Add <prosody rate="...">

README.md

@@ -242,7 +242,7 @@ For example:
   <break time="3s" />
   <voice name="en_US/cmu-arctic_low#slt">
     <s>
-      <prosody volume="soft">
+      <prosody volume="soft" rate="150%">
         This is a <say-as interpret-as="number" format="ordinal">2</say-as> voice.
       </prosody>
     </s>

mimic3-http/mimic3_http/app.py

@@ -17,6 +17,7 @@
 import argparse
 import asyncio
 import dataclasses
+import json
 import logging
 import typing
 from pathlib import Path
@@ -119,6 +120,17 @@ def get_app(args: argparse.Namespace, request_queue: Queue, temp_dir: str):
     def _to_bool(s: str) -> bool:
         return s.strip().lower() in {"true", "1", "yes", "on"}
 
+    class VoiceEncoder(json.JSONEncoder):
+        """Encode a voice to JSON"""
+
+        def default(self, o):
+            if isinstance(o, set):
+                return list(o)
+
+            return json.JSONEncoder.default(self, o)
+
+    app.json_encoder = VoiceEncoder  # type: ignore
+
     @app.route("/img/<path:filename>", methods=["GET"])
     async def img(filename) -> Response:
         """Image static endpoint."""

mimic3-tts/README.md

@@ -178,6 +178,10 @@ A subset of [SSML](https://www.w3.org/TR/speech-synthesis11/) (Speech Synthesis
             * number in [0, 100] - 0 is silent, 100 is loudest (default)
             * +X, -X, +X%, -X% - absolute/percent offset from current volume
             * one of "default", "silent", "x-loud", "loud", "medium", "soft", "x-soft"
+        * `rate` - speaking rate
+            * number - 1 is default rate, < 1 is slower, > 1 is faster
+            * X% - 100% is default rate, 50% is half speed, 200% is twice as fast
+            * one of "default", "x-fast", "fast", "medium", "slow", "x-slow"
 * `<say-as interpret-as="">` - force interpretation of inner text
     * `interpret-as` one of "spell-out", "date", "number", "time", or "currency"
     * `format` - way to format text depending on `interpret-as`

mimic3-tts/mimic3_tts/const.py

@@ -25,3 +25,4 @@ DEFAULT_VOICES_URL_FORMAT = (
 DEFAULT_VOICES_DOWNLOAD_DIR = Path(XDG().XDG_DATA_HOME) / "mimic3" / "voices"
 
 DEFAULT_VOLUME = 100.0
+DEFAULT_RATE = 1.0

mimic3-tts/mimic3_tts/tts.py

@@ -40,6 +40,7 @@ from ._resources import _VOICES
 from .config import TrainingConfig
 from .const import (
     DEFAULT_LANGUAGE,
+    DEFAULT_RATE,
     DEFAULT_VOICE,
     DEFAULT_VOICES_DOWNLOAD_DIR,
     DEFAULT_VOICES_URL_FORMAT,
@@ -113,6 +114,9 @@ class Mimic3Settings:
     volume: float = DEFAULT_VOLUME
     """Voice volume in [0, 100]"""
 
+    rate: float = DEFAULT_RATE
+    """Voice speaking rate (< 1 is slower, > 1 is faster)"""
+
 
 @dataclass
 class Mimic3Phonemes:
@@ -318,6 +322,14 @@ class Mimic3TextToSpeechSystem(TextToSpeechSystem):
     def volume(self, new_volume: float):
         self.settings.volume = max(0, min(100, new_volume))
 
+    @property
+    def rate(self) -> float:
+        return self.settings.rate
+
+    @rate.setter
+    def rate(self, new_rate: float):
+        self.settings.rate = new_rate
+
     def begin_utterance(self):
         pass
 
@@ -456,6 +468,7 @@ class Mimic3TextToSpeechSystem(TextToSpeechSystem):
             length_scale=settings.length_scale,
             noise_scale=settings.noise_scale,
             noise_w=settings.noise_w,
+            rate=settings.rate,
         )
 
         audio_bytes = audio.tobytes()

mimic3-tts/mimic3_tts/voice.py

@@ -32,8 +32,9 @@ import onnxruntime
 import phonemes2ids
 from gruut_ipa import IPA
 
-from mimic3_tts.config import Phonemizer, TrainingConfig
+from .config import Phonemizer, TrainingConfig
-from mimic3_tts.utils import audio_float_to_int16
+from .const import DEFAULT_RATE
+from .utils import audio_float_to_int16
 
 # -----------------------------------------------------------------------------
 
@@ -159,11 +160,16 @@ class Mimic3Voice(metaclass=ABCMeta):
         length_scale: typing.Optional[float] = None,
         noise_scale: typing.Optional[float] = None,
         noise_w: typing.Optional[float] = None,
+        rate: float = DEFAULT_RATE,
     ) -> np.ndarray:
         """Synthesize audio from phoneme ids usng Onnx voice model (see generator.onnx)"""
         if length_scale is None:
             length_scale = self.config.inference.length_scale
 
+        # Scale length by rate
+        if rate > 0:
+            length_scale /= rate
+
         if noise_scale is None:
             noise_scale = self.config.inference.noise_scale
 

opentts-abc/opentts_abc/__init__.py

@@ -218,6 +218,15 @@ class TextToSpeechSystem(AbstractContextManager, metaclass=ABCMeta):
     def volume(self, new_volume: float):
         """Set the current volume in [0, 100]"""
 
+    @property
+    @abstractmethod
+    def rate(self) -> float:
+        """Get the current speaking rate"""
+
+    @rate.setter
+    def rate(self, new_rate: float):
+        """Set the current speaker rate"""
+
     def shutdown(self):
         """Called by the host program when the text to speech system should be stopped"""
 

opentts-abc/opentts_abc/ssml.py

@@ -63,6 +63,7 @@ class ParsingState(int, enum.Enum):
 
 
 _DEFAULT_VOLUME: float = 100.0
+_DEFAULT_RATE: float = 1.0
 
 
 @dataclass
@@ -70,6 +71,10 @@ class ProsodyState:
     """Current prosody settings"""
 
     volume: float = _DEFAULT_VOLUME
+    """Currrent volume setting in [0, 100]"""
+
+    rate: float = _DEFAULT_RATE
+    """Current rate setting (< 1 is slower, > 1 is faster)"""
 
 
 # -----------------------------------------------------------------------------
@@ -84,12 +89,29 @@ _DEFAULT_VOLUME_MAP = {
     "silent": 0.0,
 }
 
+_DEFAULT_RATE_MAP = {
+    "default": _DEFAULT_RATE,
+    "x-fast": _DEFAULT_RATE * 3,
+    "fast": _DEFAULT_RATE * 2,
+    "medium": _DEFAULT_RATE,
+    "slow": _DEFAULT_RATE * 0.5,
+    "x-slow": _DEFAULT_RATE * 0.25,
+}
+
 
 @dataclass
 class SSMLSettings:
+    """Settings for SSML named constants"""
+
     volume_map: typing.Mapping[str, float] = field(
         default_factory=lambda: _DEFAULT_VOLUME_MAP
     )
+    """Volume named constant to volume level"""
+
+    rate_map: typing.Mapping[str, float] = field(
+        default_factory=lambda: _DEFAULT_RATE_MAP
+    )
+    """Rate named constant to speaking rate"""
 
 
 # -----------------------------------------------------------------------------
@@ -449,10 +471,15 @@ class SSMLSpeaker:
                 volume_str, current_volume=self._prosody.volume
             )
 
+        rate_str = attrib_no_namespace(elem, "rate")
+        if rate_str is not None:
+            new_prosody.rate = self._parse_rate(rate_str)
+
         LOG.debug("prosody: %s", new_prosody)
         self._push_prosody(new_prosody)
 
         self.tts.volume = new_prosody.volume
+        self.tts.rate = new_prosody.rate
 
     def _handle_end_prosody(self):
         """Handle </prosody>"""
@@ -607,6 +634,32 @@ class SSMLSpeaker:
 
         return max(0, min(_DEFAULT_VOLUME, volume))
 
+    def _parse_rate(self, rate_str: str) -> float:
+        """Parse SSML rate from <prosody> into float"""
+        rate = _DEFAULT_RATE
+        rate_str = rate_str.strip().lower()
+
+        maybe_rate = self.settings.rate_map.get(rate_str)
+        if maybe_rate is not None:
+            rate = maybe_rate
+        elif rate_str:
+            is_percent = False
+
+            if rate_str[-1] == "%":
+                is_percent = True
+                rate_str = rate_str[:-1]
+
+            rate_value = float(rate_str)
+
+            if is_percent:
+                # 50% = 0.5
+                rate = rate_value / 100.0
+            else:
+                # Absolute value
+                rate = rate_value
+
+        return rate
+
 
 # -----------------------------------------------------------------------------