceBoundary metadata type is sent."
+                                    ) from exception
                                 metadata_text = metadata["Metadata"][0]["Data"]["text"][
                                     "Text"
                                 ]
-                                if boundary_type == 1:
+                                if metadata_type == "WordBoundary":
                                     yield (
                                         [
                                             metadata_offset,
@@ -372,31 +365,32 @@ class Communicate:  # pylint: disable=too-few-public-methods
                                         metadata_text,
                                         None,
                                     )
+                                elif metadata_type == "SentenceBoundary":
+                                    raise NotImplementedError(
+                                        "SentenceBoundary is not supported due to being broken."
+                                    )
                                 else:
-                                    if metadata_type == "WordBoundary":
+                                    raise NotImplementedError(
-                                        if current_subtitle:
+                                        f"Unknown metadata type: {metadata_type}"
-                                            current_subtitle += " "
+                                    )
-                                        current_subtitle += metadata_text
+                            elif (
-                                        if first_offset is None:
+                                "Path" in parameters
-                                            first_offset = metadata_offset
+                                and parameters["Path"] == "response"
-                                        last_offset = [
+                            ):
-                                            metadata_offset,
+                                # TODO: implement this:
-                                            metadata_duration,
+                                """
-                                        ]
+                                X-RequestId:xxxxxxxxxxxxxxxxxxxxxxxxx
-                                    elif metadata_type == "SentenceBoundary":
+                                Content-Type:application/json; charset=utf-8
-                                        if current_subtitle:
+                                Path:response
-                                            yield (
-                                                [
-                                                    first_offset,
-                                                    sum(last_offset) - first_offset,
-                                                ],
-                                                current_subtitle,
-                                                None,
-                                            )
-                                        current_subtitle = ""
-                                        first_offset = None
-                                        last_offset = None
 
+                                {"context":{"serviceTag":"yyyyyyyyyyyyyyyyyyy"},"audio":{"type":"inline","streamId":"zzzzzzzzzzzzzzzzz"}}
+                                """
+                                pass
+                            else:
+                                raise ValueError(
+                                    "The response from the service is not recognized.\n"
+                                    + received.data
+                                )
                         elif received.type == aiohttp.WSMsgType.BINARY:
                             if download:
                                 yield (
@@ -406,10 +400,8 @@ class Communicate:  # pylint: disable=too-few-public-methods
                                         received.data.split(b"Path:audio\r\n")[1:]
                                     ),
                                 )
-                if current_subtitle:
+                            else:
-                    yield (
+                                raise ValueError(
-                        [first_offset, sum(last_offset) - first_offset],
+                                    "The service sent a binary message, but we are not expecting one."
-                        current_subtitle,
+                                )
-                        None,
-                    )
                 await websocket.close()

src/edge_tts/list_voices.py

@@ -9,7 +9,7 @@ import aiohttp
 from .constants import VOICE_LIST
 
 
-async def list_voices():
+async def list_voices(proxy=None):
     """
     List all available voices and their attributes.
 
@@ -36,6 +36,7 @@ async def list_voices():
                 "Accept-Encoding": "gzip, deflate, br",
                 "Accept-Language": "en-US,en;q=0.9",
             },
+            proxy=proxy,
         ) as url:
             data = json.loads(await url.text())
     return data

src/edge_tts/submaker.py

@@ -39,7 +39,7 @@ class SubMaker:
     SubMaker class
     """
 
-    def __init__(self, overlapping=5):
+    def __init__(self, overlapping=1):
         """
         SubMaker constructor.
 
@@ -48,7 +48,6 @@ class SubMaker:
                                subtitles should overlap.
         """
         self.subs_and_offset = []
-        self.broken_offset = 0
         self.overlapping = overlapping * (10**7)
 
     def create_sub(self, timestamp, text):
@@ -64,13 +63,6 @@ class SubMaker:
             None
         """
         timestamp[1] += timestamp[0]
-
-        if len(self.subs_and_offset) >= 2:
-            if self.subs_and_offset[-2][1] >= timestamp[0] + self.broken_offset:
-                self.broken_offset = self.subs_and_offset[-2][1]
-            timestamp[0] += self.broken_offset
-            timestamp[1] += self.broken_offset
-
         self.subs_and_offset.append(timestamp)
         self.subs_and_offset.append(text)
 

src/edge_tts/util.py

@@ -10,11 +10,11 @@ import sys
 from edge_tts import Communicate, SubMaker, list_voices
 
 
-async def _list_voices():
+async def _list_voices(proxy):
     """
     List available voices.
     """
-    for idx, voice in enumerate(await list_voices()):
+    for idx, voice in enumerate(await list_voices(proxy=proxy)):
         if idx != 0:
             print()
 
@@ -112,13 +112,13 @@ async def _main():
         "-O",
         "--overlapping",
         help="overlapping subtitles in seconds",
-        default=5,
+        default=1,
         type=float,
     )
     parser.add_argument(
         "-b",
         "--boundary-type",
-        help="set boundary type for subtitles. Default 0 for none. Set 1 for word_boundary, 2 for sentence_boundary",
+        help="set boundary type for subtitles. Default 0 for none. Set 1 for word_boundary.",
         default=0,
         type=int,
     )
@@ -136,7 +136,7 @@ async def _main():
     args = parser.parse_args()
 
     if args.list_voices:
-        await _list_voices()
+        await _list_voices(args.proxy)
         sys.exit(0)
 
     if args.text is not None or args.file is not None: