switching to high quality piper tts and added label translations
This commit is contained in:
Matthias Hinrichs
@@ -0,0 +1,290 @@
|
||||
"""Flask web server with HTTP API for Piper."""
|
||||
|
||||
import argparse
|
||||
import io
|
||||
import json
|
||||
import logging
|
||||
import wave
|
||||
from pathlib import Path
|
||||
from typing import Any, Dict, List, Optional
|
||||
from urllib.request import urlopen
|
||||
|
||||
from flask import Flask, request
|
||||
|
||||
from . import PiperVoice, SynthesisConfig
|
||||
from .download_voices import VOICES_JSON, download_voice
|
||||
|
||||
_LOGGER = logging.getLogger()
|
||||
|
||||
|
||||
def main() -> None:
|
||||
"""Run HTTP server."""
|
||||
parser = argparse.ArgumentParser()
|
||||
parser.add_argument("--host", default="0.0.0.0", help="HTTP server host")
|
||||
parser.add_argument("--port", type=int, default=5000, help="HTTP server port")
|
||||
#
|
||||
parser.add_argument("-m", "--model", required=True, help="Path to Onnx model file")
|
||||
#
|
||||
parser.add_argument("-s", "--speaker", type=int, help="Id of speaker (default: 0)")
|
||||
parser.add_argument(
|
||||
"--length-scale", "--length_scale", type=float, help="Phoneme length"
|
||||
)
|
||||
parser.add_argument(
|
||||
"--noise-scale", "--noise_scale", type=float, help="Generator noise"
|
||||
)
|
||||
parser.add_argument(
|
||||
"--noise-w-scale",
|
||||
"--noise_w_scale",
|
||||
"--noise-w",
|
||||
"--noise_w",
|
||||
type=float,
|
||||
help="Phoneme width noise",
|
||||
)
|
||||
#
|
||||
parser.add_argument("--cuda", action="store_true", help="Use GPU")
|
||||
#
|
||||
parser.add_argument(
|
||||
"--sentence-silence",
|
||||
"--sentence_silence",
|
||||
type=float,
|
||||
default=0.0,
|
||||
help="Seconds of silence after each sentence",
|
||||
)
|
||||
#
|
||||
parser.add_argument(
|
||||
"--data-dir",
|
||||
"--data_dir",
|
||||
action="append",
|
||||
default=[str(Path.cwd())],
|
||||
help="Data directory to check for downloaded models (default: current directory)",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--download-dir",
|
||||
"--download_dir",
|
||||
help="Path to download voices (default: first data dir)",
|
||||
)
|
||||
#
|
||||
parser.add_argument(
|
||||
"--debug", action="store_true", help="Print DEBUG messages to console"
|
||||
)
|
||||
args = parser.parse_args()
|
||||
logging.basicConfig(level=logging.DEBUG if args.debug else logging.INFO)
|
||||
_LOGGER.debug(args)
|
||||
|
||||
if not args.download_dir:
|
||||
# Download voices to first data directory if not specified
|
||||
args.download_dir = args.data_dir[0]
|
||||
|
||||
download_dir = Path(args.download_dir)
|
||||
|
||||
# Download voice if file doesn't exist
|
||||
model_path = Path(args.model)
|
||||
if not model_path.exists():
|
||||
# Look in data directories
|
||||
voice_name = args.model
|
||||
for data_dir in args.data_dir:
|
||||
maybe_model_path = Path(data_dir) / f"{voice_name}.onnx"
|
||||
_LOGGER.debug("Checking '%s'", maybe_model_path)
|
||||
if maybe_model_path.exists():
|
||||
model_path = maybe_model_path
|
||||
break
|
||||
|
||||
if not model_path.exists():
|
||||
raise ValueError(
|
||||
f"Unable to find voice: {model_path} (use piper.download_voices)"
|
||||
)
|
||||
|
||||
default_model_id = model_path.name.rstrip(".onnx")
|
||||
|
||||
# Load voice
|
||||
default_voice = PiperVoice.load(model_path, use_cuda=args.cuda)
|
||||
loaded_voices: Dict[str, PiperVoice] = {default_model_id: default_voice}
|
||||
|
||||
# Create web server
|
||||
app = Flask(__name__)
|
||||
|
||||
@app.route("/voices", methods=["GET"])
|
||||
def app_voices() -> Dict[str, Any]:
|
||||
"""List downloaded voices.
|
||||
|
||||
Outputs a JSON object with the format:
|
||||
{
|
||||
"<voice name>": { <voice config> },
|
||||
...
|
||||
}
|
||||
|
||||
for each voice in your data directories.
|
||||
"""
|
||||
voices_dict: Dict[str, Any] = {}
|
||||
config_paths: List[Path] = [Path(f"{model_path}.json")]
|
||||
|
||||
for data_dir in args.data_dir:
|
||||
for onnx_path in Path(data_dir).glob("*.onnx"):
|
||||
config_path = Path(f"{onnx_path}.json")
|
||||
if config_path.exists():
|
||||
config_paths.append(config_path)
|
||||
|
||||
for config_path in config_paths:
|
||||
model_id = config_path.name.rstrip(".onnx.json")
|
||||
if model_id in voices_dict:
|
||||
continue
|
||||
|
||||
with open(config_path, "r", encoding="utf-8") as config_file:
|
||||
voices_dict[model_id] = json.load(config_file)
|
||||
|
||||
return voices_dict
|
||||
|
||||
@app.route("/all-voices", methods=["GET"])
|
||||
def app_all_voices() -> Dict[str, Any]:
|
||||
"""List all Piper voices.
|
||||
|
||||
Outputs voices.json from the piper-voices repo on HuggingFace.
|
||||
See: https://huggingface.co/rhasspy/piper-voices
|
||||
"""
|
||||
with urlopen(VOICES_JSON) as response:
|
||||
return json.load(response)
|
||||
|
||||
@app.route("/download", methods=["POST"])
|
||||
def app_download() -> str:
|
||||
"""Download a voice.
|
||||
|
||||
Downloads the .onnx and .onnx.json file from piper-voices repo on HuggingFace.
|
||||
See: https://huggingface.co/rhasspy/piper-voices
|
||||
|
||||
Expects a JSON object with the format:
|
||||
{
|
||||
"voice": "<voice name>", (required)
|
||||
"force_redownload": false (optional)
|
||||
}
|
||||
|
||||
Returns the name of the voice.
|
||||
Voice format must be <language>-<name>-<quality> like "en_US-lessac-medium".
|
||||
"""
|
||||
data = json.loads(request.data)
|
||||
model_id = data.get("voice")
|
||||
if not model_id:
|
||||
raise ValueError("voice is required")
|
||||
|
||||
force_redownload = data.get("force_redownload", False)
|
||||
download_voice(model_id, download_dir, force_redownload=force_redownload)
|
||||
|
||||
return model_id
|
||||
|
||||
@app.route("/", methods=["POST"])
|
||||
def app_synthesize() -> bytes:
|
||||
"""Synthesize audio from text.
|
||||
|
||||
Expects a JSON object with the format:
|
||||
{
|
||||
"text": "Text to speak.", (required)
|
||||
"voice": "<voice name>", (optional)
|
||||
"speaker": "<speaker name>", (optional)
|
||||
"speaker_id": "<speaker id>", (optional, overrides speaker)
|
||||
"length_scale": 1.0, (optional)
|
||||
"noise_scale": 0.667, (optional)
|
||||
"length_w_scale": 0.8 (optional)
|
||||
}
|
||||
"""
|
||||
data = json.loads(request.data)
|
||||
text = data.get("text", "").strip()
|
||||
if not text:
|
||||
raise ValueError("No text provided")
|
||||
|
||||
_LOGGER.debug(data)
|
||||
|
||||
model_id = data.get("voice", default_model_id)
|
||||
voice = loaded_voices.get(model_id)
|
||||
if voice is None:
|
||||
for data_dir in args.data_dir:
|
||||
maybe_model_path = Path(data_dir) / f"{model_id}.onnx"
|
||||
if maybe_model_path.exists():
|
||||
_LOGGER.debug("Loading voice %s", model_id)
|
||||
voice = PiperVoice.load(maybe_model_path, use_cuda=args.cuda)
|
||||
loaded_voices[model_id] = voice
|
||||
break
|
||||
|
||||
if voice is None:
|
||||
_LOGGER.warning("Voice not found: %s. Using default voice.", model_id)
|
||||
voice = default_voice
|
||||
|
||||
speaker_id: Optional[int] = data.get("speaker_id")
|
||||
if (voice.config.num_speakers > 1) and (speaker_id is None):
|
||||
speaker = data.get("speaker")
|
||||
if speaker:
|
||||
speaker_id = voice.config.speaker_id_map.get(speaker)
|
||||
|
||||
if speaker_id is None:
|
||||
_LOGGER.warning(
|
||||
"Speaker not found: '%s' in %s",
|
||||
speaker,
|
||||
voice.config.speaker_id_map.keys(),
|
||||
)
|
||||
speaker_id = args.speaker or 0
|
||||
|
||||
if (speaker_id is not None) and (speaker_id > voice.config.num_speakers):
|
||||
speaker_id = 0
|
||||
|
||||
syn_config = SynthesisConfig(
|
||||
speaker_id=speaker_id,
|
||||
length_scale=float(
|
||||
data.get(
|
||||
"length_scale",
|
||||
(
|
||||
args.length_scale
|
||||
if args.length_scale is not None
|
||||
else voice.config.length_scale
|
||||
),
|
||||
)
|
||||
),
|
||||
noise_scale=float(
|
||||
data.get(
|
||||
"noise_scale",
|
||||
(
|
||||
args.noise_scale
|
||||
if args.noise_scale is not None
|
||||
else voice.config.noise_scale
|
||||
),
|
||||
)
|
||||
),
|
||||
noise_w_scale=float(
|
||||
data.get(
|
||||
"noise_w_scale",
|
||||
(
|
||||
args.noise_w_scale
|
||||
if args.noise_w_scale is not None
|
||||
else voice.config.noise_w_scale
|
||||
),
|
||||
)
|
||||
),
|
||||
)
|
||||
|
||||
_LOGGER.debug("Synthesizing text: '%s' with config=%s", text, syn_config)
|
||||
with io.BytesIO() as wav_io:
|
||||
wav_file: wave.Wave_write = wave.open(wav_io, "wb")
|
||||
with wav_file:
|
||||
wav_params_set = False
|
||||
for i, audio_chunk in enumerate(voice.synthesize(text, syn_config)):
|
||||
if not wav_params_set:
|
||||
wav_file.setframerate(audio_chunk.sample_rate)
|
||||
wav_file.setsampwidth(audio_chunk.sample_width)
|
||||
wav_file.setnchannels(audio_chunk.sample_channels)
|
||||
wav_params_set = True
|
||||
|
||||
if i > 0:
|
||||
wav_file.writeframes(
|
||||
bytes(
|
||||
int(
|
||||
voice.config.sample_rate * args.sentence_silence * 2
|
||||
)
|
||||
)
|
||||
)
|
||||
|
||||
wav_file.writeframes(audio_chunk.audio_int16_bytes)
|
||||
|
||||
return wav_io.getvalue()
|
||||
|
||||
app.run(host=args.host, port=args.port)
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
Reference in New Issue
Block a user