refactor: split tts.py into per-backend packages

1 files changed, 68 insertions, 0 deletions

diff --git a/app/converter/clients/__init__.py b/app/converter/clients/__init__.py
new file mode 100644
index 0000000..e216011
--- /dev/null
+++ b/app/converter/clients/__init__.py
@@ -0,0 +1,68 @@
+"""TTS client implementations — one module per backend server.
+
+Public API: the three client classes (QwenTTSClient, FasterTTSClient,
+AudioCppTTSClient), the backend/voice-mode vocabulary, and the shared
+helpers (normalize_language, speaker tables, whisper transcription) that
+the UIs and setup wizards build on.
+"""
+
+# The TTS backends a conversion can use, in Convert-form order. Each has a
+# client module here; the backends package mirrors these keys for its
+# install/setup wizards.
+BACKEND_QWEN = "qwen"
+BACKEND_FASTER = "faster"
+BACKEND_AUDIOCPP = "audiocpp"
+BACKENDS = (BACKEND_AUDIOCPP, BACKEND_QWEN, BACKEND_FASTER)
+
+from .base import BaseTTSClient, ConversionCancelled, VOICE_MODE_CLONE, \
+    VOICE_MODE_CUSTOM, VOICE_MODES, resolve_request_seed
+from .languages import LANGUAGE_ISO_CODES, TTS_LANGUAGES, \
+    TTS_LANGUAGE_ALIASES, normalize_language
+from .speakers import QWEN3_TTS_SPEAKERS, SPEAKER_DISPLAY_NAMES, \
+    is_builtin_speaker, speaker_display_name, speaker_display_name_for
+from .transcribe import transcribe_reference_audio, whisper_backend_available
+from .qwen import CUSTOM_VOICE_MODEL_ID, MODEL_SIZE, QwenTTSClient
+from .faster import SAMPLE_RATE, FasterTTSClient
+from .audiocpp import (
+    AUDIOCPP_DEFAULT_FAMILY_PROFILE,
+    AUDIOCPP_FAMILY_PROFILES,
+    AUDIOCPP_FAMILY_QWEN3_TTS,
+    AUDIOCPP_LANG_DISPLAY,
+    AUDIOCPP_LANG_ISO,
+    AUDIOCPP_LANG_OMIT,
+    AUDIOCPP_SYNTHESIS_TASKS,
+    AUDIOCPP_TASK_TTS,
+    AUDIOCPP_TASK_VDES,
+    AUDIOCPP_VOICE_CLONE,
+    AUDIOCPP_VOICE_DESIGN,
+    AUDIOCPP_VOICE_SPEAKER,
+    AudioCppFamilyProfile,
+    AudioCppTTSClient,
+    audiocpp_entry_voice_capability,
+)
+
+__all__ = [
+    # vocabulary
+    "BACKEND_QWEN", "BACKEND_FASTER", "BACKEND_AUDIOCPP", "BACKENDS",
+    "VOICE_MODE_CUSTOM", "VOICE_MODE_CLONE", "VOICE_MODES",
+    # clients
+    "BaseTTSClient", "ConversionCancelled", "resolve_request_seed",
+    "QwenTTSClient", "FasterTTSClient", "AudioCppTTSClient",
+    # model facts
+    "MODEL_SIZE", "CUSTOM_VOICE_MODEL_ID", "SAMPLE_RATE",
+    # languages
+    "TTS_LANGUAGES", "TTS_LANGUAGE_ALIASES", "LANGUAGE_ISO_CODES",
+    "normalize_language",
+    # speakers
+    "QWEN3_TTS_SPEAKERS", "SPEAKER_DISPLAY_NAMES",
+    "speaker_display_name", "speaker_display_name_for", "is_builtin_speaker",
+    # transcription
+    "transcribe_reference_audio", "whisper_backend_available",
+    # audio.cpp family profiles
+    "AUDIOCPP_LANG_DISPLAY", "AUDIOCPP_LANG_ISO", "AUDIOCPP_LANG_OMIT",
+    "AUDIOCPP_FAMILY_QWEN3_TTS", "AUDIOCPP_TASK_TTS", "AUDIOCPP_TASK_VDES",
+    "AUDIOCPP_SYNTHESIS_TASKS", "AUDIOCPP_VOICE_SPEAKER",
+    "AUDIOCPP_VOICE_CLONE", "AUDIOCPP_VOICE_DESIGN",
+    "AudioCppFamilyProfile", "AUDIOCPP_DEFAULT_FAMILY_PROFILE",
+    "AUDIOCPP_FAMILY_PROFILES", "audiocpp_entry_voice_capability",
+]