Files
AudioTranscriberCombine/gui/main_window.py
2026-10-24 06:27:17 +03:00

488 lines
20 KiB
Python
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
"""Main application window — dark theme, mode selector, staged/sequential pipelines."""
from __future__ import annotations
import shutil
from pathlib import Path
from typing import Optional
from PySide6.QtCore import Qt, QUrl
from PySide6.QtGui import QColor, QDesktopServices
from PySide6.QtWidgets import (
QCheckBox, QComboBox, QFileDialog, QFormLayout, QGroupBox,
QHBoxLayout, QLabel, QLineEdit, QMainWindow, QMessageBox,
QPlainTextEdit, QProgressBar, QPushButton, QSpinBox, QTabWidget,
QVBoxLayout, QWidget,
)
from core.config import ConvertConfig, PipelineConfig, SeparateConfig, TranscribeConfig
from gui.worker import PipelineWorker
LANGUAGES = ["auto", "ru", "en", "de", "fr", "es", "it", "uk", "ja", "zh", "ko"]
SAMPLE_RATES = [16000, 22050, 44100, 48000]
ENGINES = ["whisperx", "gigaam"]
STAGE_LABELS = {
"file": "Файлы",
"convert": "Конвертация",
"separate": "Отделение",
"transcribe": "Транскрибация",
}
MODE_LABELS = {
"sequential": "Последовательно (файл за файлом)",
"staged": "Пакетно (по этапам)",
}
def _detect_device_info() -> tuple[str, str]:
"""Return (display_text, color_hex) describing the compute device."""
try:
import torch
except ImportError:
return "⚠ PyTorch не установлен", "#f38ba8"
if torch.cuda.is_available():
try:
name = torch.cuda.get_device_name(0)
vram = torch.cuda.get_device_properties(0).total_memory / (1024 ** 3)
return f"✓ CUDA: {name} · {vram:.1f} GB VRAM", "#a6e3a1"
except Exception: # noqa: BLE001
return "✓ CUDA доступна", "#a6e3a1"
return "⚠ CUDA не обнаружена — расчёт пойдёт на CPU (медленно)", "#f9e2af"
class MainWindow(QMainWindow):
def __init__(self) -> None:
super().__init__()
self.setWindowTitle("Song Transcriber")
self.resize(1100, 850)
self._worker: Optional[PipelineWorker] = None
self._build_ui()
self._apply_defaults()
self._sync_tab_indicators()
self._sync_engine_fields()
self._log_device_info()
# ---------------- UI ----------------
def _build_ui(self) -> None:
central = QWidget()
self.setCentralWidget(central)
root = QVBoxLayout(central)
root.setContentsMargins(18, 16, 18, 16)
root.setSpacing(12)
# --- header row: title + device badge ---
header = QHBoxLayout()
title = QLabel("🎵 Song Transcriber")
title.setObjectName("appTitle")
header.addWidget(title)
header.addStretch(1)
self.device_label = QLabel("")
self.device_label.setObjectName("deviceInfo")
header.addWidget(self.device_label)
root.addLayout(header)
# --- top settings ---
top_box = QGroupBox("Общие настройки")
top_layout = QFormLayout(top_box)
top_layout.setSpacing(10)
self.input_edit = QLineEdit()
top_layout.addRow("Входная папка:", self._with_browse(self.input_edit, self._pick_input_dir))
self.output_edit = QLineEdit()
top_layout.addRow("Выходная папка:", self._with_browse(self.output_edit, self._pick_output_dir))
self.mode_cb = QComboBox()
for key, label in MODE_LABELS.items():
self.mode_cb.addItem(label, key)
top_layout.addRow("Режим:", self.mode_cb)
opts = QHBoxLayout()
self.recursive_cb = QCheckBox("Рекурсивный обход")
self.keep_intermediate_cb = QCheckBox("Сохранять промежуточные")
self.overwrite_cb = QCheckBox("Перезаписывать")
opts.addWidget(self.recursive_cb)
opts.addWidget(self.keep_intermediate_cb)
opts.addWidget(self.overwrite_cb)
opts.addStretch(1)
top_layout.addRow("Опции:", opts)
root.addWidget(top_box)
# --- stages ---
self.tabs = QTabWidget()
self.tabs.addTab(self._build_convert_tab(), "Конвертация")
self.tabs.addTab(self._build_separate_tab(), "Отделение вокала")
self.tabs.addTab(self._build_transcribe_tab(), "Транскрибация")
root.addWidget(self.tabs)
self.convert_enabled_cb.toggled.connect(self._sync_tab_indicators)
self.separate_enabled_cb.toggled.connect(self._sync_tab_indicators)
self.transcribe_enabled_cb.toggled.connect(self._sync_tab_indicators)
self.engine_cb.currentTextChanged.connect(self._sync_engine_fields)
# --- controls row ---
row = QHBoxLayout()
row.setSpacing(10)
self.start_btn = QPushButton("▶ Запустить")
self.start_btn.setObjectName("primary")
self.start_btn.clicked.connect(self._on_start)
self.stop_btn = QPushButton("⏹ Стоп")
self.stop_btn.setObjectName("danger")
self.stop_btn.setEnabled(False)
self.stop_btn.clicked.connect(self._on_stop)
self.open_out_btn = QPushButton("📂 Открыть результат")
self.open_out_btn.clicked.connect(self._open_output_dir)
self.save_preset_btn = QPushButton("💾 Сохранить пресет")
self.save_preset_btn.clicked.connect(self._save_preset)
self.load_preset_btn = QPushButton("📁 Загрузить пресет")
self.load_preset_btn.clicked.connect(self._load_preset)
row.addWidget(self.start_btn)
row.addWidget(self.stop_btn)
row.addStretch(1)
row.addWidget(self.open_out_btn)
row.addWidget(self.save_preset_btn)
row.addWidget(self.load_preset_btn)
root.addLayout(row)
# --- progress ---
self.progress = QProgressBar()
self.progress.setFormat("Ожидание")
root.addWidget(self.progress)
# --- log ---
self.log_view = QPlainTextEdit()
self.log_view.setObjectName("log")
self.log_view.setReadOnly(True)
self.log_view.setPlaceholderText("Здесь появится лог выполнения…")
root.addWidget(self.log_view, 1)
def _with_browse(self, edit: QLineEdit, picker) -> QWidget:
w = QWidget()
h = QHBoxLayout(w)
h.setContentsMargins(0, 0, 0, 0)
h.setSpacing(6)
h.addWidget(edit)
btn = QPushButton("…")
btn.setFixedWidth(36)
btn.clicked.connect(picker)
h.addWidget(btn)
return w
def _build_convert_tab(self) -> QWidget:
w = QWidget()
form = QFormLayout(w)
form.setContentsMargins(16, 16, 16, 16)
form.setSpacing(10)
self.convert_enabled_cb = QCheckBox("Включён")
form.addRow("", self.convert_enabled_cb)
self.convert_sr = QComboBox()
for sr in SAMPLE_RATES:
self.convert_sr.addItem(f"{sr} Hz", sr)
form.addRow("Sample rate:", self.convert_sr)
self.convert_ch = QComboBox()
self.convert_ch.addItem("Моно", 1)
self.convert_ch.addItem("Стерео", 2)
form.addRow("Каналы:", self.convert_ch)
self.convert_ffmpeg = QLineEdit()
form.addRow("ffmpeg:", self.convert_ffmpeg)
return w
def _build_separate_tab(self) -> QWidget:
w = QWidget()
form = QFormLayout(w)
form.setContentsMargins(16, 16, 16, 16)
form.setSpacing(10)
self.separate_enabled_cb = QCheckBox("Включён")
form.addRow("", self.separate_enabled_cb)
self.separate_model = QLineEdit()
form.addRow("Модель:", self.separate_model)
self.separate_stem = QLineEdit()
form.addRow("Isolate stem:", self.separate_stem)
self.separate_unblend = QLineEdit()
form.addRow("unblend CLI:", self.separate_unblend)
self.separate_kernels_cb = QCheckBox("Использовать кастомные ядра")
form.addRow("", self.separate_kernels_cb)
return w
def _build_transcribe_tab(self) -> QWidget:
w = QWidget()
form = QFormLayout(w)
form.setContentsMargins(16, 16, 16, 16)
form.setSpacing(10)
self.transcribe_enabled_cb = QCheckBox("Включён")
form.addRow("", self.transcribe_enabled_cb)
self.engine_cb = QComboBox()
self.engine_cb.addItems(ENGINES)
form.addRow("Движок:", self.engine_cb)
self.model_edit = QLineEdit()
form.addRow("Модель:", self.model_edit)
self.lang_cb = QComboBox()
self.lang_cb.setEditable(True)
self.lang_cb.addItems(LANGUAGES)
self.lang_cb.setToolTip(
"'auto' — WhisperX определяет язык на каждом файле отдельно.\n"
"Это заметно медленнее. Если знаешь язык — выбери его явно."
)
form.addRow("Язык:", self.lang_cb)
self.device_cb = QComboBox()
self.device_cb.addItems(["auto", "cuda", "cpu"])
form.addRow("Device:", self.device_cb)
self.compute_type_cb = QComboBox()
self.compute_type_cb.addItems(["auto", "int8_float16", "int8", "float16", "float32"])
form.addRow("Compute type:", self.compute_type_cb)
self.batch_spin = QSpinBox()
self.batch_spin.setRange(1, 512)
self.batch_spin.setToolTip(
"Размер батча сегментов ВНУТРИ одного файла (не файлов!).\n"
"Больше ≠ всегда быстрее: упирается в VRAM.\n"
"8–16 для 8 ГБ VRAM, 32–64 для 24 ГБ.\n"
"При 'CUDA out of memory' — уменьшайте."
)
form.addRow("Batch size:", self.batch_spin)
self.gigaam_provider_cb = QComboBox()
self.gigaam_provider_cb.addItems(["auto", "cuda", "cpu"])
form.addRow("GigaAM provider:", self.gigaam_provider_cb)
self.local_files_only_cb = QCheckBox("Не проверять HuggingFace (offline)")
self.local_files_only_cb.setToolTip(
"Если модель уже скачана — включите, чтобы не ждать сетевую проверку.\n"
"При первом запуске (модели нет) оставьте выключенным."
)
form.addRow("", self.local_files_only_cb)
self.vad_filter_cb = QCheckBox("Использовать VAD (pyannote)")
self.vad_filter_cb.setChecked(True)
self.vad_filter_cb.setToolTip(
"Отключите, если загрузка модели виснет на 'Performing voice activity detection'."
)
form.addRow("", self.vad_filter_cb)
return w
# ---------------- visual sync ----------------
def _log_device_info(self) -> None:
text, color = _detect_device_info()
self.device_label.setText(text)
self.device_label.setStyleSheet(f"color: {color};")
self._append_log(text)
def _sync_tab_indicators(self) -> None:
active = QColor("#89b4fa")
inactive = QColor("#585b70")
bar = self.tabs.tabBar()
bar.setTabTextColor(0, active if self.convert_enabled_cb.isChecked() else inactive)
bar.setTabTextColor(1, active if self.separate_enabled_cb.isChecked() else inactive)
bar.setTabTextColor(2, active if self.transcribe_enabled_cb.isChecked() else inactive)
def _sync_engine_fields(self) -> None:
is_whisperx = self.engine_cb.currentText() == "whisperx"
for w in (self.lang_cb, self.device_cb, self.compute_type_cb, self.batch_spin):
w.setEnabled(is_whisperx)
self.gigaam_provider_cb.setEnabled(not is_whisperx)
# ---------------- defaults & config <-> UI ----------------
def _apply_defaults(self) -> None:
self._apply_config(PipelineConfig())
def _apply_config(self, cfg: PipelineConfig) -> None:
self.input_edit.setText(cfg.input_dir)
self.output_edit.setText(str(Path(cfg.output_dir).absolute()))
self.recursive_cb.setChecked(cfg.recursive)
self.keep_intermediate_cb.setChecked(cfg.keep_intermediate)
self.overwrite_cb.setChecked(cfg.overwrite)
i = self.mode_cb.findData(cfg.mode)
if i >= 0:
self.mode_cb.setCurrentIndex(i)
self.convert_enabled_cb.setChecked(cfg.convert.enabled)
i = self.convert_sr.findData(cfg.convert.sample_rate)
if i >= 0:
self.convert_sr.setCurrentIndex(i)
i = self.convert_ch.findData(cfg.convert.channels)
if i >= 0:
self.convert_ch.setCurrentIndex(i)
self.convert_ffmpeg.setText(cfg.convert.ffmpeg_path)
self.separate_enabled_cb.setChecked(cfg.separate.enabled)
self.separate_model.setText(cfg.separate.model)
self.separate_stem.setText(cfg.separate.isolate_stem)
self.separate_unblend.setText(cfg.separate.unblend_path)
self.separate_kernels_cb.setChecked(cfg.separate.custom_kernels)
self.transcribe_enabled_cb.setChecked(cfg.transcribe.enabled)
self.engine_cb.setCurrentText(cfg.transcribe.engine)
self.model_edit.setText(cfg.transcribe.model)
self.lang_cb.setCurrentText(cfg.transcribe.language or "auto")
self.device_cb.setCurrentText(cfg.transcribe.device)
self.compute_type_cb.setCurrentText(cfg.transcribe.compute_type or "auto")
self.batch_spin.setValue(cfg.transcribe.batch_size)
self.gigaam_provider_cb.setCurrentText(cfg.transcribe.gigaam_provider)
self.local_files_only_cb.setChecked(cfg.transcribe.local_files_only)
self.vad_filter_cb.setChecked(cfg.transcribe.vad_filter)
def _collect_config(self) -> PipelineConfig:
cfg = PipelineConfig()
cfg.input_dir = self.input_edit.text().strip()
cfg.output_dir = self.output_edit.text().strip() or "transcribed"
cfg.recursive = self.recursive_cb.isChecked()
cfg.keep_intermediate = self.keep_intermediate_cb.isChecked()
cfg.overwrite = self.overwrite_cb.isChecked()
cfg.mode = self.mode_cb.currentData()
cfg.convert = ConvertConfig(
enabled=self.convert_enabled_cb.isChecked(),
sample_rate=int(self.convert_sr.currentData()),
channels=int(self.convert_ch.currentData()),
ffmpeg_path=self.convert_ffmpeg.text().strip() or "ffmpeg",
)
cfg.separate = SeparateConfig(
enabled=self.separate_enabled_cb.isChecked(),
model=self.separate_model.text().strip(),
isolate_stem=self.separate_stem.text().strip() or "vocals",
unblend_path=self.separate_unblend.text().strip() or "unblend",
custom_kernels=self.separate_kernels_cb.isChecked(),
)
lang = self.lang_cb.currentText().strip()
compute = self.compute_type_cb.currentText().strip()
cfg.transcribe = TranscribeConfig(
enabled=self.transcribe_enabled_cb.isChecked(),
engine=self.engine_cb.currentText(),
model=self.model_edit.text().strip() or "large-v3",
language=None if lang in ("", "auto") else lang,
device=self.device_cb.currentText(),
compute_type=None if compute == "auto" else compute,
batch_size=self.batch_spin.value(),
gigaam_provider=self.gigaam_provider_cb.currentText(),
)
cfg.transcribe.local_files_only = self.local_files_only_cb.isChecked()
cfg.transcribe.vad_filter = self.vad_filter_cb.isChecked()
return cfg
# ---------------- handlers ----------------
def _pick_input_dir(self) -> None:
d = QFileDialog.getExistingDirectory(self, "Выберите входную папку")
if d:
self.input_edit.setText(d)
def _pick_output_dir(self) -> None:
d = QFileDialog.getExistingDirectory(self, "Выберите выходную папку")
if d:
self.output_edit.setText(d)
def _open_output_dir(self) -> None:
p = Path(self.output_edit.text().strip() or "transcribed").absolute()
p.mkdir(parents=True, exist_ok=True)
QDesktopServices.openUrl(QUrl.fromLocalFile(str(p)))
def _save_preset(self) -> None:
path, _ = QFileDialog.getSaveFileName(self, "Сохранить пресет", "preset.json", "JSON (*.json)")
if not path:
return
try:
self._collect_config().save(Path(path))
except Exception as exc: # noqa: BLE001
QMessageBox.critical(self, "Ошибка", str(exc))
def _load_preset(self) -> None:
path, _ = QFileDialog.getOpenFileName(self, "Загрузить пресет", "", "JSON (*.json)")
if not path:
return
try:
self._apply_config(PipelineConfig.load(Path(path)))
except Exception as exc: # noqa: BLE001
QMessageBox.critical(self, "Ошибка", str(exc))
def _on_start(self) -> None:
cfg = self._collect_config()
if not cfg.input_dir or not Path(cfg.input_dir).is_dir():
QMessageBox.warning(self, "Нет входной папки", "Выберите существующую входную папку.")
return
if not (cfg.convert.enabled or cfg.separate.enabled or cfg.transcribe.enabled):
QMessageBox.warning(self, "Нет этапов", "Включите хотя бы один этап.")
return
if cfg.convert.enabled and not shutil.which(cfg.convert.ffmpeg_path):
QMessageBox.warning(self, "ffmpeg не найден",
f"Команда '{cfg.convert.ffmpeg_path}' не найдена в PATH.")
return
if cfg.separate.enabled and not shutil.which(cfg.separate.unblend_path):
QMessageBox.warning(self, "unblend не найден",
f"Команда '{cfg.separate.unblend_path}' не найдена в PATH.")
return
self.log_view.clear()
self._log_device_info()
if cfg.transcribe.enabled and cfg.transcribe.language is None:
self._append_log("⚠ Язык = auto: WhisperX определит язык на каждом файле. "
"Если язык известен — задайте его явно, будет быстрее.")
self.progress.setValue(0)
self.progress.setFormat("Запуск…")
self.start_btn.setEnabled(False)
self.stop_btn.setEnabled(True)
self._worker = PipelineWorker(cfg)
self._worker.log.connect(self._append_log)
self._worker.progress.connect(self._on_progress)
self._worker.finished_ok.connect(self._on_finished)
self._worker.failed.connect(self._on_failed)
self._worker.start()
def _on_stop(self) -> None:
if self._worker is not None:
self._worker.stop()
self._append_log("⏹ Запрошена остановка…")
def _append_log(self, text: str) -> None:
self.log_view.appendPlainText(text)
sb = self.log_view.verticalScrollBar()
sb.setValue(sb.maximum())
def _on_progress(self, done: int, total: int, stage: str) -> None:
self.progress.setMaximum(total)
self.progress.setValue(done)
label = STAGE_LABELS.get(stage, stage)
self.progress.setFormat(f"{label} — %v / %m")
def _on_finished(self, results: list) -> None:
self.start_btn.setEnabled(True)
self.stop_btn.setEnabled(False)
ok = sum(1 for r in results if r.error is None)
fail = len(results) - ok
self.progress.setFormat(f"Готово — успешно {ok}, с ошибкой {fail}")
self._append_log(f"\n✅ Готово. Успешно: {ok}, с ошибкой: {fail}.")
def _on_failed(self, message: str) -> None:
self.start_btn.setEnabled(True)
self.stop_btn.setEnabled(False)
self.progress.setFormat("Ошибка")
self._append_log(f"\n✗ Ошибка пайплайна: {message}")
QMessageBox.critical(self, "Ошибка", message)