ttsforge 0.1.1__py3-none-any.whl → 0.1.2__py3-none-any.whl

ttsforge/__init__.py

@@ -18,7 +18,7 @@ from pykokoro.tokenizer import (
     Tokenizer,
 )
 from pykokoro.constants import SUPPORTED_LANGUAGES
-
+from pykokoro.onnx_backend import VOICE_NAMES_BY_VARIANT
 from .constants import (
     DEFAULT_CONFIG,
     LANGUAGE_DESCRIPTIONS,
@@ -27,23 +27,7 @@ from .constants import (
 )
 
 # Import from pykokoro
-try:
-    from pykokoro.constants import SAMPLE_RATE
-    from pykokoro.onnx_backend import LANG_CODE_TO_ONNX
-except ImportError:
-    # Fallback values if pykokoro not installed
-    SAMPLE_RATE = 24000
-    LANG_CODE_TO_ONNX = {
-        "a": "en-us",
-        "b": "en-gb",
-        "e": "es",
-        "f": "fr-fr",
-        "h": "hi",
-        "i": "it",
-        "j": "ja",
-        "p": "pt",
-        "z": "zh",
-    }
+from pykokoro.constants import SAMPLE_RATE
 
 from .conversion import (
     Chapter,
@@ -73,6 +57,7 @@ __all__ = [
     "LANGUAGE_DESCRIPTIONS",
     "SUPPORTED_OUTPUT_FORMATS",
     "VOICES",
+    "VOICE_NAMES_BY_VARIANT",
     # Conversion
     "Chapter",
     "ConversionOptions",

ttsforge/_version.py

@@ -28,7 +28,7 @@ version_tuple: VERSION_TUPLE
 commit_id: COMMIT_ID
 __commit_id__: COMMIT_ID
 
-__version__ = version = '0.1.1'
-__version_tuple__ = version_tuple = (0, 1, 1)
+__version__ = version = '0.1.2'
+__version_tuple__ = version_tuple = (0, 1, 2)
 
 __commit_id__ = commit_id = None

ttsforge/cli/commands_conversion.py

@@ -17,6 +17,7 @@ from typing import Literal, TypedDict, cast
 
 import click
 import numpy as np
+from pykokoro.onnx_backend import DEFAULT_MODEL_QUALITY, ModelQuality
 from rich.panel import Panel
 from rich.progress import (
     BarColumn,
@@ -37,7 +38,6 @@ from ..constants import (
     LANGUAGE_DESCRIPTIONS,
     SUPPORTED_OUTPUT_FORMATS,
     VOICE_PREFIX_TO_LANG,
-    VOICES,
 )
 from ..conversion import (
     Chapter,
@@ -54,6 +54,7 @@ from ..utils import (
     load_config,
     resolve_conversion_defaults,
 )
+from .commands_utility import _resolve_model_source_and_variant, _resolve_voice_names
 from .helpers import DEFAULT_SAMPLE_TEXT, console, parse_voice_parameter
 
 
@@ -64,6 +65,14 @@ class ContentItem(TypedDict):
     page_number: NotRequired[int]
 
 
+def get_voices() -> list[str]:
+    """Get the list of available voices."""
+    cfg = load_config()
+
+    model_source, model_variant = _resolve_model_source_and_variant(cfg)
+    return _resolve_voice_names(model_source, model_variant)
+
+
 @click.command()
 @click.argument("epub_file", type=click.Path(exists=True, path_type=Path))
 @click.option(
@@ -82,7 +91,7 @@ class ContentItem(TypedDict):
 @click.option(
     "-v",
     "--voice",
-    type=click.Choice(VOICES),
+    type=click.Choice(get_voices()),
     help="Voice to use for TTS.",
 )
 @click.option(
@@ -150,6 +159,12 @@ class ContentItem(TypedDict):
     default=None,
     help="Pause mode: 'tts', 'manual', or 'auto' (default: auto).",
 )
+@click.option(
+    "--enable-short-sentence/--disable-short-sentence",
+    "enable_short_sentence",
+    default=None,
+    help="Enable/disable special handling for short sentences.",
+)
 @click.option(
     "--announce-chapters/--no-announce-chapters",
     "announce_chapters",
@@ -296,6 +311,7 @@ def convert(  # noqa: C901
     pause_paragraph: float | None,
     pause_variance: float | None,
     pause_mode: str | None,
+    enable_short_sentence: bool | None,
     announce_chapters: bool | None,
     chapter_pause: float | None,
     title: str | None,
@@ -325,6 +341,10 @@ def convert(  # noqa: C901
     config = load_config()
     model_path = ctx.obj.get("model_path") if ctx.obj else None
     voices_path = ctx.obj.get("voices_path") if ctx.obj else None
+    model_source, model_variant = _resolve_model_source_and_variant(config)
+    model_quality = cast(
+        ModelQuality, config.get("model_quality", DEFAULT_MODEL_QUALITY)
+    )
 
     # Get format first (needed for output path construction)
     fmt = output_format or config.get("default_format", "m4b")
@@ -467,6 +487,9 @@ def convert(  # noqa: C901
         language=language or "a",
         speed=speed or config.get("default_speed", 1.0),
         use_gpu=use_gpu if use_gpu is not None else config.get("use_gpu", False),
+        model_source=model_source,
+        model_variant=model_variant,
+        model_quality=model_quality,
         num_chapters=len(selected_indices) if selected_indices else len(epub_chapters),
         title=effective_title,
         author=effective_author,
@@ -510,6 +533,9 @@ def convert(  # noqa: C901
         output_format=output_format or config.get("default_format", "m4b"),
         output_dir=output.parent,
         use_gpu=use_gpu if use_gpu is not None else config.get("use_gpu", False),
+        model_quality=model_quality,
+        model_source=model_source,
+        model_variant=model_variant,
         silence_between_chapters=silence or config.get("silence_between_chapters", 2.0),
         lang=lang or config.get("phonemization_lang"),
         use_mixed_language=(
@@ -556,6 +582,11 @@ def convert(  # noqa: C901
         pause_mode=(
             pause_mode if pause_mode is not None else config.get("pause_mode", "auto")
         ),
+        enable_short_sentence=(
+            enable_short_sentence
+            if enable_short_sentence is not None
+            else config.get("enable_short_sentence", None)
+        ),
         announce_chapters=(
             announce_chapters
             if announce_chapters is not None
@@ -947,6 +978,10 @@ def sample(
 
     # Load config for defaults
     user_config = load_config()
+    model_source, model_variant = _resolve_model_source_and_variant(user_config)
+    model_quality = cast(
+        ModelQuality, user_config.get("model_quality", DEFAULT_MODEL_QUALITY)
+    )
     resolved_defaults = resolve_conversion_defaults(
         user_config,
         {
@@ -980,6 +1015,9 @@ def sample(
         use_gpu=resolved_defaults["use_gpu"],
         split_mode=resolved_defaults["split_mode"],
         lang=resolved_defaults["lang"],
+        model_quality=model_quality,
+        model_source=model_source,
+        model_variant=model_variant,
         use_mixed_language=(
             use_mixed_language or user_config.get("use_mixed_language", False)
         ),
@@ -1117,6 +1155,9 @@ def _show_conversion_summary(
     language: str,
     speed: float,
     use_gpu: bool,
+    model_source: str,
+    model_variant: str,
+    model_quality: str | None,
     num_chapters: int,
     title: str,
     author: str,
@@ -1139,6 +1180,9 @@ def _show_conversion_summary(
     table.add_row("Chapters", str(num_chapters))
     table.add_row("Voice", voice)
     table.add_row("Language", LANGUAGE_DESCRIPTIONS.get(language, language))
+    table.add_row("Model Source", model_source)
+    table.add_row("Model Variant", model_variant)
+    table.add_row("Model Quality", str(model_quality))
     if lang:
         table.add_row("Phonemization Lang", f"{lang} (override)")
     if use_mixed_language:
@@ -1167,7 +1211,7 @@ def _show_conversion_summary(
 @click.option(
     "-v",
     "--voice",
-    type=click.Choice(VOICES),
+    type=click.Choice(get_voices()),
     help="TTS voice to use.",
 )
 @click.option(
@@ -1271,6 +1315,11 @@ def _show_conversion_summary(
     default=None,
     help="Trim leading/trailing silence from audio.",
 )
+@click.option(
+    "--enable-short-sentence/--disable-short-sentence",
+    default=None,
+    help="Enable special handling for short sentences.",
+)
 @click.pass_context
 def read(  # noqa: C901
     ctx: click.Context,
@@ -1293,6 +1342,7 @@ def read(  # noqa: C901
     pause_paragraph: float | None,
     pause_variance: float | None,
     pause_mode: str | None,
+    enable_short_sentence: bool | None,
 ) -> None:
     """Read an EPUB or text file aloud with streaming playback.
 
@@ -1340,6 +1390,10 @@ def read(  # noqa: C901
 
     # Load config for defaults
     config = load_config()
+    model_source, model_variant = _resolve_model_source_and_variant(config)
+    model_quality = cast(
+        ModelQuality, config.get("model_quality", DEFAULT_MODEL_QUALITY)
+    )
     resolved_defaults = resolve_conversion_defaults(
         config,
         {
@@ -1389,6 +1443,11 @@ def read(  # noqa: C901
     effective_pause_mode = (
         pause_mode if pause_mode is not None else config.get("pause_mode", "auto")
     )
+    effective_enable_short_sentence = (
+        enable_short_sentence
+        if enable_short_sentence is not None
+        else config.get("enable_short_sentence", None)
+    )
 
     # Get language code for TTS
     espeak_lang = LANG_CODE_TO_ONNX.get(effective_language, "en-us")
@@ -1645,11 +1704,15 @@ def read(  # noqa: C901
             model_path=model_path,
             voices_path=voices_path,
             use_gpu=effective_use_gpu,
+            model_quality=model_quality,
+            model_source=model_source,
+            model_variant=model_variant,
         )
         generation = GenerationConfig(
             speed=effective_speed,
             lang=espeak_lang,
             pause_mode=cast(Literal["tts", "manual", "auto"], effective_pause_mode),
+            enable_short_sentence=effective_enable_short_sentence,
             pause_clause=effective_pause_clause,
             pause_sentence=effective_pause_sentence,
             pause_paragraph=effective_pause_paragraph,
@@ -1658,6 +1721,9 @@ def read(  # noqa: C901
         pipeline_config = PipelineConfig(
             voice=effective_voice,
             generation=generation,
+            model_quality=model_quality,
+            model_source=model_source,
+            model_variant=model_variant,
             model_path=model_path,
             voices_path=voices_path,
         )

ttsforge/cli/commands_phonemes.py

@@ -10,9 +10,10 @@ This module contains commands for working with phonemes and pre-tokenized conten
 import re
 import sys
 from pathlib import Path
-from typing import Any
+from typing import Any, cast
 
 import click
+from pykokoro.onnx_backend import DEFAULT_MODEL_QUALITY, ModelQuality
 from rich.progress import (
     BarColumn,
     Progress,
@@ -37,6 +38,7 @@ from ..utils import (
     format_filename_template,
     load_config,
 )
+from .commands_utility import _resolve_model_source_and_variant
 from .helpers import console, parse_voice_parameter
 
 
@@ -500,6 +502,10 @@ def phonemes_convert(
     config = load_config()
     model_path = ctx.obj.get("model_path") if ctx.obj else None
     voices_path = ctx.obj.get("voices_path") if ctx.obj else None
+    model_source, model_variant = _resolve_model_source_and_variant(config)
+    model_quality = cast(
+        ModelQuality, config.get("model_quality", DEFAULT_MODEL_QUALITY)
+    )
 
     # Get book info and metadata
     book_info = book.get_info()
@@ -599,6 +605,9 @@ def phonemes_convert(
         speed=speed,
         output_format=fmt,
         use_gpu=gpu,
+        model_quality=model_quality,
+        model_source=model_source,
+        model_variant=model_variant,
         silence_between_chapters=silence,
         pause_clause=(
             pause_clause
@@ -834,6 +843,12 @@ def phonemes_preview(
             # Auto-detect if voice is a blend
             parsed_voice, parsed_voice_blend = parse_voice_parameter(voice)
 
+            config = load_config()
+            model_source, model_variant = _resolve_model_source_and_variant(config)
+            model_quality = cast(
+                ModelQuality, config.get("model_quality", DEFAULT_MODEL_QUALITY)
+            )
+
             # Initialize converter
             options = ConversionOptions(
                 phoneme_dictionary_path=str(phoneme_dict) if phoneme_dict else None,
@@ -841,6 +856,9 @@ def phonemes_preview(
                 voice_blend=parsed_voice_blend,
                 language=language,
                 output_format="wav",  # Explicitly set WAV format
+                model_quality=model_quality,
+                model_source=model_source,
+                model_variant=model_variant,
             )
             converter = TTSConverter(options)
 

ttsforge/cli/commands_utility.py

@@ -555,6 +555,14 @@ def _resolve_model_source_and_variant(cfg: dict) -> tuple[ModelSource, ModelVari
     return cast(ModelSource, source), cast(ModelVariant, variant)
 
 
+def _resolve_voice_names(
+    model_source: ModelSource = "huggingface",
+    model_variant: ModelVariant = "v1.0",
+) -> list[str]:
+    """Return the list of voice names for the given model variant."""
+    return VOICE_NAMES_BY_VARIANT.get(model_variant, VOICE_NAMES)
+
+
 def _get_cache_voices_path(
     model_source: ModelSource,
     model_variant: ModelVariant,
@@ -708,7 +716,7 @@ def download(ctx: click.Context, force: bool, quality: str | None) -> None:
 
         # ---- voices
         if model_source == "huggingface":
-            voice_names = VOICE_NAMES_BY_VARIANT.get(model_variant, VOICE_NAMES)
+            voice_names = _resolve_voice_names(model_source, model_variant)
             total_voices = len(voice_names)
             voices_task = progress.add_task(
                 f"Downloading voices (0/{total_voices})...", total=total_voices
@@ -1269,6 +1277,12 @@ def list_names(  # noqa: C901
         )
         console.print("[dim]Type 'q' to quit, 's' to skip, 'r' to replay.[/dim]\n")
 
+        cfg = load_config()
+        model_source, model_variant = _resolve_model_source_and_variant(cfg)
+        model_quality = cast(
+            ModelQuality, cfg.get("model_quality", DEFAULT_MODEL_QUALITY)
+        )
+
         # Initialize converter with phoneme dictionary
         try:
             # Auto-detect if voice is a blend
@@ -1279,6 +1293,9 @@ def list_names(  # noqa: C901
                 voice=parsed_voice or "af_sky",
                 voice_blend=parsed_voice_blend,
                 language=language,
+                model_quality=model_quality,
+                model_source=model_source,
+                model_variant=model_variant,
             )
             converter = TTSConverter(options)
 

ttsforge/cli/helpers.py

@@ -50,6 +50,7 @@ DEFAULT_SAMPLE_TEXT = (
 DEMO_TEXT = {
     "a": "Hello! This audio was generated by {voice}. How do you like it?",
     "b": "Hello! This audio was generated by {voice}. How do you like it?",
+    "d": "Hallo! Dieses Audio wurde von {voice} erzeugt. Wie gefallt es Ihnen?",
     "e": "Hola! Este audio fue generado por {voice}. Que te parece?",
     "f": "Bonjour! Cet audio a ete genere par {voice}. Comment le trouvez-vous?",
     "h": "Namaste! Yah audio {voice} dwara banaya gaya hai. Aapko kaisa laga?",

ttsforge/constants.py

@@ -3,10 +3,10 @@
 # from pykokoro.onnx_backend import VOICE_NAMES_V1_0
 # from pykokoro.onnx_backend import VOICE_NAMES_V1_1_ZH, VOICE_NAMES_V1_1_DE
 
-from pykokoro.onnx_backend import VOICE_NAMES_V1_0 as VOICE_NAMES
+from pykokoro.onnx_backend import DEFAULT_MODEL_SOURCE, VOICE_NAMES_V1_0
 
 # Re-export from pykokoro for convenience
-VOICES = VOICE_NAMES
+VOICES = VOICE_NAMES_V1_0
 
 # Audio constants from pykokoro
 try:
@@ -24,6 +24,7 @@ PROGRAM_DESCRIPTION = "Generate audiobooks from EPUB files using Kokoro ONNX TTS
 LANGUAGE_DESCRIPTIONS = {
     "a": "American English",
     "b": "British English",
+    "d": "German",
     "e": "Spanish",
     "f": "French",
     "h": "Hindi",
@@ -35,6 +36,8 @@ LANGUAGE_DESCRIPTIONS = {
 
 # ISO language code to ttsforge language code mapping
 ISO_TO_LANG_CODE = {
+    "de": "d",
+    "de-de": "d",
     "en": "a",  # Default to American English
     "en-us": "a",
     "en-gb": "b",
@@ -62,6 +65,8 @@ VOICE_PREFIX_TO_LANG = {
     "am": "a",  # American Male
     "bf": "b",  # British Female
     "bm": "b",  # British Male
+    "df": "d",  # German Female
+    "dm": "d",  # German Male
     "ef": "e",  # Spanish Female
     "em": "e",  # Spanish Male
     "ff": "f",  # French Female
@@ -82,6 +87,7 @@ VOICE_PREFIX_TO_LANG = {
 DEFAULT_VOICE_FOR_LANG = {
     "a": "af_heart",
     "b": "bf_emma",
+    "d": "df_eva",
     "e": "ef_dora",
     "f": "ff_siwis",
     "h": "hf_alpha",
@@ -115,6 +121,7 @@ DEFAULT_CONFIG = {
     "use_gpu": False,  # GPU requires onnxruntime-gpu
     # Model quality: fp32, fp16, q8, q8f16, q4, q4f16, uint8, uint8f16
     "model_quality": "fp32",
+    "model_source": DEFAULT_MODEL_SOURCE,
     "model_variant": "v1.0",
     "silence_between_chapters": 2.0,
     "save_chapters_separately": False,
@@ -128,6 +135,7 @@ DEFAULT_CONFIG = {
     "pause_paragraph": 0.9,
     "pause_variance": 0.05,
     "pause_mode": "auto",  # "tts", "manual", or "auto
+    "enable_short_sentence": None,
     # Language override for phonemization (e.g., 'de', 'fr', 'en-us')
     # If None, language is determined from voice prefix
     "phonemization_lang": None,
@@ -154,6 +162,7 @@ AUDIO_CHANNELS = 1
 SAMPLE_TEXTS = {
     "a": "This is a sample of the selected voice.",
     "b": "This is a sample of the selected voice.",
+    "d": "Dies ist ein Beispiel für die ausgewählte Stimme.",
     "e": "Este es una muestra de la voz seleccionada.",
     "f": "Ceci est un exemple de la voix sélectionnée.",
     "h": "यह चयनित आवाज़ का एक नमूना है।",  # noqa: E501

ttsforge/conversion.py

@@ -11,6 +11,14 @@ from pathlib import Path
 from typing import Any, Literal, Optional, cast
 
 import soundfile as sf
+from pykokoro.onnx_backend import (
+    DEFAULT_MODEL_QUALITY,
+    DEFAULT_MODEL_SOURCE,
+    DEFAULT_MODEL_VARIANT,
+    ModelQuality,
+    ModelSource,
+    ModelVariant,
+)
 
 from .audio_merge import AudioMerger, MergeMeta
 from .constants import (
@@ -123,12 +131,16 @@ class ConversionState:
     speed: float = 1.0
     split_mode: str = "auto"
     output_format: str = "m4b"
+    model_quality: ModelQuality | None = DEFAULT_MODEL_QUALITY
+    model_source: ModelSource = DEFAULT_MODEL_SOURCE
+    model_variant: ModelVariant = DEFAULT_MODEL_VARIANT
     silence_between_chapters: float = 2.0
     pause_clause: float = 0.3
     pause_sentence: float = 0.5
     pause_paragraph: float = 0.9
     pause_variance: float = 0.05
     pause_mode: str = "auto"  # "tts", "manual", or "auto
+    enable_short_sentence: bool | None = None
     lang: str | None = None  # Language override for phonemization
     chapters: list[ChapterState] = field(default_factory=list)
     started_at: str = ""
@@ -183,8 +195,16 @@ class ConversionState:
                 data["pause_variance"] = 0.05
             if "pause_mode" not in data:
                 data["pause_mode"] = "auto"
+            if "enable_short_sentence" not in data:
+                data["enable_short_sentence"] = None
             if "lang" not in data:
                 data["lang"] = None
+            if "model_quality" not in data:
+                data["model_quality"] = DEFAULT_MODEL_QUALITY
+            if "model_source" not in data:
+                data["model_source"] = DEFAULT_MODEL_SOURCE
+            if "model_variant" not in data:
+                data["model_variant"] = DEFAULT_MODEL_VARIANT
 
             return cls(**data)
         except (json.JSONDecodeError, TypeError, KeyError):
@@ -204,12 +224,16 @@ class ConversionState:
             "speed": self.speed,
             "split_mode": self.split_mode,
             "output_format": self.output_format,
+            "model_quality": self.model_quality,
+            "model_source": self.model_source,
+            "model_variant": self.model_variant,
             "silence_between_chapters": self.silence_between_chapters,
             "pause_clause": self.pause_clause,
             "pause_sentence": self.pause_sentence,
             "pause_paragraph": self.pause_paragraph,
             "pause_variance": self.pause_variance,
             "pause_mode": self.pause_mode,
+            "enable_short_sentence": self.enable_short_sentence,
             "lang": self.lang,
             "chapters": [
                 {
@@ -294,6 +318,7 @@ class ConversionOptions:
     pause_paragraph: float = 0.9  # For paragraph boundaries
     pause_variance: float = 0.05  # Standard deviation for natural variation
     pause_mode: str = "auto"  # "tts", "manual", or "auto
+    enable_short_sentence: bool | None = None  # Enable short sentence handling
     # Chapter announcement settings
     announce_chapters: bool = True  # Read chapter titles aloud before content
     chapter_pause_after_title: float = 2.0  # Pause after chapter title (seconds)
@@ -315,6 +340,9 @@ class ConversionOptions:
     # Filename template for chapter files
     chapter_filename_template: str = "{chapter_num:03d}_{book_title}_{chapter_title}"
     # Custom ONNX model path (None = use default downloaded model)
+    model_quality: ModelQuality | None = DEFAULT_MODEL_QUALITY
+    model_source: ModelSource = DEFAULT_MODEL_SOURCE
+    model_variant: ModelVariant = DEFAULT_MODEL_VARIANT
     model_path: Path | None = None
     # Custom voices.bin path (None = use default downloaded voices)
     voices_path: Path | None = None
@@ -420,6 +448,9 @@ class TTSConverter:
             pause_sentence=self.options.pause_sentence,
             pause_paragraph=self.options.pause_paragraph,
             pause_variance=self.options.pause_variance,
+            model_quality=self.options.model_quality,
+            model_source=self.options.model_source,
+            model_variant=self.options.model_variant,
             model_path=self.options.model_path,
             voices_path=self.options.voices_path,
             voice_blend=self.options.voice_blend,
@@ -600,53 +631,79 @@ class TTSConverter:
                         )
                         state = None
                     else:
-                        # Check if settings differ from saved state
-                        settings_changed = (
-                            state.voice != self.options.voice
-                            or state.language != self.options.language
-                            or state.speed != self.options.speed
-                            or state.split_mode != self.options.split_mode
-                            or state.silence_between_chapters
-                            != self.options.silence_between_chapters
-                            or state.pause_clause != self.options.pause_clause
-                            or state.pause_sentence != self.options.pause_sentence
-                            or state.pause_paragraph != self.options.pause_paragraph
-                            or state.pause_variance != self.options.pause_variance
-                            or state.pause_mode != self.options.pause_mode
-                            or state.lang != self.options.lang
+                        model_settings_changed = (
+                            state.model_quality != self.options.model_quality
+                            or state.model_source != self.options.model_source
+                            or state.model_variant != self.options.model_variant
                         )
 
-                        if settings_changed:
+                        if model_settings_changed:
                             self.log(
-                                f"Restoring settings from previous session: "
-                                f"voice={state.voice}, language={state.language}, "
-                                f"lang_override={state.lang}, "
-                                f"speed={state.speed}, "
-                                f"split_mode={state.split_mode}, "
-                                f"silence={state.silence_between_chapters}s, "
-                                f"pauses: clause={state.pause_clause}s "
-                                f"sent={state.pause_sentence}s "
-                                f"para={state.pause_paragraph}s "
-                                f"var={state.pause_variance}s "
-                                f"pause_mode={state.pause_mode}",
-                                "info",
+                                "Model settings changed, starting fresh conversion",
+                                "warning",
+                            )
+                            state = None
+                        else:
+                            # Check if settings differ from saved state
+                            settings_changed = (
+                                state.voice != self.options.voice
+                                or state.language != self.options.language
+                                or state.speed != self.options.speed
+                                or state.split_mode != self.options.split_mode
+                                or state.silence_between_chapters
+                                != self.options.silence_between_chapters
+                                or state.pause_clause != self.options.pause_clause
+                                or state.pause_sentence != self.options.pause_sentence
+                                or state.pause_paragraph != self.options.pause_paragraph
+                                or state.pause_variance != self.options.pause_variance
+                                or state.pause_mode != self.options.pause_mode
+                                or state.enable_short_sentence
+                                != self.options.enable_short_sentence
+                                or state.lang != self.options.lang
                             )
 
-                        # Apply saved settings to options for consistency
-                        self.options.voice = state.voice
-                        self.options.language = state.language
-                        self.options.speed = state.speed
-                        self.options.split_mode = state.split_mode
-                        self.options.output_format = state.output_format
-                        self.options.silence_between_chapters = (
-                            state.silence_between_chapters
-                        )
-                        self.options.pause_clause = state.pause_clause
-                        self.options.pause_sentence = state.pause_sentence
-                        self.options.pause_paragraph = state.pause_paragraph
-                        self.options.pause_variance = state.pause_variance
-                        self.options.pause_mode = state.pause_mode
-                        self.options.lang = state.lang
+                            if settings_changed:
+                                self.log(
+                                    f"Restoring settings from previous session: "
+                                    f"voice={state.voice}, language={state.language}, "
+                                    f"lang_override={state.lang}, "
+                                    f"speed={state.speed}, "
+                                    f"split_mode={state.split_mode}, "
+                                    f"silence={state.silence_between_chapters}s, "
+                                    f"pauses: clause={state.pause_clause}s "
+                                    f"sent={state.pause_sentence}s "
+                                    f"para={state.pause_paragraph}s "
+                                    f"var={state.pause_variance}s "
+                                    f"pause_mode={state.pause_mode}, "
+                                    f"enable_short_sentence="
+                                    f"{state.enable_short_sentence}, "
+                                    f"model_source={state.model_source}, "
+                                    f"model_variant={state.model_variant}, "
+                                    f"model_quality={state.model_quality}",
+                                    "info",
+                                )
+
+                            # Apply saved settings to options for consistency
+                            self.options.voice = state.voice
+                            self.options.language = state.language
+                            self.options.speed = state.speed
+                            self.options.split_mode = state.split_mode
+                            self.options.output_format = state.output_format
+                            self.options.silence_between_chapters = (
+                                state.silence_between_chapters
+                            )
+                            self.options.pause_clause = state.pause_clause
+                            self.options.pause_sentence = state.pause_sentence
+                            self.options.pause_paragraph = state.pause_paragraph
+                            self.options.pause_variance = state.pause_variance
+                            self.options.pause_mode = state.pause_mode
+                            self.options.enable_short_sentence = (
+                                state.enable_short_sentence
+                            )
+                            self.options.lang = state.lang
+                            self.options.model_quality = state.model_quality
+                            self.options.model_source = state.model_source
+                            self.options.model_variant = state.model_variant
 
             if state is None:
                 # Create new state
@@ -661,12 +718,16 @@ class TTSConverter:
                     speed=self.options.speed,
                     split_mode=self.options.split_mode,
                     output_format=self.options.output_format,
+                    model_quality=self.options.model_quality,
+                    model_source=self.options.model_source,
+                    model_variant=self.options.model_variant,
                     silence_between_chapters=self.options.silence_between_chapters,
                     pause_clause=self.options.pause_clause,
                     pause_sentence=self.options.pause_sentence,
                     pause_paragraph=self.options.pause_paragraph,
                     pause_variance=self.options.pause_variance,
                     pause_mode=self.options.pause_mode,
+                    enable_short_sentence=self.options.enable_short_sentence,
                     lang=self.options.lang,
                     chapters=[
                         ChapterState(

ttsforge/kokoro_runner.py

@@ -7,11 +7,19 @@ from typing import Any, Literal, Protocol, cast
 import numpy as np
 from pykokoro import GenerationConfig, KokoroPipeline, PipelineConfig
 from pykokoro.onnx_backend import (
+    DEFAULT_MODEL_QUALITY,
+    DEFAULT_MODEL_SOURCE,
+    DEFAULT_MODEL_VARIANT,
     Kokoro,
+    ModelQuality,
+    ModelSource,
+    ModelVariant,
     VoiceBlend,
     are_models_downloaded,
     download_all_models,
+    download_all_models_github,
 )
+from pykokoro.pipeline import build_pipeline
 from pykokoro.stages.audio_generation.onnx import OnnxAudioGenerationAdapter
 from pykokoro.stages.audio_postprocessing.onnx import OnnxAudioPostprocessingAdapter
 from pykokoro.stages.phoneme_processing.onnx import OnnxPhonemeProcessorAdapter
@@ -26,6 +34,9 @@ class KokoroRunOptions:
     pause_sentence: float
     pause_paragraph: float
     pause_variance: float
+    model_quality: ModelQuality | None = DEFAULT_MODEL_QUALITY
+    model_source: ModelSource = DEFAULT_MODEL_SOURCE
+    model_variant: ModelVariant = DEFAULT_MODEL_VARIANT
     model_path: Any | None = None
     voices_path: Any | None = None
     voice_blend: str | None = None
@@ -48,15 +59,32 @@ class KokoroRunner:
         if self._pipeline is not None:
             return
 
-        if not are_models_downloaded():
-            self.log("Downloading ONNX model files...")
-            download_all_models()
+        if self.opts.model_path is None or self.opts.voices_path is None:
+            model_quality = self.opts.model_quality or DEFAULT_MODEL_QUALITY
+            model_source = self.opts.model_source or DEFAULT_MODEL_SOURCE
+            if model_source == "github":
+                if not are_models_downloaded(quality=model_quality):
+                    self.log("Downloading ONNX model files from GitHub...")
+                download_all_models_github(
+                    variant=self.opts.model_variant,
+                    quality=model_quality,
+                )
+            else:
+                if not are_models_downloaded(quality=model_quality):
+                    self.log("Downloading ONNX model files...")
+                download_all_models(
+                    variant=self.opts.model_variant,
+                    quality=model_quality,
+                )
 
         self._kokoro = Kokoro(
             model_path=self.opts.model_path,
             voices_path=self.opts.voices_path,
             use_gpu=self.opts.use_gpu,
             tokenizer_config=self.opts.tokenizer_config,
+            model_quality=self.opts.model_quality,
+            model_source=self.opts.model_source,
+            model_variant=self.opts.model_variant,
         )
 
         assert self._kokoro is not None
@@ -88,14 +116,18 @@ class KokoroRunner:
         pipeline_cfg = PipelineConfig(
             voice=self._voice_style,
             generation=GenerationConfig(speed=self.opts.speed, lang="en-us"),
+            model_quality=self.opts.model_quality,
+            model_source=self.opts.model_source,
+            model_variant=self.opts.model_variant,
             model_path=self.opts.model_path,
             voices_path=self.opts.voices_path,
             tokenizer_config=self.opts.tokenizer_config,
         )
 
         # Use the same adapters everywhere (text + phonemes)
-        self._pipeline = KokoroPipeline(
-            pipeline_cfg,
+        self._pipeline = build_pipeline(
+            config=pipeline_cfg,
+            backend=self._kokoro,
             phoneme_processing=OnnxPhonemeProcessorAdapter(self._kokoro),
             audio_generation=OnnxAudioGenerationAdapter(self._kokoro),
             audio_postprocessing=OnnxAudioPostprocessingAdapter(self._kokoro),
@@ -116,6 +148,7 @@ class KokoroRunner:
             lang=lang_code,
             is_phonemes=is_phonemes,
             pause_mode=pause_mode,
+            enable_short_sentence=self.opts.enable_short_sentence,
             pause_clause=self.opts.pause_clause,
             pause_sentence=self.opts.pause_sentence,
             pause_paragraph=self.opts.pause_paragraph,

ttsforge/phoneme_conversion.py

@@ -15,6 +15,14 @@ from typing import Any, Literal, Optional, cast
 
 import numpy as np
 import soundfile as sf
+from pykokoro.onnx_backend import (
+    DEFAULT_MODEL_QUALITY,
+    DEFAULT_MODEL_SOURCE,
+    DEFAULT_MODEL_VARIANT,
+    ModelQuality,
+    ModelSource,
+    ModelVariant,
+)
 
 from .audio_merge import AudioMerger, MergeMeta
 from .chapter_selection import parse_chapter_selection
@@ -94,12 +102,16 @@ class PhonemeConversionState:
     voice: str = ""
     speed: float = 1.0
     output_format: str = "m4b"
+    model_quality: ModelQuality | None = DEFAULT_MODEL_QUALITY
+    model_source: ModelSource = DEFAULT_MODEL_SOURCE
+    model_variant: ModelVariant = DEFAULT_MODEL_VARIANT
     silence_between_chapters: float = 2.0
     pause_clause: float = 0.3
     pause_sentence: float = 0.5
     pause_paragraph: float = 0.9
     pause_variance: float = 0.05
     pause_mode: str = "auto"
+    enable_short_sentence: bool | None = None
     lang: str | None = None  # Language override for phonemization
     chapters: list[PhonemeChapterState] = field(default_factory=list)
     started_at: str = ""
@@ -154,8 +166,16 @@ class PhonemeConversionState:
                 data["pause_variance"] = 0.05
             if "pause_mode" not in data:
                 data["pause_mode"] = "auto"
+            if "enable_short_sentence" not in data:
+                data["enable_short_sentence"] = None
             if "lang" not in data:
                 data["lang"] = None
+            if "model_quality" not in data:
+                data["model_quality"] = DEFAULT_MODEL_QUALITY
+            if "model_source" not in data:
+                data["model_source"] = DEFAULT_MODEL_SOURCE
+            if "model_variant" not in data:
+                data["model_variant"] = DEFAULT_MODEL_VARIANT
 
             return cls(**data)
         except (json.JSONDecodeError, TypeError, KeyError):
@@ -172,12 +192,16 @@ class PhonemeConversionState:
             "voice": self.voice,
             "speed": self.speed,
             "output_format": self.output_format,
+            "model_quality": self.model_quality,
+            "model_source": self.model_source,
+            "model_variant": self.model_variant,
             "silence_between_chapters": self.silence_between_chapters,
             "pause_clause": self.pause_clause,
             "pause_sentence": self.pause_sentence,
             "pause_paragraph": self.pause_paragraph,
             "pause_variance": self.pause_variance,
             "pause_mode": self.pause_mode,
+            "enable_short_sentence": self.enable_short_sentence,
             "lang": self.lang,
             "chapters": [
                 {
@@ -215,6 +239,7 @@ class PhonemeConversionOptions:
     pause_paragraph: float = 0.9  # For paragraph boundaries
     pause_variance: float = 0.05  # Standard deviation for natural variation
     pause_mode: str = "auto"  # "tts", "manual", or "auto"
+    enable_short_sentence: bool | None = None  # Enable short sentence handling
     # Chapter announcement settings
     announce_chapters: bool = True  # Read chapter titles aloud before content
     chapter_pause_after_title: float = 2.0  # Pause after chapter title (seconds)
@@ -235,6 +260,9 @@ class PhonemeConversionOptions:
     # Filename template for chapter files
     chapter_filename_template: str = "{chapter_num:03d}_{book_title}_{chapter_title}"
     # Custom ONNX model path (None = use default downloaded model)
+    model_quality: ModelQuality | None = DEFAULT_MODEL_QUALITY
+    model_source: ModelSource = DEFAULT_MODEL_SOURCE
+    model_variant: ModelVariant = DEFAULT_MODEL_VARIANT
     model_path: Path | None = None
     # Custom voices.bin path (None = use default downloaded voices)
     voices_path: Path | None = None
@@ -583,6 +611,11 @@ class PhonemeConverter:
                         or state.pause_paragraph != self.options.pause_paragraph
                         or state.pause_variance != self.options.pause_variance
                         or state.pause_mode != self.options.pause_mode
+                        or state.enable_short_sentence
+                        != self.options.enable_short_sentence
+                        or state.model_quality != self.options.model_quality
+                        or state.model_source != self.options.model_source
+                        or state.model_variant != self.options.model_variant
                     ):
                         self.log(
                             f"Restoring settings from previous session: "
@@ -592,7 +625,11 @@ class PhonemeConverter:
                             f"pause_sentence={state.pause_sentence}s, "
                             f"pause_paragraph={state.pause_paragraph}s, "
                             f"pause_variance={state.pause_variance}s, "
-                            f"pause_mode={state.pause_mode}",
+                            f"pause_mode={state.pause_mode}, "
+                            f"enable_short_sentence={state.enable_short_sentence}, "
+                            f"model_source={state.model_source}, "
+                            f"model_variant={state.model_variant}, "
+                            f"model_quality={state.model_quality}",
                             "info",
                         )
                         # Apply saved settings for consistency
@@ -607,6 +644,10 @@ class PhonemeConverter:
                         self.options.pause_paragraph = state.pause_paragraph
                         self.options.pause_variance = state.pause_variance
                         self.options.pause_mode = state.pause_mode
+                        self.options.enable_short_sentence = state.enable_short_sentence
+                        self.options.model_quality = state.model_quality
+                        self.options.model_source = state.model_source
+                        self.options.model_variant = state.model_variant
 
             if state is None:
                 # Create new state
@@ -617,12 +658,16 @@ class PhonemeConverter:
                     voice=self.options.voice,
                     speed=self.options.speed,
                     output_format=self.options.output_format,
+                    model_quality=self.options.model_quality,
+                    model_source=self.options.model_source,
+                    model_variant=self.options.model_variant,
                     silence_between_chapters=self.options.silence_between_chapters,
                     pause_clause=self.options.pause_clause,
                     pause_sentence=self.options.pause_sentence,
                     pause_paragraph=self.options.pause_paragraph,
                     pause_variance=self.options.pause_variance,
                     pause_mode=self.options.pause_mode,
+                    enable_short_sentence=self.options.enable_short_sentence,
                     chapters=[
                         PhonemeChapterState(
                             index=idx,
@@ -648,6 +693,9 @@ class PhonemeConverter:
                 pause_sentence=self.options.pause_sentence,
                 pause_paragraph=self.options.pause_paragraph,
                 pause_variance=self.options.pause_variance,
+                model_quality=self.options.model_quality,
+                model_source=self.options.model_source,
+                model_variant=self.options.model_variant,
                 model_path=self.options.model_path,
                 voices_path=self.options.voices_path,
                 voice_blend=self.options.voice_blend,
@@ -848,6 +896,9 @@ class PhonemeConverter:
                 pause_sentence=self.options.pause_sentence,
                 pause_paragraph=self.options.pause_paragraph,
                 pause_variance=self.options.pause_variance,
+                model_quality=self.options.model_quality,
+                model_source=self.options.model_source,
+                model_variant=self.options.model_variant,
                 model_path=self.options.model_path,
                 voices_path=self.options.voices_path,
                 voice_blend=self.options.voice_blend,

{ttsforge-0.1.1.dist-info → ttsforge-0.1.2.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ttsforge
-Version: 0.1.1
+Version: 0.1.2
 Summary: Generate audiobooks from EPUB files using Kokoro ONNX TTS.
 Author-email: Holger Nahrstaedt <nahrstaedt@gmail.com>
 License: MIT License
@@ -617,6 +617,7 @@ ttsforge convert book.epub --gpu
 | `pause_paragraph`           | `0.9`          | Paragraph pause (seconds)            |
 | `pause_variance`            | `0.05`         | Pause variance (seconds)             |
 | `pause_mode`                | `auto`         | Pause mode (`tts`, `manual`, `auto`) |
+| `enable_short_sentence`     | `None`         | Handle short sentences               |
 | `announce_chapters`         | `true`         | Speak chapter titles                 |
 | `chapter_pause_after_title` | `2.0`          | Pause after chapter title            |
 | `phonemization_lang`        | `None`         | Override phonemization language      |

ttsforge-0.1.2.dist-info/RECORD

@@ -0,0 +1,27 @@
+ttsforge/__init__.py,sha256=5mT7eXwuq0Z-Qn_WjYWVjA8VUOUy0lGTyaqQwNdcEOE,2149
+ttsforge/_version.py,sha256=Ok5oAXdWgR9aghaFXTafTeDW6sYO3uVe6d2Nket57R4,704
+ttsforge/audio_merge.py,sha256=Tt7o8GBNrkcfiSKycUpWvblj-y4zwlULoX-eCblqYpo,5666
+ttsforge/audio_player.py,sha256=HYc4vv46yDXjVXaWRlj1tUtWLiwNTwbzT6oDfOUB5vA,14351
+ttsforge/chapter_selection.py,sha256=a-XlEO4HMzeUBfhvnh6gQOQmDuM0wMpVCi0pw6oM2hQ,2579
+ttsforge/constants.py,sha256=c3b9s41mNIb8NHK6C7XKB-xzX4wJhFZpo0auUqCwT2s,5394
+ttsforge/conversion.py,sha256=2ZzeBfg-Kwf7kR9Fwht30Vx9ZcoAEQOVAiFsmfF2IdA,45011
+ttsforge/input_reader.py,sha256=b49SBT-mL4SnR74D8xwyWHC_smPhsJ5jpPAj4QQ5WKo,14068
+ttsforge/kokoro_lang.py,sha256=8603b5whfk0KzGrNK7pqRjzoH1Ge9TKoX7AMzKsX0sk,376
+ttsforge/kokoro_runner.py,sha256=AKvEMaBfCTCLR3KcHoE04nQnUkHDwv62BvQNb1vGC8U,5923
+ttsforge/name_extractor.py,sha256=CxxBadCO0Pcoepcj7gZwkfWPMud2oa9477h_lDYWrIA,9578
+ttsforge/phoneme_conversion.py,sha256=nbDV0adWi--XyRt2RblJbav_ImzDeXrC0xPvoPx8_9c,41093
+ttsforge/phonemes.py,sha256=EUZ1Qr-0rPThRpSeuJQe5Z3J3nz7rX1Xs3Rjjw19qIQ,15517
+ttsforge/ssmd_generator.py,sha256=LknVBSETKH9cY4CoAUBjd1vEfr5VXi0xqEKcft2CR8I,13346
+ttsforge/utils.py,sha256=3BiNFyScV3Dy_xhVm2EigpxUb4Z6YwIQPzzxwDzfCzI,24942
+ttsforge/cli/__init__.py,sha256=CTqYeUAJaKV7YTYqcmr7-VxjwJfjLcnPYM2OKyws0Oc,2103
+ttsforge/cli/commands_conversion.py,sha256=T4hPiU4EXDQ2Wkbd5I5TuHvuAT8rlfEjhfC4mlSMrzg,66007
+ttsforge/cli/commands_phonemes.py,sha256=k3CtXKnUTpbGHj1oPucjIB6syA3LWjxFpg3OIC-tzJ8,33183
+ttsforge/cli/commands_utility.py,sha256=65NSHUFYjRPOWVLnpeBKbR-TbsaDLYAa5xDGcWHS-fk,48630
+ttsforge/cli/helpers.py,sha256=IJt0VpIMPOC-lnBeR3-1keh31MuAdSemDsLh6FpiHLk,2778
+ttsforge/vocab/__init__.py,sha256=lMgS0dY9VbOYI20LnPjjqrWcjLIQ1FKkR4-xcXsvrqc,3641
+ttsforge-0.1.2.dist-info/licenses/LICENSE,sha256=9csb1sDNn0HdUPKgOTUwtb4CkvYPcFXHnkxKCS99EWQ,1074
+ttsforge-0.1.2.dist-info/METADATA,sha256=T0xJ8RtNsBidb4sa_JYR-QnugMK-fLLyVg71R-zi0QU,19655
+ttsforge-0.1.2.dist-info/WHEEL,sha256=wUyA8OaulRlbfwMtmQsvNngGrxQHAvkKcvRmdizlJi0,92
+ttsforge-0.1.2.dist-info/entry_points.txt,sha256=SrcNdlhQpoUCzPzhVbOmMzATQeV7j7XYl0DPrVjZ-ks,47
+ttsforge-0.1.2.dist-info/top_level.txt,sha256=rNLi-3muicHF8UvZu_FuA2ML_Dz9sVPCjik2E8XnCVk,9
+ttsforge-0.1.2.dist-info/RECORD,,

ttsforge-0.1.1.dist-info/RECORD

@@ -1,27 +0,0 @@
-ttsforge/__init__.py,sha256=Jg8_0vPttTVWrnt4HBqrTOKYfmcgpVpfddSVcU4HKXo,2432
-ttsforge/_version.py,sha256=m8HxkqoKGw_wAJtc4ZokpJKNLXqp4zwnNhbnfDtro7w,704
-ttsforge/audio_merge.py,sha256=Tt7o8GBNrkcfiSKycUpWvblj-y4zwlULoX-eCblqYpo,5666
-ttsforge/audio_player.py,sha256=HYc4vv46yDXjVXaWRlj1tUtWLiwNTwbzT6oDfOUB5vA,14351
-ttsforge/chapter_selection.py,sha256=a-XlEO4HMzeUBfhvnh6gQOQmDuM0wMpVCi0pw6oM2hQ,2579
-ttsforge/constants.py,sha256=Y_u8BanHjZBGiAnc9FVPulVHHGIRYlOJhq74BTtxPYA,5108
-ttsforge/conversion.py,sha256=gz6ePEHjC7Nu1zlPArdsIOZA2rb_EEfJMaV6lA-0V2o,41623
-ttsforge/input_reader.py,sha256=b49SBT-mL4SnR74D8xwyWHC_smPhsJ5jpPAj4QQ5WKo,14068
-ttsforge/kokoro_lang.py,sha256=8603b5whfk0KzGrNK7pqRjzoH1Ge9TKoX7AMzKsX0sk,376
-ttsforge/kokoro_runner.py,sha256=ZGBx70_rHcfwKiUgywa_3-7d5u-wQ_0pPOukQRuACu0,4390
-ttsforge/name_extractor.py,sha256=CxxBadCO0Pcoepcj7gZwkfWPMud2oa9477h_lDYWrIA,9578
-ttsforge/phoneme_conversion.py,sha256=nLnFk1uuMMD1Mw4a24zZeZHDHG_4mG3kW1xiOcKM9YA,38274
-ttsforge/phonemes.py,sha256=EUZ1Qr-0rPThRpSeuJQe5Z3J3nz7rX1Xs3Rjjw19qIQ,15517
-ttsforge/ssmd_generator.py,sha256=LknVBSETKH9cY4CoAUBjd1vEfr5VXi0xqEKcft2CR8I,13346
-ttsforge/utils.py,sha256=3BiNFyScV3Dy_xhVm2EigpxUb4Z6YwIQPzzxwDzfCzI,24942
-ttsforge/cli/__init__.py,sha256=CTqYeUAJaKV7YTYqcmr7-VxjwJfjLcnPYM2OKyws0Oc,2103
-ttsforge/cli/commands_conversion.py,sha256=fvbC62gswEb-qpVoU_9S81nFOApH09v9g29nG0HE9X4,63421
-ttsforge/cli/commands_phonemes.py,sha256=Y1_t4jLDuEFVLY6sQC9uwcU9ptt4Y7FrRBRWtkSAF0o,32371
-ttsforge/cli/commands_utility.py,sha256=_8KMUjVYVqp63PH_gjOjS-fw6ZCujaMDXFxoKfUzlko,48013
-ttsforge/cli/helpers.py,sha256=5Co2EvDhYspKhjW2-P3sNxj9MFFgWyTFeqOyJbPy2yA,2697
-ttsforge/vocab/__init__.py,sha256=lMgS0dY9VbOYI20LnPjjqrWcjLIQ1FKkR4-xcXsvrqc,3641
-ttsforge-0.1.1.dist-info/licenses/LICENSE,sha256=9csb1sDNn0HdUPKgOTUwtb4CkvYPcFXHnkxKCS99EWQ,1074
-ttsforge-0.1.1.dist-info/METADATA,sha256=tVUrSm-pBcKWQ0gyrsazqadW0ufARWPXp00qHJsf6ag,19567
-ttsforge-0.1.1.dist-info/WHEEL,sha256=wUyA8OaulRlbfwMtmQsvNngGrxQHAvkKcvRmdizlJi0,92
-ttsforge-0.1.1.dist-info/entry_points.txt,sha256=SrcNdlhQpoUCzPzhVbOmMzATQeV7j7XYl0DPrVjZ-ks,47
-ttsforge-0.1.1.dist-info/top_level.txt,sha256=rNLi-3muicHF8UvZu_FuA2ML_Dz9sVPCjik2E8XnCVk,9
-ttsforge-0.1.1.dist-info/RECORD,,