tabs/inference/inference.py

import os, sys
import gradio as gr
import regex as re
import shutil
import datetime
import random

from core import (
    run_infer_script,
    run_batch_infer_script,
)

from assets.i18n.i18n import I18nAuto

from rvc.lib.utils import format_title

i18n = I18nAuto()

now_dir = os.getcwd()
sys.path.append(now_dir)

model_root = os.path.join(now_dir, "logs")
audio_root = os.path.join(now_dir, "assets", "audios")

model_root_relative = os.path.relpath(model_root, now_dir)
audio_root_relative = os.path.relpath(audio_root, now_dir)

sup_audioext = {
    "wav",
    "mp3",
    "flac",
    "ogg",
    "opus",
    "m4a",
    "mp4",
    "aac",
    "alac",
    "wma",
    "aiff",
    "webm",
    "ac3",
}

names = [
    os.path.join(root, file)
    for root, _, files in os.walk(model_root_relative, topdown=False)
    for file in files
    if (
        file.endswith((".pth", ".onnx"))
        and not (file.startswith("G_") or file.startswith("D_"))
    )
]

indexes_list = [
    os.path.join(root, name)
    for root, _, files in os.walk(model_root_relative, topdown=False)
    for name in files
    if name.endswith(".index") and "trained" not in name
]

audio_paths = [
    os.path.join(root, name)
    for root, _, files in os.walk(audio_root_relative, topdown=False)
    for name in files
    if name.endswith(tuple(sup_audioext))
    and root == audio_root_relative
    and "_output" not in name
]


def output_path_fn(input_audio_path):
    original_name_without_extension = os.path.basename(input_audio_path).rsplit(".", 1)[
        0
    ]
    new_name = original_name_without_extension + "_output.wav"
    output_path = os.path.join(os.path.dirname(input_audio_path), new_name)
    return output_path


def change_choices():
    names = [
        os.path.join(root, file)
        for root, _, files in os.walk(model_root_relative, topdown=False)
        for file in files
        if (
            file.endswith((".pth", ".onnx"))
            and not (file.startswith("G_") or file.startswith("D_"))
        )
    ]

    indexes_list = [
        os.path.join(root, name)
        for root, _, files in os.walk(model_root_relative, topdown=False)
        for name in files
        if name.endswith(".index") and "trained" not in name
    ]

    audio_paths = [
        os.path.join(root, name)
        for root, _, files in os.walk(audio_root_relative, topdown=False)
        for name in files
        if name.endswith(tuple(sup_audioext))
        and root == audio_root_relative
        and "_output" not in name
    ]

    return (
        {"choices": sorted(names), "__type__": "update"},
        {"choices": sorted(indexes_list), "__type__": "update"},
        {"choices": sorted(audio_paths), "__type__": "update"},
    )


def get_indexes():
    indexes_list = [
        os.path.join(dirpath, filename)
        for dirpath, _, filenames in os.walk(model_root_relative)
        for filename in filenames
        if filename.endswith(".index") and "trained" not in filename
    ]

    return indexes_list if indexes_list else ""


def match_index(model_file: str) -> tuple:
    model_files_trip = re.sub(r"\.pth|\.onnx$", "", model_file)
    model_file_name = os.path.split(model_files_trip)[
        -1
    ]  # Extract only the name, not the directory

    # Check if the sid0strip has the specific ending format _eXXX_sXXX
    if re.match(r".+_e\d+_s\d+$", model_file_name):
        base_model_name = model_file_name.rsplit("_", 2)[0]
    else:
        base_model_name = model_file_name

    sid_directory = os.path.join(model_root_relative, base_model_name)
    double_sid_directory = os.path.join(sid_directory, base_model_name)
    directories_to_search = [sid_directory] if os.path.exists(sid_directory) else []
    directories_to_search += (
        [double_sid_directory] if os.path.exists(double_sid_directory) else []
    )
    directories_to_search.append(model_root_relative)
    matching_index_files = []

    for directory in directories_to_search:
        for filename in os.listdir(directory):
            if filename.endswith(".index") and "trained" not in filename:
                # Condition to match the name
                name_match = any(
                    name.lower() in filename.lower()
                    for name in [model_file_name, base_model_name]
                )

                # If in the specific directory, it's automatically a match
                folder_match = directory == sid_directory

                if name_match or folder_match:
                    index_path = os.path.join(directory, filename)
                    if index_path in indexes_list:
                        matching_index_files.append(
                            (
                                index_path,
                                os.path.getsize(index_path),
                                " " not in filename,
                            )
                        )
    if matching_index_files:
        # Sort by favoring files without spaces and by size (largest size first)
        matching_index_files.sort(key=lambda x: (-x[2], -x[1]))
        best_match_index_path = matching_index_files[0][0]
        return best_match_index_path

    return ""


def save_to_wav(record_button):
    if record_button is None:
        pass
    else:
        path_to_file = record_button
        new_name = datetime.datetime.now().strftime("%Y-%m-%d_%H-%M-%S") + ".wav"
        target_path = os.path.join(audio_root_relative, os.path.basename(new_name))

        shutil.move(path_to_file, target_path)
        return target_path, output_path_fn(target_path)


def save_to_wav2(upload_audio):
    file_path = upload_audio
    formated_name = format_title(os.path.basename(file_path))
    target_path = os.path.join(audio_root_relative, formated_name)

    if os.path.exists(target_path):
        os.remove(target_path)

    shutil.copy(file_path, target_path)
    return target_path, output_path_fn(target_path)


def delete_outputs():
    for root, _, files in os.walk(audio_root_relative, topdown=False):
        for name in files:
            if name.endswith(tuple(sup_audioext)) and name.__contains__("_output"):
                os.remove(os.path.join(root, name))
    gr.Info(f"Outputs cleared!")


# Inference tab
def inference_tab():
    default_weight = random.choice(names) if names else None
    with gr.Row():
        with gr.Row():
            model_file = gr.Dropdown(
                label=i18n("Voice Model"),
                choices=sorted(names, key=lambda path: os.path.getsize(path)),
                interactive=True,
                value=default_weight,
                allow_custom_value=True,
            )

            index_file = gr.Dropdown(
                label=i18n("Index File"),
                choices=get_indexes(),
                value=match_index(default_weight) if default_weight else "",
                interactive=True,
                allow_custom_value=True,
            )
        with gr.Column():
            refresh_button = gr.Button(i18n("Refresh"))
            unload_button = gr.Button(i18n("Unload Voice"))

            unload_button.click(
                fn=lambda: ({"value": "", "__type__": "update"}),
                inputs=[],
                outputs=[model_file],
            )

            model_file.select(
                fn=match_index,
                inputs=[model_file],
                outputs=[index_file],
            )

    # Single inference tab
    with gr.Tab(i18n("Single")):
        with gr.Row():
            with gr.Column():
                upload_audio = gr.Audio(
                    label=i18n("Upload Audio"), type="filepath", editable=False
                )
                with gr.Row():
                    audio = gr.Dropdown(
                        label=i18n("Select Audio"),
                        choices=sorted(audio_paths),
                        value=audio_paths[0] if audio_paths else "",
                        interactive=True,
                        allow_custom_value=True,
                    )

        with gr.Accordion(i18n("Advanced Settings"), open=False):
            with gr.Column():
                clear_outputs = gr.Button(
                    i18n("Clear Outputs (Deletes all audios in assets/audios)")
                )
                output_path = gr.Textbox(
                    label=i18n("Output Path"),
                    placeholder=i18n("Enter output path"),
                    value=output_path_fn(audio_paths[0])
                    if audio_paths
                    else os.path.join(now_dir, "assets", "audios", "output.wav"),
                    interactive=True,
                )
                split_audio = gr.Checkbox(
                    label=i18n("Split Audio"),
                    visible=True,
                    value=False,
                    interactive=True,
                )
                pitch = gr.Slider(
                    minimum=-24,
                    maximum=24,
                    step=1,
                    label=i18n("Pitch"),
                    value=0,
                    interactive=True,
                )
                filter_radius = gr.Slider(
                    minimum=0,
                    maximum=7,
                    label=i18n(
                        "If >=3: apply median filtering to the harvested pitch results. The value represents the filter radius and can reduce breathiness"
                    ),
                    value=3,
                    step=1,
                    interactive=True,
                )
                index_rate = gr.Slider(
                    minimum=0,
                    maximum=1,
                    label=i18n("Search Feature Ratio"),
                    value=0.75,
                    interactive=True,
                )
                hop_length = gr.Slider(
                    minimum=1,
                    maximum=512,
                    step=1,
                    label=i18n("Hop Length"),
                    value=128,
                    interactive=True,
                )
            with gr.Column():
                f0method = gr.Radio(
                    label=i18n("Pitch extraction algorithm"),
                    choices=[
                        "pm",
                        "harvest",
                        "dio",
                        "crepe",
                        "crepe-tiny",
                        "rmvpe",
                    ],
                    value="rmvpe",
                    interactive=True,
                )

        convert_button1 = gr.Button(i18n("Convert"))

        with gr.Row():  # Defines output info + output audio download after conversion
            vc_output1 = gr.Textbox(label=i18n("Output Information"))
            vc_output2 = gr.Audio(label=i18n("Export Audio"))

    # Batch inference tab
    with gr.Tab(i18n("Batch")):
        with gr.Row():
            with gr.Column():
                input_folder_batch = gr.Textbox(
                    label=i18n("Input Folder"),
                    placeholder=i18n("Enter input path"),
                    value=os.path.join(now_dir, "assets", "audios"),
                    interactive=True,
                )
                output_folder_batch = gr.Textbox(
                    label=i18n("Output Folder"),
                    placeholder=i18n("Enter output path"),
                    value=os.path.join(now_dir, "assets", "audios"),
                    interactive=True,
                )
        with gr.Accordion(i18n("Advanced Settings"), open=False):
            with gr.Column():
                clear_outputs = gr.Button(
                    i18n("Clear Outputs (Deletes all audios in assets/audios)")
                )
                pitch_batch = gr.Slider(
                    minimum=-24,
                    maximum=24,
                    step=1,
                    label=i18n("Pitch"),
                    value=0,
                    interactive=True,
                )
                filter_radius_batch = gr.Slider(
                    minimum=0,
                    maximum=7,
                    label=i18n(
                        "If >=3: apply median filtering to the harvested pitch results. The value represents the filter radius and can reduce breathiness"
                    ),
                    value=3,
                    step=1,
                    interactive=True,
                )
                index_rate_batch = gr.Slider(
                    minimum=0,
                    maximum=1,
                    label=i18n("Search Feature Ratio"),
                    value=0.75,
                    interactive=True,
                )
                hop_length_batch = gr.Slider(
                    minimum=1,
                    maximum=512,
                    step=1,
                    label=i18n("Hop Length"),
                    value=128,
                    interactive=True,
                )
            with gr.Column():
                f0method_batch = gr.Radio(
                    label=i18n("Pitch extraction algorithm"),
                    choices=[
                        "pm",
                        "harvest",
                        "dio",
                        "crepe",
                        "crepe-tiny",
                        "rmvpe",
                    ],
                    value="rmvpe",
                    interactive=True,
                )

        convert_button2 = gr.Button(i18n("Convert"))

        with gr.Row():  # Defines output info + output audio download after conversion
            vc_output3 = gr.Textbox(label=i18n("Output Information"))

    def toggle_visible(checkbox):
        return {"visible": checkbox, "__type__": "update"}

    refresh_button.click(
        fn=change_choices,
        inputs=[],
        outputs=[model_file, index_file, audio],
    )
    audio.change(
        fn=output_path_fn,
        inputs=[audio],
        outputs=[output_path],
    )
    upload_audio.upload(
        fn=save_to_wav2,
        inputs=[upload_audio],
        outputs=[audio, output_path],
    )
    upload_audio.stop_recording(
        fn=save_to_wav,
        inputs=[upload_audio],
        outputs=[audio, output_path],
    )
    clear_outputs.click(
        fn=delete_outputs,
        inputs=[],
        outputs=[],
    )
    convert_button1.click(
        fn=run_infer_script,
        inputs=[
            pitch,
            filter_radius,
            index_rate,
            hop_length,
            f0method,
            audio,
            output_path,
            model_file,
            index_file,
            split_audio,
        ],
        outputs=[vc_output1, vc_output2],
    )
    convert_button2.click(
        fn=run_batch_infer_script,
        inputs=[
            pitch_batch,
            filter_radius_batch,
            index_rate_batch,
            hop_length_batch,
            f0method_batch,
            input_folder_batch,
            output_folder_batch,
            model_file,
            index_file,
        ],
        outputs=[vc_output3],
    )