get rid of nltk dependency

edge-tts.py

@@ -35,8 +35,65 @@ def removeIncompatibleControlChars(s):
         output += [ ch ]
     return "".join(output)
 
-# From https://github.com/pndurette/gTTS/blob/master/gtts/utils.py
+def list_voices():
+    with urllib.request.urlopen(voiceList) as url:
+        debug("Loading json from %s" % voiceList)
+        data = json.loads(url.read().decode())
+        debug("JSON Loaded")
+        for voice in data:
+            print()
+            for key in voice.keys():
+                debug("Processing key %s" % key)
+                if key in ["Name", "SuggestedCodec", "FriendlyName", "Status"]:
+                    debug("Key %s skipped" % key)
+                    continue
+                print("%s: %s" % ("Name" if key == "ShortName" else key, voice[key]))
+    print()
+
+def mkssmlmsg(text=""):
+    message='X-RequestId:'+connectId()+'\r\nContent-Type:application/ssml+xml\r\n'
+    message+='X-Timestamp:'+formatdate()+'Z\r\nPath:ssml\r\n\r\n'
+    message+="<speak version='1.0' xmlns='http://www.w3.org/2001/10/synthesis' xml:lang='en-US'>"
+    message+="<voice  name='" + voice + "'>" + "<prosody pitch='" + pitchString + "' rate ='" + rateString + "' volume='" + volumeString + "'>" + text + '</prosody></voice></speak>'
+    return message
+
+async def run_tts(msg):
+    debug("Doing %s!" % msg)
+    async with websockets.connect(wssUrl, ssl=ssl_context) as ws:
+        message='X-Timestamp:'+formatdate()+'\r\nContent-Type:application/json; charset=utf-8\r\nPath:speech.config\r\n\r\n'
+        message+='{"context":{"synthesis":{"audio":{"metadataoptions":{"sentenceBoundaryEnabled":"'+sentenceBoundaryEnabled+'","wordBoundaryEnabled":"'+wordBoundaryEnabled+'"},"outputFormat":"' + codec + '"}}}}\r\n'
+        await ws.send(message)
+        debug("> %s" % message)
+        await ws.send(msg)
+        debug("> %s" % msg)
+        while True:
+            recv = await ws.recv()
+            recv = recv.encode() if type(recv) is not bytes else recv
+            debug("< %s" % recv)
+            if b'turn.end' in recv:
+                break
+            elif b'Path:audio\r\n' in recv:
+                sys.stdout.buffer.write(recv.split(b'Path:audio\r\n')[1])
+
+# From https://github.com/pndurette/gTTS/blob/6d9309f05b3ad26ca356654732f3b5b9c3bec538/gtts/utils.py#L13-L54
 def _minimize(the_string, delim, max_size):
+    """Recursively split a string in the largest chunks
+    possible from the highest position of a delimiter all the way
+    to a maximum size
+    Args:
+        the_string (string): The string to split.
+        delim (string): The delimiter to split on.
+        max_size (int): The maximum size of a chunk.
+    Returns:
+        list: the minimized string in tokens
+    Every chunk size will be at minimum ``the_string[0:idx]`` where ``idx``
+    is the highest index of ``delim`` found in ``the_string``; and at maximum
+    ``the_string[0:max_size]`` if no ``delim`` was found in ``the_string``.
+    In the latter case, the split will occur at ``the_string[max_size]``
+    which can be any character. The function runs itself again on the rest of
+    ``the_string`` (``the_string[idx:]``) until no chunk is larger than
+    ``max_size``.
+    """
     # Remove `delim` from start of `the_string`
     # i.e. prevent a recursive infinite loop on `the_string[0:0]`
     # if `the_string` starts with `delim` and is larger than `max_size`
@@ -58,42 +115,6 @@ def _minimize(the_string, delim, max_size):
     else:
         return [the_string]
 
-def list_voices():
-	with urllib.request.urlopen(voiceList) as url:
-		debug("Loading json from %s" % voiceList)
-		data = json.loads(url.read().decode())
-		debug("JSON Loaded")
-		for voice in data:
-			print()
-			for key in voice.keys():
-				debug("Processing key %s" % key)
-				if key in ["Name", "SuggestedCodec", "FriendlyName", "Status"]:
-					debug("Key %s skipped" % key)
-					continue
-				print("%s: %s" % ("Name" if key == "ShortName" else key, voice[key]))
-	print()
-
-async def run_tts():
-	async with websockets.connect(wssUrl, ssl=ssl_context) as ws:
-		message='X-Timestamp:'+formatdate()+'\r\nContent-Type:application/json; charset=utf-8\r\nPath:speech.config\r\n\r\n'
-		message+='{"context":{"synthesis":{"audio":{"metadataoptions":{"sentenceBoundaryEnabled":"'+sentenceBoundaryEnabled+'","wordBoundaryEnabled":"'+wordBoundaryEnabled+'"},"outputFormat":"' + codec + '"}}}}\r\n'
-		await ws.send(message)
-		debug("> %s" % message)
-		message='X-RequestId:'+connectId()+'\r\nContent-Type:application/ssml+xml\r\n'
-		message+='X-Timestamp:'+formatdate()+'Z\r\nPath:ssml\r\n\r\n'
-		message+="<speak version='1.0' xmlns='http://www.w3.org/2001/10/synthesis' xml:lang='en-US'>"
-		message+="<voice  name='" + voice + "'>" + "<prosody pitch='" + pitchString + "' rate ='" + rateString + "' volume='" + volumeString + "'>" + escape(text) + '</prosody></voice></speak>'
-		await ws.send(message)
-		debug("> %s" % message)
-		while True:
-			recv = await ws.recv()
-			recv = recv.encode() if type(recv) is not bytes else recv
-			debug("< %s" % recv)
-			if b'turn.end' in recv:
-				break
-			elif b'Path:audio\r\n' in recv:
-				sys.stdout.buffer.write(recv.split(b'Path:audio\r\n')[1])
-
 if __name__ == "__main__":
     parser = argparse.ArgumentParser(description="Microsoft Edge's Online TTS Reader")
     group = parser.add_mutually_exclusive_group(required=True)
@@ -107,7 +128,6 @@ if __name__ == "__main__":
     parser.add_argument('-V', '--volume', help="set TTS volume. Default +0%%. For more info check https://bit.ly/3eAE5Nx", default="+0%")
     parser.add_argument('-s', '--enable-sentence-boundary', help="enable sentence boundary (not implemented but set)", action='store_true')
     parser.add_argument('-w', '--enable-word-boundary', help="enable word boundary (not implemented but set)", action='store_true')
-	parser.add_argument('-S', '--dont-split-sentences', help="sends entire text as is (careful because limit is unknown)", action='store_true')
     parser.add_argument('-D', '--debug', help="some debugging", action='store_true')
     args = parser.parse_args()
     DEBUG = args.debug
@@ -130,23 +150,11 @@ if __name__ == "__main__":
         volumeString = args.volume
         sentenceBoundaryEnabled = 'true' if args.enable_sentence_boundary else 'false'
         wordBoundaryEnabled = 'true' if args.enable_word_boundary else 'false'
-		# Websocket max is 65536, lets say that overhead is approx. 5k
+        # https://hpbn.co/websocket/ says client must also send a masking key,
-		max_size = 65536 - 5000
+        # which adds an extra 4 bytes to the header, resulting in 6–14 bytes over overhead
-		if not args.dont_split_sentences:
+        overhead = len(mkssmlmsg()) + 14
-			try:
+        wsmax = 65536 - overhead
-				from nltk.tokenize import sent_tokenize
+        for text in _minimize(escape(removeIncompatibleControlChars(args.text)), " ", wsmax):
-				debug("Was able to load nltk module")
+            asyncio.get_event_loop().run_until_complete(run_tts(mkssmlmsg(text)))
-			except Exception as e:
-				print("You need nltk for sentence splitting.", file=sys.stderr)
-				print("If you can't install it you could use the --dont-split-sentences flag.", file=sys.stderr)
-				debug("Exception was %s %s" % (e.message, e.args))
-				sys.exit(1)
-			for text in _minimize(" ".join(sent_tokenize(removeIncompatibleControlChars(args.text))), " ", max_size):
-				debug ("Sent %s to be TTSed!" % text)
-				asyncio.get_event_loop().run_until_complete(run_tts())
-		else:
-			for text in _minimize(removeIncompatibleControlChars(args.text), " ", max_size):
-				debug ("Sent %s to be TTSed!" % text)
-				asyncio.get_event_loop().run_until_complete(run_tts())
     elif args.list_voices:
         list_voices()