tts-service / core /__init__.py
jlopez00's picture
Upload folder using huggingface_hub
bbf5262 verified
raw
history blame
3.22 kB
import os
import subprocess
import sys
from functools import lru_cache
from rvc.lib.tools.prerequisites_download import prequisites_download_pipeline
from tts_service.utils import cache_path
from tts_service.voices import voice_manager
python = sys.executable
@lru_cache(maxsize=None)
def import_voice_converter():
from rvc.infer.infer import VoiceConverter
return VoiceConverter()
# TTS
def run_tts_script(
tts_text: str,
voice_name: str,
tts_rate: int,
) -> tuple[str, str]:
tts_script_path = os.path.join("rvc", "lib", "tools", "tts.py")
voice = voice_manager.voices[voice_name]
format = "wav"
output_tts_path = cache_path(voice.tts, "", tts_rate, tts_text, extension=format)
if not os.path.exists(output_tts_path):
command_tts = [
*map(
str,
[
python,
tts_script_path,
"", # tts_file
tts_text,
voice.tts,
tts_rate,
output_tts_path,
],
),
]
subprocess.run(command_tts)
output_rvc_path = cache_path(voice.tts, voice.name, tts_rate, tts_text, extension=format)
if not os.path.exists(output_rvc_path):
infer_pipeline = import_voice_converter()
infer_pipeline.convert_audio(
pitch=voice.pitch,
filter_radius=voice.filter_radius,
index_rate=voice.index_rate,
volume_envelope=voice.rms_mix_rate,
protect=voice.protect,
hop_length=voice.hop_length,
f0_method=voice.f0_method,
audio_input_path=str(output_tts_path),
audio_output_path=str(output_rvc_path),
model_path=voice.model,
index_path=voice.index,
split_audio=False,
f0_autotune=voice.autotune is not None,
f0_autotune_strength=voice.autotune,
clean_audio=voice.clean is not None,
clean_strength=voice.clean,
export_format=format.upper(),
upscale_audio=voice.upscale,
f0_file=None,
embedder_model=voice.embedder_model,
embedder_model_custom=None,
sid=0,
formant_shifting=None,
formant_qfrency=None,
formant_timbre=None,
post_process=None,
reverb=None,
pitch_shift=None,
limiter=None,
gain=None,
distortion=None,
chorus=None,
bitcrush=None,
clipping=None,
compressor=None,
delay=None,
sliders=None,
)
return "Text synthesized successfully.", str(output_rvc_path)
# Prerequisites
def run_prerequisites_script(
pretraineds_v1_f0: bool,
pretraineds_v1_nof0: bool,
pretraineds_v2_f0: bool,
pretraineds_v2_nof0: bool,
models: bool,
voices: bool,
):
prequisites_download_pipeline(
pretraineds_v1_f0,
pretraineds_v1_nof0,
pretraineds_v2_f0,
pretraineds_v2_nof0,
models,
voices,
)
return "Prerequisites installed successfully."