More work on UI for TTS embed

src/calibre/ebooks/oeb/polish/tts.py

@@ -7,6 +7,7 @@ import os
 import sys
 from collections import defaultdict
 from contextlib import suppress
+from functools import partial
 from typing import NamedTuple
 
 from lxml.etree import ElementBase as Element
@@ -380,14 +381,15 @@ class ReportProgress:
     def __init__(self):
         self.current_stage = ''
 
-    def __call__(self, stage: str, item: str, count: int, total: int) -> None:
+    def __call__(self, stage: str, item: str, count: int, total: int) -> bool:
         if stage != self.current_stage:
             self.current_stage = stage
             print()
             print(self.current_stage)
-            return
+            return False
         frac = count / total
         print(f'\r{frac:4.0%} {item}', end='')
+        return False
 
 
 def make_par(container, seq, html_href, audio_href, elem_id, pos, duration) -> None:
@@ -436,12 +438,13 @@ def remove_embedded_tts(container):
         container.remove_item(aname)
 
 
-def embed_tts(container, report_progress=None, parent_widget=None):
+def embed_tts(container, report_progress=None, callback_to_download_voices=None):
     report_progress = report_progress or ReportProgress()
     if container.book_type != 'epub':
         raise UnsupportedContainerType(_('Only the EPUB format has support for embedding speech overlay audio'))
     if container.opf_version_parsed[0] < 3:
-        report_progress(_('Updating book internals'), '', 0, 0)
+        if report_progress(_('Updating book internals'), '', 0, 0):
+            return False
         upgrade_book(container, print)
     remove_embedded_tts(container)
 
@@ -455,7 +458,8 @@ def embed_tts(container, report_progress=None, parent_widget=None):
         if container.mime_map.get(name) in OEB_DOCS:
             name_map[name] = PerFileData(name)
     stage = _('Processing HTML')
-    report_progress(stage, '', 0, len(name_map))
+    if report_progress(stage, '', 0, len(name_map)):
+        return False
     all_voices = set()
     total_num_sentences = 0
     for i, (name, pfd) in enumerate(name_map.items()):
@@ -467,10 +471,16 @@ def embed_tts(container, report_progress=None, parent_widget=None):
             pfd.key_map[key].append(s)
             all_voices.add(key)
         container.dirty(name)
-        report_progress(stage, name, i+1, len(name_map))
-    piper.ensure_voices_downloaded(iter(all_voices), parent=parent_widget)
+        if report_progress(stage, name, i+1, len(name_map)):
+            return False
+    if callback_to_download_voices is None:
+        piper.ensure_voices_downloaded(iter(all_voices))
+    else:
+        if not callback_to_download_voices(partial(piper.ensure_voices_downloaded, iter(all_voices))):
+            return False
     stage = _('Converting text to speech')
-    report_progress(stage, '', 0, total_num_sentences)
+    if report_progress(stage, '', 0, total_num_sentences):
+        return False
     snum = 0
     size_of_audio_data = 0
     mmap = {container.href_to_name(item.get('href'), container.opf_name):item for item in container.manifest_items}
@@ -484,7 +494,8 @@ def embed_tts(container, report_progress=None, parent_widget=None):
                 audio_map[s] = audio_data, duration
                 size_of_audio_data += len(audio_data)
                 snum += 1
-                report_progress(stage, _('Sentence number: {}').format(snum), snum, total_num_sentences)
+                if report_progress(stage, _('Sentence number: {}').format(snum), snum, total_num_sentences):
+                    return False
         wav = io.BytesIO()
         wav.write(wav_header_for_pcm_data(size_of_audio_data, HIGH_QUALITY_SAMPLE_RATE))
         afitem = container.generate_item(name + '.m4a', id_prefix='tts-')

src/calibre/gui2/tts/piper.py

@@ -628,10 +628,13 @@ class PiperEmbedded:
                 raw_data = resample_raw_audio_16bit(raw_data, self._current_audio_rate, sample_rate)
             yield raw_data, duration_of_raw_audio_data(raw_data, sample_rate)
 
-    def ensure_voices_downloaded(self, specs: Iterable[tuple[str, str]], parent: QObject = None) -> None:
+    def ensure_voices_downloaded(self, specs: Iterable[tuple[str, str]], parent: QObject = None) -> bool:
         for lang, voice_name in specs:
             voice = self.resolve_voice(lang, voice_name)
-            download_voice(voice, parent=parent, headless=parent is None)
+            m, c = download_voice(voice, parent=parent, headless=parent is None)
+            if not m:
+                return False
+        return True
 
     def shutdown(self):
         if self._process is not None:

src/calibre/gui2/tweak_book/tts.py

@@ -1,14 +1,18 @@
 #!/usr/bin/env python
 # License: GPLv3 Copyright: 2024, Kovid Goyal <kovid at kovidgoyal.net>
 
+import os
 import sys
+import traceback
 
-from qt.core import QHBoxLayout, QStackedLayout, QTextBrowser, QVBoxLayout, QWidget
+from qt.core import QDialogButtonBox, QHBoxLayout, QIcon, QStackedLayout, Qt, QTextBrowser, QVBoxLayout, QWidget, pyqtSignal
 
+from calibre.gui2 import error_dialog
 from calibre.gui2.tweak_book.widgets import Dialog
+from calibre.gui2.widgets import BusyCursor
 
 
-class ConfigWidget(QWidget):
+class EngineSettingsWidget(QWidget):
 
     def __init__(self, parent=None):
         from calibre.gui2.tts.config import EmbeddingConfig
@@ -18,20 +22,138 @@ class ConfigWidget(QWidget):
         self.conf = c = EmbeddingConfig(self)
         h.addWidget(c)
         self.help = q = QTextBrowser(self)
-        h.addWidget(q)
+        h.addWidget(q, 10)
+        q.setHtml(_('''
+<h2>Add Text-to-speech narration</h2>
+
+<p>Add an audio overlay to this book using Text-to-speech technology. Then users reading this book in a reader that supports
+audio overlays, such as the calibre viewer, will be able to hear the text read to them, if they wish.
+
+<p>You can mark different passages to be spoken by different voices as shown in the example below:
+
+<div><code>&lt;p data-calibre-tts="{0}"&gt;This will be voiced by "{0}"&lt;/p&gt;</code></div>
+<div><code>&lt;p data-calibre-tts="{1}"&gt;This will be voiced by "{1}"&lt;/p&gt;</code></div>
+
+<p style="font-size: small">Note that generating the Text-to-speech audio will be quite slow,
+at the rate of approximately one sentence per couple of seconds, depending on your computer's hardware,
+so consider leave it running overnight.
+''').format('cory', 'ryan'))
+        self.save_settings = c.save_settings
+
+
+class Progress(QWidget):
+
+    cancel_requested: bool = False
+    current_stage: str = ''
+
+    def __init__(self, parent: QWidget = None):
+        super().__init__(parent)
+        self.v = v = QVBoxLayout(self)
+        v.setAlignment(Qt.AlignmentFlag.AlignCenter)
+        v.setContentsMargins(0, 0, 0, 0)
+
+    def __call__(self, stage: str, item: str, count: int, total: int) -> bool:
+        return self.cancel_requested
 
 
 class TTSEmbed(Dialog):
 
+    report_progress = pyqtSignal(object, object)
+    worker_done = pyqtSignal(object)
+    ensure_voices_downloaded_signal = pyqtSignal(object, object)
+
     def __init__(self, container, parent=None):
         self.container = container
+        from threading import Thread
+        self.worker_thread = Thread(target=self.worker, daemon=True)
+        self.worker_done.connect(self.on_worker_done, type=Qt.ConnectionType.QueuedConnection)
+        self.ensure_voices_downloaded_signal.connect(self.do_ensure_voices_downloaded, type=Qt.ConnectionType.QueuedConnection)
         super().__init__(_('Add Text-to-speech narration'), 'tts-overlay-dialog', parent=parent)
 
     def setup_ui(self):
         self.v = v = QVBoxLayout(self)
-        self.stack = s = QStackedLayout(self)
+        self.engine_settings_widget = e = EngineSettingsWidget(self)
+        self.stack = s = QStackedLayout()
+        s.addWidget(e)
+        s.setCurrentIndex(0)
         v.addLayout(s)
+
+        self.progress = p = Progress(self)
+        self.report_progress.connect(self.do_report_progress, type=Qt.ConnectionType.QueuedConnection)
+        s.addWidget(p)
+
+        self.remove_media_button = b = self.bb.addButton(_('&Remove existing audio'), QDialogButtonBox.ButtonRole.ActionRole)
+        b.setToolTip(_('Remove any exisiting audio overlays, such as a previously created Text-to-speech narration from this book'))
+        b.setIcon(QIcon.ic('trash.png'))
+        b.clicked.connect(self.remove_media)
         v.addWidget(self.bb)
+        self.update_button_box()
+        self.stack.currentChanged.connect(self.update_button_box)
+
+    def update_button_box(self):
+        if self.stack.currentIndex() == 0:
+            self.bb.setStandardButtons(QDialogButtonBox.StandardButton.Ok | QDialogButtonBox.StandardButton.Cancel)
+            self.remove_media_button.setVisible(True)
+        else:
+            self.bb.setStandardButtons(QDialogButtonBox.StandardButton.Cancel)
+            self.remove_media_button.setVisible(False)
+
+    def remove_media(self):
+        from calibre.ebooks.oeb.polish.tts import remove_embedded_tts
+        remove_embedded_tts(self.container)
+        super().accept()
+
+    def accept(self):
+        if self.stack.currentIndex() == 0:
+            self.engine_settings_widget.save_settings()
+            self.stack.setCurrentIndex(1)
+            self.worker_thread.start()
+
+    def do_report_progress(self, a, kw):
+        self.progress(*a, **kw)
+
+    def worker(self):
+        from calibre.ebooks.oeb.polish.tts import embed_tts
+        def report_progress(*a, **kw):
+            self.report_progress.emit(a, kw)
+            return self.progress.cancel_requested
+        try:
+            err = embed_tts(self.container, report_progress, self.ensure_voices_downloaded)
+        except Exception as e:
+            err = e
+            err.det_msg = traceback.format_exc()
+        self.worker_done.emit(err)
+
+    def ensure_voices_downloaded(self, callback):
+        from queue import Queue
+        queue = Queue()
+        self.ensure_voices_downloaded_signal.emit(callback, queue)
+        e = queue.get()
+        if isinstance(e, Exception):
+            raise e
+        return e
+
+    def do_ensure_voices_downloaded(self, callback, queue):
+        try:
+            queue.put(callback(self))
+        except Exception as e:
+            e.det_msg = traceback.format_exc()
+            queue.put(e)
+
+    def on_worker_done(self, err_or_ok):
+        if isinstance(err_or_ok, Exception):
+            error_dialog(self, _('Text-to-speech narration failed'), str(err_or_ok), det_msg=getattr(err_or_ok, 'det_msg', ''), show=True)
+            return super().reject()
+        return super().accept() if err_or_ok else super().reject()
+
+    def reject(self):
+        if self.stack.currentIndex() == 0:
+            return super().reject()
+        with BusyCursor():
+            self.progress.cancel_requested = True
+            self.bb.setEnabled(False)
+        return super().reject()
+
 
 
 def develop():
@@ -44,6 +166,11 @@ def develop():
     d.exec()
     del d
     del app
+    b, e = os.path.splitext(path)
+    outpath = b + '-tts' + e
+    container.commit(outpath)
+    print('Output saved to:', outpath)
+
 
 
 if __name__ == '__main__':