~

app.py

@@ -2666,335 +2666,6 @@ def _pair_audio_tracks_and_gain(
         if audio_track
     ]
 
-@cache
-def _get_voice_converter() -> VoiceConverter:
-    """
-    Get a voice converter.
-
-    Returns
-    -------
-    VoiceConverter
-        A voice converter.
-
-    """
-    from ultimate_rvc.rvc.infer.infer import VoiceConverter  # noqa: PLC0415
-
-    return VoiceConverter()
-
-
-def convert(
-    audio_track: StrPath,
-    directory: StrPath,
-    model_name: str,
-    n_octaves: int = 0,
-    n_semitones: int = 0,
-    f0_methods: Sequence[F0Method] | None = None,
-    index_rate: float = 0.3,
-    rms_mix_rate: float = 1.0,
-    protect_rate: float = 0.33,
-    hop_length: int = 128,
-    split_audio: bool = False,
-    autotune_audio: bool = False,
-    autotune_strength: float = 1.0,
-    clean_audio: bool = False,
-    clean_strength: float = 0.7,
-    embedder_model: EmbedderModel = EmbedderModel.CONTENTVEC,
-    custom_embedder_model: str | None = None,
-    sid: int = 0,
-    content_type: RVCContentType = RVCContentType.AUDIO,
-    make_directory: bool = False,
-) -> Path:
-    """
-    Convert an audio track using an RVC model.
-
-    Parameters
-    ----------
-    audio_track : StrPath
-        The path to the audio track to convert.
-    directory : StrPath
-        The path to the directory where the converted audio track
-        will be saved.
-    model_name : str
-        The name of the model to use for voice conversion.
-    n_octaves : int, default=0
-        The number of octaves to pitch-shift the converted audio by.
-    n_semitones : int, default=0
-        The number of semitones to pitch-shift the converted audio by.
-    f0_methods : Sequence[F0Method], optional
-        The methods to use for pitch extraction. If None, the method
-        used is rmvpe.
-    index_rate : float, default=0.3
-        The influence of the index file on the voice conversion.
-    rms_mix_rate : float, default = 1.0
-        The blending rate of the volume envelope of the converted
-        audio.
-    protect_rate : float, default=0.33
-        The protection rate for consonants and breathing sounds.
-    hop_length : int, default=128
-        The hop length to use for CREPE-based pitch extraction.
-    split_audio : bool, default=False
-        Whether to split the audio track into smaller segments before
-        converting it.
-    autotune_audio : bool, default=False
-        Whether to apply autotune to the converted audio.
-    autotune_strength : float, default=1.0
-        The strength of the autotune to apply to the converted audio.
-    clean_audio : bool, default=False
-        Whether to clean the converted audio.
-    clean_strength : float, default=0.7
-        The intensity of the cleaning to apply to the converted audio.
-    embedder_model : EmbedderModel, default=EmbedderModel.CONTENTVEC
-        The model to use for generating speaker embeddings.
-    custom_embedder_model : str, optional
-        The name of a custom embedder model to use for generating
-        speaker embeddings.
-    sid : int, default=0
-        The speaker id to use for multi-speaker models.
-    content_type : RVCContentType, default=RVCContentType.AUDIO
-        The type of content to convert. Determines what is shown in
-        display mesages and saved file names.
-    make_directory : bool, default=False
-        Whether to create the directory where the converted audio
-        track will be saved if it does not exist.
-
-    Returns
-    -------
-    Path
-        The path to the converted audio track.
-
-    """
-    match content_type:
-        case RVCContentType.VOCALS:
-            track_entity = Entity.VOCALS_TRACK
-            directory_entity = Entity.SONG_DIR
-        case RVCContentType.VOICE:
-            track_entity = Entity.VOICE_TRACK
-            directory_entity = Entity.DIRECTORY
-        case RVCContentType.SPEECH:
-            track_entity = Entity.SPEECH_TRACK
-            directory_entity = Entity.DIRECTORY
-        case RVCContentType.AUDIO:
-            track_entity = Entity.AUDIO_TRACK
-            directory_entity = Entity.DIRECTORY
-    audio_path = validate_audio_file_exists(audio_track, track_entity)
-    if make_directory:
-        Path(directory).mkdir(parents=True, exist_ok=True)
-    directory_path = validate_audio_dir_exists(directory, directory_entity)
-    validate_model(model_name, Entity.VOICE_MODEL)
-    custom_embedder_model_path = None
-    if embedder_model == EmbedderModel.CUSTOM:
-        custom_embedder_model_path = validate_model(
-            custom_embedder_model,
-            Entity.CUSTOM_EMBEDDER_MODEL,
-        )
-
-    audio_path = wavify(
-        audio_path,
-        directory_path,
-        "20_Input",
-        accepted_formats={AudioExt.M4A, AudioExt.AAC},
-    )
-
-    n_semitones = n_octaves * 12 + n_semitones
-    f0_methods_set = set(f0_methods) if f0_methods else {F0Method.RMVPE}
-
-    args_dict = RVCAudioMetaData(
-        audio_track=FileMetaData(
-            name=audio_path.name,
-            hash_id=get_file_hash(audio_path),
-        ),
-        model_name=model_name,
-        n_semitones=n_semitones,
-        f0_methods=sorted(f0_methods_set),
-        index_rate=index_rate,
-        rms_mix_rate=rms_mix_rate,
-        protect_rate=protect_rate,
-        hop_length=hop_length,
-        split_audio=split_audio,
-        autotune_audio=autotune_audio,
-        autotune_strength=autotune_strength,
-        clean_audio=clean_audio,
-        clean_strength=clean_strength,
-        embedder_model=embedder_model,
-        custom_embedder_model=custom_embedder_model,
-        sid=sid,
-    ).model_dump()
-
-    paths = [
-        get_unique_base_path(
-            directory_path,
-            f"21_{content_type.capitalize()}_Converted",
-            args_dict,
-        ).with_suffix(suffix)
-        for suffix in [".wav", ".json"]
-    ]
-
-    converted_audio_path, converted_audio_json_path = paths
-
-    if not all(path.exists() for path in paths):
-        rvc_model_path, rvc_index_path = _get_rvc_files(model_name)
-
-        voice_converter = _get_voice_converter()
-
-        voice_converter.convert_audio(
-            audio_input_path=str(audio_path),
-            audio_output_path=str(converted_audio_path),
-            model_path=str(rvc_model_path),
-            index_path=str(rvc_index_path) if rvc_index_path else "",
-            pitch=n_semitones,
-            f0_methods=f0_methods_set,
-            index_rate=index_rate,
-            volume_envelope=rms_mix_rate,
-            protect=protect_rate,
-            hop_length=hop_length,
-            split_audio=split_audio,
-            f0_autotune=autotune_audio,
-            f0_autotune_strength=autotune_strength,
-            embedder_model=embedder_model,
-            embedder_model_custom=(
-                str(custom_embedder_model_path)
-                if custom_embedder_model_path is not None
-                else None
-            ),
-            clean_audio=clean_audio,
-            clean_strength=clean_strength,
-            post_process=False,
-            resample_sr=0,
-            sid=sid,
-        )
-        json_dump(args_dict, converted_audio_json_path)
-    return converted_audio_path
-
-
-@cache
-def _get_audio_separator(
-    output_dir: StrPath = INTERMEDIATE_AUDIO_BASE_DIR,
-    output_format: str = AudioExt.WAV,
-    segment_size: int = SegmentSize.SEG_256,
-    sample_rate: int = 44100,
-) -> Separator:
-    static_ffmpeg.add_paths(weak=True)
-    from audio_separator.separator import Separator  # noqa: PLC0415
-
-    """
-    Get an audio separator.
-
-    Parameters
-    ----------
-    output_dir : StrPath, default=INTERMEDIATE_AUDIO_BASE_DIR
-        The directory to save the separated audio to.
-    output_format : str, default=AudioExt.WAV
-        The format to save the separated audio in.
-    segment_size : int, default=SegmentSize.SEG_256
-        The segment size to use for separation.
-    sample_rate : int, default=44100
-        The sample rate to use for separation.
-
-    Returns
-    -------
-    Separator
-        An audio separator.
-
-    """
-    return Separator(
-        model_file_dir=SEPARATOR_MODELS_DIR,
-        use_autocast=False,
-        output_dir=output_dir,
-        output_format=output_format,
-        sample_rate=sample_rate,
-        mdx_params={
-            "hop_length": 1024,
-            "segment_size": segment_size,
-            "overlap": 0.25,
-            "batch_size": 1,
-            "enable_denoise": True,
-        },
-    )
-
-
-
-@spaces.GPU(duration=zgpuduration)
-def separate_audio(
-    audio_track: StrPath,
-    song_dir: StrPath,
-    model_name: SeparationModel,
-    segment_size: int,
-) -> tuple[Path, Path]:
-    """
-    Separate an audio track into a primary stem and a secondary stem.
-
-    Parameters
-    ----------
-    audio_track : StrPath
-        The path to the audio track to separate.
-    song_dir : StrPath
-        The path to the song directory where the separated primary stem
-        and secondary stem will be saved.
-    model_name : str
-        The name of the model to use for audio separation.
-    segment_size : int
-        The segment size to use for audio separation.
-
-    Returns
-    -------
-    primary_path : Path
-        The path to the separated primary stem.
-    secondary_path : Path
-        The path to the separated secondary stem.
-
-    """
-    audio_path = validate_audio_file_exists(audio_track, Entity.AUDIO_TRACK)
-    song_dir_path = validate_audio_dir_exists(song_dir, Entity.SONG_DIR)
-
-    args_dict = SeparatedAudioMetaData(
-        audio_track=FileMetaData(
-            name=audio_path.name,
-            hash_id=get_file_hash(audio_path),
-        ),
-        model_name=model_name,
-        segment_size=segment_size,
-    ).model_dump()
-
-    paths = [
-        get_unique_base_path(
-            song_dir_path,
-            prefix,
-            args_dict,
-        ).with_suffix(suffix)
-        for prefix in ["11_Stem_Primary", "11_Stem_Secondary"]
-        for suffix in [".wav", ".json"]
-    ]
-
-    (
-        primary_path,
-        primary_json_path,
-        secondary_path,
-        secondary_json_path,
-    ) = paths
-
-    if not all(path.exists() for path in paths):
-        audio_separator = _get_audio_separator(
-            output_dir=song_dir_path,
-            segment_size=segment_size,
-        )
-        audio_separator.load_model(model_name)
-        audio_separator.separate(
-            str(audio_path),
-            custom_output_names={
-                audio_separator.model_instance.primary_stem_name: str(
-                    primary_path.with_suffix(""),
-                ),
-                audio_separator.model_instance.secondary_stem_name: str(
-                    secondary_path.with_suffix(""),
-                ),
-            },
-        )
-        json_dump(args_dict, primary_json_path)
-        json_dump(args_dict, secondary_json_path)
-
-    return primary_path, secondary_path
-
-
 def run_pipeline2(
     source: str,
     model_name: str,