Tools / Modules /Generate_Speech.py
Nymbo's picture
Update Modules/Generate_Speech.py
ae1b42d verified
raw
history blame
25 kB
from __future__ import annotations
import json
import os
import time
from contextlib import contextmanager
from typing import Optional, Annotated
from unicodedata import normalize
import re
import uuid
import io
import wave
import numpy as np
import onnxruntime as ort
import scipy.io.wavfile
import gradio as gr
from .File_System import ROOT_DIR
from app import _log_call_end, _log_call_start, _truncate_for_log
from ._docstrings import autodoc
try:
import torch # type: ignore
except Exception: # pragma: no cover
torch = None # type: ignore
try:
from kokoro import KModel, KPipeline # type: ignore
except Exception: # pragma: no cover
KModel = None # type: ignore
KPipeline = None # type: ignore
try:
from huggingface_hub import snapshot_download, list_repo_files
except ImportError:
snapshot_download = None
list_repo_files = None
# --- Supertonic Helper Classes & Functions ---
class UnicodeProcessor:
def __init__(self, unicode_indexer_path: str):
with open(unicode_indexer_path, "r") as f:
self.indexer = json.load(f)
def _preprocess_text(self, text: str) -> str:
# TODO: add more preprocessing
text = normalize("NFKD", text)
return text
def _get_text_mask(self, text_ids_lengths: np.ndarray) -> np.ndarray:
text_mask = length_to_mask(text_ids_lengths)
return text_mask
def _text_to_unicode_values(self, text: str) -> np.ndarray:
unicode_values = np.array(
[ord(char) for char in text], dtype=np.uint16
) # 2 bytes
return unicode_values
def __call__(self, text_list: list[str]) -> tuple[np.ndarray, np.ndarray]:
text_list = [self._preprocess_text(t) for t in text_list]
text_ids_lengths = np.array([len(text) for text in text_list], dtype=np.int64)
text_ids = np.zeros((len(text_list), text_ids_lengths.max()), dtype=np.int64)
for i, text in enumerate(text_list):
unicode_vals = self._text_to_unicode_values(text)
text_ids[i, : len(unicode_vals)] = np.array(
[self.indexer[val] for val in unicode_vals], dtype=np.int64
)
text_mask = self._get_text_mask(text_ids_lengths)
return text_ids, text_mask
class Style:
def __init__(self, style_ttl_onnx: np.ndarray, style_dp_onnx: np.ndarray):
self.ttl = style_ttl_onnx
self.dp = style_dp_onnx
class TextToSpeech:
def __init__(
self,
cfgs: dict,
text_processor: UnicodeProcessor,
dp_ort: ort.InferenceSession,
text_enc_ort: ort.InferenceSession,
vector_est_ort: ort.InferenceSession,
vocoder_ort: ort.InferenceSession,
):
self.cfgs = cfgs
self.text_processor = text_processor
self.dp_ort = dp_ort
self.text_enc_ort = text_enc_ort
self.vector_est_ort = vector_est_ort
self.vocoder_ort = vocoder_ort
self.sample_rate = cfgs["ae"]["sample_rate"]
self.base_chunk_size = cfgs["ae"]["base_chunk_size"]
self.chunk_compress_factor = cfgs["ttl"]["chunk_compress_factor"]
self.ldim = cfgs["ttl"]["latent_dim"]
def sample_noisy_latent(
self, duration: np.ndarray
) -> tuple[np.ndarray, np.ndarray]:
bsz = len(duration)
wav_len_max = duration.max() * self.sample_rate
wav_lengths = (duration * self.sample_rate).astype(np.int64)
chunk_size = self.base_chunk_size * self.chunk_compress_factor
latent_len = ((wav_len_max + chunk_size - 1) / chunk_size).astype(np.int32)
latent_dim = self.ldim * self.chunk_compress_factor
noisy_latent = np.random.randn(bsz, latent_dim, latent_len).astype(np.float32)
latent_mask = get_latent_mask(
wav_lengths, self.base_chunk_size, self.chunk_compress_factor
)
noisy_latent = noisy_latent * latent_mask
return noisy_latent, latent_mask
def _infer(
self, text_list: list[str], style: Style, total_step: int, speed: float = 1.05
) -> tuple[np.ndarray, np.ndarray]:
assert (
len(text_list) == style.ttl.shape[0]
), "Number of texts must match number of style vectors"
bsz = len(text_list)
text_ids, text_mask = self.text_processor(text_list)
dur_onnx, *_ = self.dp_ort.run(
None, {"text_ids": text_ids, "style_dp": style.dp, "text_mask": text_mask}
)
dur_onnx = dur_onnx / speed
text_emb_onnx, *_ = self.text_enc_ort.run(
None,
{"text_ids": text_ids, "style_ttl": style.ttl, "text_mask": text_mask},
) # dur_onnx: [bsz]
xt, latent_mask = self.sample_noisy_latent(dur_onnx)
total_step_np = np.array([total_step] * bsz, dtype=np.float32)
for step in range(total_step):
current_step = np.array([step] * bsz, dtype=np.float32)
xt, *_ = self.vector_est_ort.run(
None,
{
"noisy_latent": xt,
"text_emb": text_emb_onnx,
"style_ttl": style.ttl,
"text_mask": text_mask,
"latent_mask": latent_mask,
"current_step": current_step,
"total_step": total_step_np,
},
)
wav, *_ = self.vocoder_ort.run(None, {"latent": xt})
return wav, dur_onnx
def __call__(
self,
text: str,
style: Style,
total_step: int,
speed: float = 1.05,
silence_duration: float = 0.3,
max_len: int = 300,
) -> tuple[np.ndarray, np.ndarray]:
assert (
style.ttl.shape[0] == 1
), "Single speaker text to speech only supports single style"
text_list = chunk_text(text, max_len=max_len)
wav_cat = None
dur_cat = None
for text in text_list:
wav, dur_onnx = self._infer([text], style, total_step, speed)
if wav_cat is None:
wav_cat = wav
dur_cat = dur_onnx
else:
silence = np.zeros(
(1, int(silence_duration * self.sample_rate)), dtype=np.float32
)
wav_cat = np.concatenate([wav_cat, silence, wav], axis=1)
dur_cat += dur_onnx + silence_duration
return wav_cat, dur_cat
def stream(
self,
text: str,
style: Style,
total_step: int,
speed: float = 1.05,
silence_duration: float = 0.3,
max_len: int = 300,
):
assert (
style.ttl.shape[0] == 1
), "Single speaker text to speech only supports single style"
text_list = chunk_text(text, max_len=max_len)
for i, text in enumerate(text_list):
wav, _ = self._infer([text], style, total_step, speed)
yield wav.flatten()
if i < len(text_list) - 1:
silence = np.zeros(
(int(silence_duration * self.sample_rate),), dtype=np.float32
)
yield silence
def batch(
self, text_list: list[str], style: Style, total_step: int, speed: float = 1.05
) -> tuple[np.ndarray, np.ndarray]:
return self._infer(text_list, style, total_step, speed)
def length_to_mask(lengths: np.ndarray, max_len: Optional[int] = None) -> np.ndarray:
"""
Convert lengths to binary mask.
Args:
lengths: (B,)
max_len: int
Returns:
mask: (B, 1, max_len)
"""
max_len = max_len or lengths.max()
ids = np.arange(0, max_len)
mask = (ids < np.expand_dims(lengths, axis=1)).astype(np.float32)
return mask.reshape(-1, 1, max_len)
def get_latent_mask(
wav_lengths: np.ndarray, base_chunk_size: int, chunk_compress_factor: int
) -> np.ndarray:
latent_size = base_chunk_size * chunk_compress_factor
latent_lengths = (wav_lengths + latent_size - 1) // latent_size
latent_mask = length_to_mask(latent_lengths)
return latent_mask
def load_onnx(
onnx_path: str, opts: ort.SessionOptions, providers: list[str]
) -> ort.InferenceSession:
return ort.InferenceSession(onnx_path, sess_options=opts, providers=providers)
def load_onnx_all(
onnx_dir: str, opts: ort.SessionOptions, providers: list[str]
) -> tuple[
ort.InferenceSession,
ort.InferenceSession,
ort.InferenceSession,
ort.InferenceSession,
]:
dp_onnx_path = os.path.join(onnx_dir, "duration_predictor.onnx")
text_enc_onnx_path = os.path.join(onnx_dir, "text_encoder.onnx")
vector_est_onnx_path = os.path.join(onnx_dir, "vector_estimator.onnx")
vocoder_onnx_path = os.path.join(onnx_dir, "vocoder.onnx")
dp_ort = load_onnx(dp_onnx_path, opts, providers)
text_enc_ort = load_onnx(text_enc_onnx_path, opts, providers)
vector_est_ort = load_onnx(vector_est_onnx_path, opts, providers)
vocoder_ort = load_onnx(vocoder_onnx_path, opts, providers)
return dp_ort, text_enc_ort, vector_est_ort, vocoder_ort
def load_cfgs(onnx_dir: str) -> dict:
cfg_path = os.path.join(onnx_dir, "tts.json")
with open(cfg_path, "r") as f:
cfgs = json.load(f)
return cfgs
def load_text_processor(onnx_dir: str) -> UnicodeProcessor:
unicode_indexer_path = os.path.join(onnx_dir, "unicode_indexer.json")
text_processor = UnicodeProcessor(unicode_indexer_path)
return text_processor
def load_text_to_speech(onnx_dir: str, use_gpu: bool = False) -> TextToSpeech:
opts = ort.SessionOptions()
if use_gpu:
raise NotImplementedError("GPU mode is not fully tested")
else:
providers = ["CPUExecutionProvider"]
print("Using CPU for inference")
cfgs = load_cfgs(onnx_dir)
dp_ort, text_enc_ort, vector_est_ort, vocoder_ort = load_onnx_all(
onnx_dir, opts, providers
)
text_processor = load_text_processor(onnx_dir)
return TextToSpeech(
cfgs, text_processor, dp_ort, text_enc_ort, vector_est_ort, vocoder_ort
)
def load_voice_style(voice_style_paths: list[str], verbose: bool = False) -> Style:
bsz = len(voice_style_paths)
# Read first file to get dimensions
with open(voice_style_paths[0], "r") as f:
first_style = json.load(f)
ttl_dims = first_style["style_ttl"]["dims"]
dp_dims = first_style["style_dp"]["dims"]
# Pre-allocate arrays with full batch size
ttl_style = np.zeros([bsz, ttl_dims[1], ttl_dims[2]], dtype=np.float32)
dp_style = np.zeros([bsz, dp_dims[1], dp_dims[2]], dtype=np.float32)
# Fill in the data
for i, voice_style_path in enumerate(voice_style_paths):
with open(voice_style_path, "r") as f:
voice_style = json.load(f)
ttl_data = np.array(
voice_style["style_ttl"]["data"], dtype=np.float32
).flatten()
ttl_style[i] = ttl_data.reshape(ttl_dims[1], ttl_dims[2])
dp_data = np.array(
voice_style["style_dp"]["data"], dtype=np.float32
).flatten()
dp_style[i] = dp_data.reshape(dp_dims[1], dp_dims[2])
if verbose:
print(f"Loaded {bsz} voice styles")
return Style(ttl_style, dp_style)
@contextmanager
def timer(name: str):
start = time.time()
print(f"{name}...")
yield
print(f" -> {name} completed in {time.time() - start:.2f} sec")
def sanitize_filename(text: str, max_len: int) -> str:
"""Sanitize filename by replacing non-alphanumeric characters with underscores"""
prefix = text[:max_len]
return re.sub(r"[^a-zA-Z0-9]", "_", prefix)
def chunk_text(text: str, max_len: int = 300) -> list[str]:
"""
Split text into chunks by paragraphs and sentences.
Args:
text: Input text to chunk
max_len: Maximum length of each chunk (default: 300)
Returns:
List of text chunks
"""
# Split by paragraph (two or more newlines)
paragraphs = [p.strip() for p in re.split(r"\n\s*\n+", text.strip()) if p.strip()]
chunks = []
for paragraph in paragraphs:
paragraph = paragraph.strip()
if not paragraph:
continue
# Split by sentence boundaries (period, question mark, exclamation mark followed by space)
# But exclude common abbreviations like Mr., Mrs., Dr., etc. and single capital letters like F.
pattern = r"(?<!Mr\.)(?<!Mrs\.)(?<!Ms\.)(?<!Dr\.)(?<!Prof\.)(?<!Sr\.)(?<!Jr\.)(?<!Ph\.D\.)(?<!etc\.)(?<!e\.g\.)(?<!i\.e\.)(?<!vs\.)(?<!Inc\.)(?<!Ltd\.)(?<!Co\.)(?<!Corp\.)(?<!St\.)(?<!Ave\.)(?<!Blvd\.)(?<!\b[A-Z]\.)(?<=[.!?])\s+"
sentences = re.split(pattern, paragraph)
current_chunk = ""
for sentence in sentences:
if len(current_chunk) + len(sentence) + 1 <= max_len:
current_chunk += (" " if current_chunk else "") + sentence
else:
if current_chunk:
chunks.append(current_chunk.strip())
current_chunk = sentence
if current_chunk:
chunks.append(current_chunk.strip())
return chunks
# --- Main Tool Logic ---
# --- Kokoro State ---
_KOKORO_STATE = {
"initialized": False,
"device": "cpu",
"model": None,
"pipelines": {},
}
# --- Supertonic State ---
_SUPERTONIC_STATE = {
"initialized": False,
"tts": None,
"assets_dir": None,
}
def _audio_np_to_int16(audio_np: np.ndarray) -> np.ndarray:
audio_clipped = np.clip(audio_np, -1.0, 1.0)
return (audio_clipped * 32767.0).astype(np.int16)
# --- Kokoro Functions ---
def get_kokoro_voices() -> list[str]:
try:
if list_repo_files:
files = list_repo_files("hexgrad/Kokoro-82M")
voice_files = [file for file in files if file.endswith(".pt") and file.startswith("voices/")]
voices = [file.replace("voices/", "").replace(".pt", "") for file in voice_files]
return sorted(voices) if voices else _get_fallback_voices()
return _get_fallback_voices()
except Exception:
return _get_fallback_voices()
def _get_fallback_voices() -> list[str]:
return [
"af_alloy", "af_aoede", "af_bella", "af_heart", "af_jessica", "af_kore", "af_nicole", "af_nova", "af_river", "af_sarah", "af_sky",
"am_adam", "am_echo", "am_eric", "am_fenrir", "am_liam", "am_michael", "am_onyx", "am_puck", "am_santa",
"bf_alice", "bf_emma", "bf_isabella", "bf_lily",
"bm_daniel", "bm_fable", "bm_george", "bm_lewis",
"ef_dora", "em_alex", "em_santa",
"ff_siwis",
"hf_alpha", "hf_beta", "hm_omega", "hm_psi",
"if_sara", "im_nicola",
"jf_alpha", "jf_gongitsune", "jf_nezumi", "jf_tebukuro", "jm_kumo",
"pf_dora", "pm_alex", "pm_santa",
"zf_xiaobei", "zf_xiaoni", "zf_xiaoxiao", "zf_xiaoyi",
"zm_yunjian", "zm_yunxi", "zm_yunxia", "zm_yunyang",
]
def _init_kokoro() -> None:
if _KOKORO_STATE["initialized"]:
return
if KModel is None or KPipeline is None:
raise RuntimeError("Kokoro is not installed. Please install the 'kokoro' package (>=0.9.4).")
device = "cpu"
if torch is not None:
try:
if torch.cuda.is_available():
device = "cuda"
except Exception:
device = "cpu"
model = KModel(repo_id="hexgrad/Kokoro-82M").to(device).eval()
pipelines = {"a": KPipeline(lang_code="a", model=False, repo_id="hexgrad/Kokoro-82M")}
try:
pipelines["a"].g2p.lexicon.golds["kokoro"] = "kˈOkəɹO"
except Exception:
pass
_KOKORO_STATE.update({"initialized": True, "device": device, "model": model, "pipelines": pipelines})
# --- Supertonic Functions ---
def _init_supertonic() -> None:
if _SUPERTONIC_STATE["initialized"]:
return
if snapshot_download is None:
raise RuntimeError("huggingface_hub is not installed.")
# Use a local assets directory within Nymbo-Tools
# Assuming this file is in Nymbo-Tools/Modules
base_dir = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
assets_dir = os.path.join(base_dir, "assets", "supertonic")
if not os.path.exists(assets_dir):
print(f"Downloading Supertonic models to {assets_dir}...")
snapshot_download(repo_id="Supertone/supertonic", local_dir=assets_dir)
onnx_dir = os.path.join(assets_dir, "onnx")
tts = load_text_to_speech(onnx_dir, use_gpu=False)
_SUPERTONIC_STATE.update({"initialized": True, "tts": tts, "assets_dir": assets_dir})
def get_supertonic_voices() -> list[str]:
# We need assets to list voices. If not initialized, try to find them or init.
if not _SUPERTONIC_STATE["initialized"]:
# Check if assets exist without full init
base_dir = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
assets_dir = os.path.join(base_dir, "assets", "supertonic")
if not os.path.exists(assets_dir):
# If we can't list, return a default list or empty
return ["F1", "F2", "M1", "M2"] # Known defaults
else:
assets_dir = _SUPERTONIC_STATE["assets_dir"]
voice_styles_dir = os.path.join(assets_dir, "voice_styles")
if not os.path.exists(voice_styles_dir):
return ["F1", "F2", "M1", "M2"]
files = os.listdir(voice_styles_dir)
voices = [f.replace('.json', '') for f in files if f.endswith('.json')]
return sorted(voices)
def List_Kokoro_Voices() -> list[str]:
return get_kokoro_voices()
def List_Supertonic_Voices() -> list[str]:
return get_supertonic_voices()
# Single source of truth for the LLM-facing tool description
TOOL_SUMMARY = (
"Synthesize speech from text using Supertonic-66M (default) or Kokoro-82M. "
"Supertonic: faster, supports steps/silence/chunking. "
"Kokoro: slower, supports many languages/accents. "
"Return the generated media to the user in this format `![Alt text](URL)`."
)
@autodoc(
summary=TOOL_SUMMARY,
)
def Generate_Speech(
text: Annotated[str, "The text to synthesize (English)."],
model: Annotated[str, "The TTS model to use: 'Supertonic' or 'Kokoro'."] = "Supertonic",
speed: Annotated[float, "Speech speed multiplier in 0.5–2.0; 1.0 = normal speed."] = 1.3,
steps: Annotated[int, "Supertonic only. Diffusion steps (1-50). Higher = better quality but slower."] = 5,
voice: Annotated[str, "Voice identifier. Default 'F1' for Supertonic, 'af_heart' for Kokoro."] = "F1",
silence_duration: Annotated[float, "Supertonic only. Silence duration between chunks (0.0-2.0s)."] = 0.3,
max_chunk_size: Annotated[int, "Supertonic only. Max text chunk length (50-1000)."] = 300,
) -> str:
_log_call_start("Generate_Speech", text=_truncate_for_log(text, 200), model=model, speed=speed, voice=voice)
if not text or not text.strip():
try:
_log_call_end("Generate_Speech", "error=empty text")
finally:
pass
raise gr.Error("Please provide non-empty text to synthesize.")
model_lower = model.lower()
# Handle default voice switching if user didn't specify appropriate voice for model
if model_lower == "kokoro" and voice == "F1":
voice = "af_heart"
elif model_lower == "supertonic" and voice == "af_heart":
voice = "F1"
try:
if model_lower == "kokoro":
return _generate_kokoro(text, speed, voice)
else:
# Default to Supertonic
return _generate_supertonic(text, speed, voice, steps, silence_duration, max_chunk_size)
except gr.Error as exc:
_log_call_end("Generate_Speech", f"gr_error={str(exc)}")
raise
except Exception as exc: # pylint: disable=broad-except
_log_call_end("Generate_Speech", f"error={str(exc)[:120]}")
raise gr.Error(f"Error during speech generation: {exc}")
def _generate_kokoro(text: str, speed: float, voice: str) -> str:
_init_kokoro()
model = _KOKORO_STATE["model"]
pipelines = _KOKORO_STATE["pipelines"]
pipeline = pipelines.get("a")
if pipeline is None:
raise gr.Error("Kokoro English pipeline not initialized.")
audio_segments = []
pack = pipeline.load_voice(voice)
segments = list(pipeline(text, voice, speed))
total_segments = len(segments)
for segment_idx, (text_chunk, ps, _) in enumerate(segments):
ref_s = pack[len(ps) - 1]
try:
audio = model(ps, ref_s, float(speed))
audio_segments.append(audio.detach().cpu().numpy())
if total_segments > 10 and (segment_idx + 1) % 5 == 0:
print(f"Progress: Generated {segment_idx + 1}/{total_segments} segments...")
except Exception as exc:
raise gr.Error(f"Error generating audio for segment {segment_idx + 1}: {exc}")
if not audio_segments:
raise gr.Error("No audio was generated (empty synthesis result).")
if len(audio_segments) == 1:
final_audio = audio_segments[0]
else:
final_audio = np.concatenate(audio_segments, axis=0)
if total_segments > 1:
duration = len(final_audio) / 24_000
print(f"Completed: {total_segments} segments concatenated into {duration:.1f} seconds of audio")
# Save to file
filename = f"speech_kokoro_{uuid.uuid4().hex[:8]}.wav"
output_path = os.path.join(ROOT_DIR, filename)
# Normalize to 16-bit PCM
audio_int16 = (final_audio * 32767).astype(np.int16)
scipy.io.wavfile.write(output_path, 24000, audio_int16)
_log_call_end("Generate_Speech", f"saved_to={os.path.basename(output_path)} duration_sec={len(final_audio)/24_000:.2f}")
return output_path
def _generate_supertonic(text: str, speed: float, voice: str, steps: int, silence_duration: float, max_chunk_size: int) -> str:
_init_supertonic()
tts = _SUPERTONIC_STATE["tts"]
assets_dir = _SUPERTONIC_STATE["assets_dir"]
voice_path = os.path.join(assets_dir, "voice_styles", f"{voice}.json")
if not os.path.exists(voice_path):
# Fallback or error?
# Try to find if it's just a name mismatch or use default
if not os.path.exists(voice_path):
raise gr.Error(f"Voice style {voice} not found for Supertonic.")
style = load_voice_style([voice_path])
sr = tts.sample_rate
# Supertonic returns a generator of chunks, or we can use __call__ for full audio
# Using __call__ to get full audio for saving
# But __call__ returns (wav_cat, dur_cat)
wav_cat, _ = tts(text, style, steps, speed, silence_duration, max_chunk_size)
if wav_cat is None or wav_cat.size == 0:
raise gr.Error("No audio generated.")
# wav_cat is (1, samples) float32
final_audio = wav_cat.flatten()
# Save to file
filename = f"speech_supertonic_{uuid.uuid4().hex[:8]}.wav"
output_path = os.path.join(ROOT_DIR, filename)
audio_int16 = _audio_np_to_int16(final_audio)
scipy.io.wavfile.write(output_path, sr, audio_int16)
_log_call_end("Generate_Speech", f"saved_to={os.path.basename(output_path)} duration_sec={len(final_audio)/sr:.2f}")
return output_path
def build_interface() -> gr.Interface:
kokoro_voices = get_kokoro_voices()
supertonic_voices = get_supertonic_voices()
all_voices = sorted(list(set(kokoro_voices + supertonic_voices)))
return gr.Interface(
fn=Generate_Speech,
inputs=[
gr.Textbox(label="Text", placeholder="Type text to synthesize…", lines=4, info="The text to synthesize (English)"),
gr.Dropdown(label="Model", choices=["Supertonic", "Kokoro"], value="Supertonic", info="The TTS model to use"),
gr.Slider(minimum=0.5, maximum=2.0, value=1.3, step=0.1, label="Speed", info="Speech speed multiplier (1.0 = normal)"),
gr.Slider(minimum=1, maximum=50, value=5, step=1, label="Steps", info="Supertonic only: Diffusion steps (1-50)"),
gr.Dropdown(
label="Voice",
choices=all_voices,
value="F1",
info="Select voice (F1/F2/M1/M2 for Supertonic, others for Kokoro)",
),
gr.Slider(minimum=0.0, maximum=2.0, value=0.3, step=0.1, label="Silence Duration", info="Supertonic only: Silence duration between chunks"),
gr.Slider(minimum=50, maximum=1000, value=300, step=10, label="Max Chunk Size", info="Supertonic only: Max text chunk length"),
],
outputs=gr.Audio(label="Audio", type="filepath", format="wav"),
title="Generate Speech",
description=(
"<div style=\"text-align:center\">Generate speech with Supertonic-66M or Kokoro-82M. Runs on CPU.</div>"
),
api_description=TOOL_SUMMARY,
flagging_mode="never",
)
__all__ = ["Generate_Speech", "List_Kokoro_Voices", "List_Supertonic_Voices", "build_interface"]