mirror of
https://github.com/denizsafak/abogen.git
synced 2026-07-18 13:40:27 +02:00
feat: add SuperTonic TTS plugin
- Add plugins/supertonic/ with Engine and EngineSession implementations - Reuse existing SupertonicPipeline from abogen.tts_backends.supertonic - Implement VoiceLister capability (M1-M5, F1-F5 voices) - Declare no streaming support via capabilities - Add 28 tests: plugin loading, protocol compliance, lifecycle, voice listing, parameters, errors - All 237 tests pass
This commit is contained in:
Artem Akymenko
@@ -0,0 +1,123 @@
|
||||
"""SuperTonic TTS Plugin for the TTS Plugin Architecture.
|
||||
|
||||
This plugin provides a SuperTonic-based TTS engine that implements the
|
||||
Plugin API contract. It wraps the existing SuperTonic backend in the
|
||||
new Engine/EngineSession architecture.
|
||||
|
||||
Exports:
|
||||
- PLUGIN_MANIFEST: PluginManifest
|
||||
- MODEL_REQUIREMENTS: list[ModelManifest]
|
||||
- create_engine: Factory function
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
from pathlib import Path
|
||||
from typing import Any
|
||||
|
||||
from abogen.tts_plugin.engine import Engine
|
||||
from abogen.tts_plugin.host_context import HostContext
|
||||
from abogen.tts_plugin.manifest import (
|
||||
AudioFormatManifest,
|
||||
EngineManifest,
|
||||
ModelManifest,
|
||||
ParameterManifest,
|
||||
PluginManifest,
|
||||
RequirementManifest,
|
||||
VoiceSourceManifest,
|
||||
)
|
||||
from abogen.tts_plugin.types import EngineConfig
|
||||
|
||||
from .engine import SuperTonicEngine
|
||||
|
||||
|
||||
def _load_supertonic_pipeline(sample_rate: int = 24000, auto_download: bool = True, total_steps: int = 5) -> Any:
|
||||
"""Lazy-load SuperTonic dependencies and create pipeline."""
|
||||
from abogen.tts_backends.supertonic import SupertonicPipeline
|
||||
|
||||
return SupertonicPipeline(
|
||||
sample_rate=sample_rate,
|
||||
auto_download=auto_download,
|
||||
total_steps=total_steps,
|
||||
)
|
||||
|
||||
|
||||
PLUGIN_MANIFEST = PluginManifest(
|
||||
id="supertonic",
|
||||
name="SuperTonic",
|
||||
version="0.1.0",
|
||||
api_version="1.0",
|
||||
description="SuperTonic TTS engine - fast high-quality text-to-speech",
|
||||
author="SuperTonic Team",
|
||||
capabilities=("voice_list",),
|
||||
requires=RequirementManifest(
|
||||
internet=False,
|
||||
),
|
||||
engine=EngineManifest(
|
||||
voiceSources=(
|
||||
VoiceSourceManifest(
|
||||
id="builtin",
|
||||
name="Built-in Voices",
|
||||
type="list",
|
||||
config={"voices": "See listVoices()"},
|
||||
),
|
||||
),
|
||||
parameters=(
|
||||
ParameterManifest(
|
||||
id="speed",
|
||||
name="Speed",
|
||||
description="Speech speed multiplier",
|
||||
type="float",
|
||||
default=1.0,
|
||||
min=0.7,
|
||||
max=2.0,
|
||||
step=0.1,
|
||||
),
|
||||
ParameterManifest(
|
||||
id="total_steps",
|
||||
name="Quality Steps",
|
||||
description="Inference steps (higher = better quality, slower)",
|
||||
type="int",
|
||||
default=5,
|
||||
min=2,
|
||||
max=15,
|
||||
step=1,
|
||||
),
|
||||
),
|
||||
audioFormats=(
|
||||
AudioFormatManifest(mime="audio/wav", extension="wav"),
|
||||
),
|
||||
),
|
||||
)
|
||||
|
||||
MODEL_REQUIREMENTS: list[ModelManifest] = []
|
||||
|
||||
|
||||
def create_engine(
|
||||
context: HostContext,
|
||||
model_path: Path | None,
|
||||
config: EngineConfig,
|
||||
) -> Engine:
|
||||
"""Create a SuperTonic engine instance.
|
||||
|
||||
This function is the plugin entry point. It must be atomic:
|
||||
succeed fully or raise EngineError and clean up.
|
||||
|
||||
Args:
|
||||
context: Host services (config dir, logger, http client).
|
||||
model_path: Resolved model path, or None for default.
|
||||
config: Engine initialization settings (device, etc.).
|
||||
|
||||
Returns:
|
||||
A fully initialized SuperTonicEngine instance.
|
||||
|
||||
Raises:
|
||||
EngineError: On failure. Cleans up partially created resources.
|
||||
"""
|
||||
try:
|
||||
pipeline = _load_supertonic_pipeline()
|
||||
engine = SuperTonicEngine(pipeline)
|
||||
return engine
|
||||
except Exception as e:
|
||||
from abogen.tts_plugin.errors import EngineError as EngineErrorClass
|
||||
raise EngineErrorClass(f"Failed to create SuperTonic engine: {e}") from e
|
||||
@@ -0,0 +1,156 @@
|
||||
"""SuperTonic Engine adapter for the TTS Plugin Architecture.
|
||||
|
||||
This module adapts the existing SuperTonic backend to the new Engine/EngineSession
|
||||
protocol. It wraps the SupertonicPipeline without modifying it.
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import io
|
||||
import logging
|
||||
from typing import Any, Optional
|
||||
|
||||
import numpy as np
|
||||
|
||||
from abogen.tts_plugin.capabilities import VoiceLister
|
||||
from abogen.tts_plugin.engine import Engine, EngineSession
|
||||
from abogen.tts_plugin.errors import EngineError, InvalidInputError
|
||||
from abogen.tts_plugin.manifest import VoiceManifest
|
||||
from abogen.tts_plugin.types import (
|
||||
AudioFormat,
|
||||
Duration,
|
||||
ParameterValues,
|
||||
SynthesisRequest,
|
||||
SynthesizedAudio,
|
||||
VoiceSelection,
|
||||
)
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
# SuperTonic voice list - source of truth
|
||||
_SUPERTONIC_VOICES = ("M1", "M2", "M3", "M4", "M5", "F1", "F2", "F3", "F4", "F5")
|
||||
|
||||
# Voice display names mapping
|
||||
_VOICE_DISPLAY_NAMES: dict[str, str] = {
|
||||
"M1": "Male 1",
|
||||
"M2": "Male 2",
|
||||
"M3": "Male 3",
|
||||
"M4": "Male 4",
|
||||
"M5": "Male 5",
|
||||
"F1": "Female 1",
|
||||
"F2": "Female 2",
|
||||
"F3": "Female 3",
|
||||
"F4": "Female 4",
|
||||
"F5": "Female 5",
|
||||
}
|
||||
|
||||
# Sample rate for SuperTonic audio
|
||||
_SUPERTONIC_SAMPLE_RATE = 24000
|
||||
|
||||
|
||||
class SuperTonicSession:
|
||||
"""EngineSession implementation for SuperTonic.
|
||||
|
||||
Owns mutable execution state for synthesis.
|
||||
NOT thread-safe.
|
||||
"""
|
||||
|
||||
def __init__(self, pipeline: Any) -> None:
|
||||
self._pipeline = pipeline
|
||||
self._disposed = False
|
||||
|
||||
def synthesize(self, request: SynthesisRequest) -> SynthesizedAudio:
|
||||
"""Synthesize audio from text using SuperTonic."""
|
||||
if self._disposed:
|
||||
raise EngineError("Session disposed")
|
||||
|
||||
try:
|
||||
import soundfile as sf
|
||||
|
||||
voice = request.voice.key
|
||||
speed = float(request.parameters.values.get("speed", 1.0))
|
||||
total_steps = request.parameters.values.get("total_steps", None)
|
||||
split_pattern = request.parameters.values.get("split_pattern", None)
|
||||
|
||||
if total_steps is not None:
|
||||
total_steps = int(total_steps)
|
||||
|
||||
audio_parts: list[np.ndarray] = []
|
||||
for segment in self._pipeline(
|
||||
request.text,
|
||||
voice=voice,
|
||||
speed=speed,
|
||||
split_pattern=split_pattern,
|
||||
total_steps=total_steps,
|
||||
):
|
||||
audio_parts.append(segment.audio)
|
||||
|
||||
if not audio_parts:
|
||||
return SynthesizedAudio(
|
||||
data=b"",
|
||||
format=AudioFormat(mime="audio/wav", extension="wav"),
|
||||
duration=Duration(seconds=0.0),
|
||||
)
|
||||
|
||||
combined = np.concatenate(audio_parts).astype("float32", copy=False)
|
||||
buf = io.BytesIO()
|
||||
sf.write(buf, combined, self._pipeline.sample_rate, format="WAV")
|
||||
audio_bytes = buf.getvalue()
|
||||
duration_seconds = len(combined) / self._pipeline.sample_rate
|
||||
|
||||
return SynthesizedAudio(
|
||||
data=audio_bytes,
|
||||
format=AudioFormat(mime="audio/wav", extension="wav"),
|
||||
duration=Duration(seconds=duration_seconds),
|
||||
)
|
||||
except EngineError:
|
||||
raise
|
||||
except Exception as e:
|
||||
raise EngineError(f"Synthesis failed: {e}") from e
|
||||
|
||||
def dispose(self) -> None:
|
||||
"""Release session resources. Idempotent."""
|
||||
self._disposed = True
|
||||
|
||||
|
||||
class SuperTonicEngine:
|
||||
"""Engine implementation for SuperTonic.
|
||||
|
||||
Factory for SuperTonicSession instances. Stateless and thread-safe.
|
||||
"""
|
||||
|
||||
def __init__(self, pipeline: Any) -> None:
|
||||
self._pipeline = pipeline
|
||||
self._disposed = False
|
||||
|
||||
def createSession(self) -> SuperTonicSession:
|
||||
"""Create a new SuperTonicSession."""
|
||||
if self._disposed:
|
||||
raise EngineError("Engine disposed")
|
||||
return SuperTonicSession(self._pipeline)
|
||||
|
||||
def dispose(self) -> None:
|
||||
"""Release engine resources. Idempotent."""
|
||||
self._disposed = True
|
||||
|
||||
def listVoices(self, sourceId: str) -> list[VoiceManifest]:
|
||||
"""List available SuperTonic voices. Implements VoiceLister capability."""
|
||||
if self._disposed:
|
||||
raise EngineError("Engine disposed")
|
||||
return [
|
||||
VoiceManifest(
|
||||
id=voice_id,
|
||||
name=_VOICE_DISPLAY_NAMES.get(voice_id, voice_id),
|
||||
tags=(_get_gender_tag(voice_id),),
|
||||
)
|
||||
for voice_id in _SUPERTONIC_VOICES
|
||||
]
|
||||
|
||||
|
||||
def _get_gender_tag(voice_id: str) -> str:
|
||||
"""Extract gender tag from voice ID."""
|
||||
if voice_id.startswith("M"):
|
||||
return "male"
|
||||
elif voice_id.startswith("F"):
|
||||
return "female"
|
||||
return "unknown"
|
||||
Reference in New Issue
Block a user