demo.py

from original import *
import shutil, glob
from easyfuncs import download_from_url, CachedModels
import os

os.makedirs("dataset", exist_ok=True)
model_library = CachedModels()

# Helper moved outside to avoid lambda issues in UI definition
def get_audio_paths(path):
    return [os.path.abspath(os.path.join(path, f)) for f in os.listdir(path) if os.path.splitext(f)[1].lower() in ('.mp3', '.wav', '.flac', '.ogg')]

with gr.Blocks(title="🔊", theme=gr.themes.Base(primary_hue="blue", neutral_hue="zinc")) as app:
    with gr.Tabs():
        with gr.Tab("Inference"):
            with gr.Row():
                voice_model = gr.Dropdown(label="Model Voice", choices=sorted(names), value=lambda:sorted(names)[0] if len(sorted(names)) > 0 else '', interactive=True)
                refresh_button = gr.Button("Refresh", variant="primary")
                spk_item = gr.Slider(
                    minimum=0,
                    maximum=2333,
                    step=1,
                    label="Speaker ID",
                    value=0,
                    visible=False,
                    interactive=True,
                )
            vc_transform0 = gr.Number(
                label="Pitch", 
                value=0
            )
            but0 = gr.Button(value="Convert", variant="primary")
            with gr.Row():
                with gr.Column():
                    with gr.Row():
                        # Fixed: Use correct parameter name
                        dropbox = gr.Audio(label="Drop your audio here & hit the Reload button.", type="filepath")
                    with gr.Row():
                        record_button = gr.Audio(sources=["microphone"], label="OR Record audio.", type="filepath")
                    with gr.Row():
                        input_audio0 = gr.Dropdown(
                            label="Input Path",
                            value="",
                            choices=[],
                            allow_custom_value=True
                        )
                    with gr.Row():
                        audio_player = gr.Audio()
                        
                        # Fixed: Proper function to update audio player
                        def update_audio_player(path):
                            if path and os.path.exists(path):
                                return path
                            return None
                        
                        input_audio0.change(
                            fn=update_audio_player,
                            inputs=[input_audio0],
                            outputs=[audio_player]
                        )
                        
                        # Fixed: Handle record button properly
                        def handle_record(audio):
                            if audio:
                                return audio
                            return ""
                        
                        record_button.change(
                            fn=handle_record,
                            inputs=[record_button], 
                            outputs=[input_audio0]
                        )
                        
                        # Fixed: Handle dropbox upload
                        def handle_upload(audio):
                            if audio:
                                return audio
                            return ""
                        
                        dropbox.change(
                            fn=handle_upload,
                            inputs=[dropbox], 
                            outputs=[input_audio0]
                        )
                
                with gr.Column():
                    with gr.Accordion("Change Index", open=False):
                        file_index2 = gr.Dropdown(
                            label="Change Index",
                            choices=[],
                            interactive=True,
                            value=""
                        )
                        index_rate1 = gr.Slider(
                            minimum=0,
                            maximum=1,
                            label="Index Strength",
                            value=0.5,
                            interactive=True,
                        )
                    vc_output2 = gr.Audio(label="Output")
                    with gr.Accordion("General Settings", open=False):
                        f0method0 = gr.Radio(
                            label="Method",
                            choices=["pm", "harvest", "crepe", "rmvpe"]
                            if config.dml == False
                            else ["pm", "harvest", "rmvpe"],
                            value="rmvpe",
                            interactive=True,
                        )
                        filter_radius0 = gr.Slider(
                            minimum=0,
                            maximum=7,
                            label="Breathiness Reduction (Harvest only)",
                            value=3,
                            step=1,
                            interactive=True,
                        )
                        resample_sr0 = gr.Slider(
                            minimum=0,
                            maximum=48000,
                            label="Resample",
                            value=0,
                            step=1,
                            interactive=True,
                            visible=False
                        )
                        rms_mix_rate0 = gr.Slider(
                            minimum=0,
                            maximum=1,
                            label="Volume Normalization",
                            value=0,
                            interactive=True,
                        )
                        protect0 = gr.Slider(
                            minimum=0,
                            maximum=0.5,
                            label="Breathiness Protection (0 is enabled, 0.5 is disabled)",
                            value=0.33,
                            step=0.01,
                            interactive=True,
                        )
                    
                    file_index1 = gr.Textbox(
                        label="Index Path",
                        interactive=True,
                        visible=False
                    )

                    # Consolidated refresh logic
                    def refresh_ui():
                        # Get updated lists
                        # Assuming change_choices is imported from original or defined elsewhere
                        try:
                            model_choices, index_choices = change_choices()
                        except:
                            # Fallback if change_choices doesn't return two values
                            model_choices = []
                            index_choices = []
                        
                        audio_paths = get_audio_paths('audios')

                        # Helper to safely get the first item
                        def safe_first(data):
                            if not data:
                                return ""
                            if isinstance(data, dict):
                                return next(iter(data.values()))
                            if isinstance(data, list) and len(data) > 0:
                                return data[0]
                            return ""

                        default_audio = safe_first(audio_paths)
                        default_model = safe_first(model_choices)
                        default_index = safe_first(index_choices)

                        return (
                            gr.update(choices=model_choices, value=default_model), # voice_model
                            gr.update(choices=index_choices, value=default_index), # file_index2
                            gr.update(choices=audio_paths, value=default_audio)    # input_audio0
                        )

                    refresh_button.click(
                        fn=refresh_ui,
                        inputs=[],
                        outputs=[voice_model, file_index2, input_audio0],
                        api_name="infer_refresh",
                    )
            
            with gr.Row():
                f0_file = gr.File(label="F0 Path", visible=False)
            with gr.Row():
                vc_output1 = gr.Textbox(label="Information", placeholder="Welcome!", visible=False)
                but0.click(
                    vc.vc_single,  
                    [
                        spk_item,
                        input_audio0,
                        vc_transform0,
                        f0_file,
                        f0method0,
                        file_index1,
                        file_index2,
                        index_rate1,
                        filter_radius0,
                        resample_sr0,
                        rms_mix_rate0,
                        protect0,
                    ],
                    [vc_output1, vc_output2],
                    api_name="infer_convert",
                )  
                voice_model.change(
                    fn=vc.get_vc,
                    inputs=[voice_model, protect0, protect0],
                    outputs=[spk_item, protect0, protect0, file_index2, file_index2],
                    api_name="infer_change_voice",
                )

        with gr.Tab("Download Models"):
            with gr.Row():
                url_input = gr.Textbox(label="URL to model", value="", placeholder="https://...", scale=6)
                name_output = gr.Textbox(label="Save as", value="", placeholder="MyModel", scale=2)
                url_download = gr.Button(value="Download Model", scale=2)
                url_download.click(
                    inputs=[url_input, name_output],
                    outputs=[url_input],
                    fn=download_from_url,
                )
            with gr.Row():
                model_browser = gr.Dropdown(choices=list(model_library.models.keys()), label="OR Search Models (Quality UNKNOWN)", scale=5)
                download_from_browser = gr.Button(value="Get", scale=2)
                download_from_browser.click(
                    inputs=[model_browser],
                    outputs=[model_browser],
                    fn=lambda model: download_from_url(model_library.models[model], model),
                )

        with gr.Tab("Train"):
            with gr.Row():
                with gr.Column():
                    training_name = gr.Textbox(label="Name your model", value="My-Voice", placeholder="My-Voice")
                    np7 = gr.Slider(
                        minimum=0,
                        maximum=config.n_cpu,
                        step=1,
                        label="Number of CPU processes used to extract pitch features",
                        value=int(np.ceil(config.n_cpu / 1.5)),
                        interactive=True,
                    )
                    sr2 = gr.Radio(
                        label="Sampling Rate",
                        choices=["40k", "32k"],
                        value="32k",
                        interactive=True,
                        visible=False
                    )
                    if_f0_3 = gr.Radio(
                        label="Will your model be used for singing? If not, you can ignore this.",
                        choices=[True, False],
                        value=True,
                        interactive=True,
                        visible=False
                    )
                    version19 = gr.Radio(
                        label="Version",
                        choices=["v1", "v2"],
                        value="v2",
                        interactive=True,
                        visible=False,
                    )
                    dataset_folder = gr.Textbox(
                        label="dataset folder", value='dataset'
                    )
                    # Fixed: Use correct file handling
                    easy_uploader = gr.File(label="Drop your audio files here", file_count="multiple", file_types=["audio"])
                    but1 = gr.Button("1. Process", variant="primary")
                    info1 = gr.Textbox(label="Information", value="", visible=True)
                    
                    # Fixed: Proper file upload handling
                    def handle_file_upload(files, folder):
                        if not folder or folder.strip() == "":
                            gr.Warning('Please enter a folder name for your dataset')
                            return []
                        if not os.path.exists(folder):
                            os.makedirs(folder, exist_ok=True)
                        
                        saved_files = []
                        for file_obj in files:
                            if hasattr(file_obj, 'name'):  # Handle Gradio file object
                                filename = os.path.basename(file_obj.name)
                                dest_path = os.path.join(folder, filename)
                                shutil.copy2(file_obj.name, dest_path)
                                saved_files.append(dest_path)
                            elif isinstance(file_obj, str):  # Handle string path
                                filename = os.path.basename(file_obj)
                                dest_path = os.path.join(folder, filename)
                                shutil.copy2(file_obj, dest_path)
                                saved_files.append(dest_path)
                        return []
                    
                    easy_uploader.upload(
                        fn=handle_file_upload,
                        inputs=[easy_uploader, dataset_folder], 
                        outputs=[]
                    )
                    
                    gpus6 = gr.Textbox(
                        label="Enter the GPU numbers to use separated by -, (e.g. 0-1-2)",
                        value=gpus,
                        interactive=True,
                        visible=F0GPUVisible,
                    )
                    gpu_info9 = gr.Textbox(
                        label="GPU Info", value=gpu_info, visible=F0GPUVisible
                    )
                    spk_id5 = gr.Slider(
                        minimum=0,
                        maximum=4,
                        step=1,
                        label="Speaker ID",
                        value=0,
                        interactive=True,
                        visible=False
                    )
                    but1.click(
                        preprocess_dataset,
                        [dataset_folder, training_name, sr2, np7],
                        [info1],
                        api_name="train_preprocess",
                    ) 
                
                with gr.Column():
                    f0method8 = gr.Radio(
                        label="F0 extraction method",
                        choices=["pm", "harvest", "dio", "rmvpe", "rmvpe_gpu"],
                        value="rmvpe_gpu",
                        interactive=True,
                    )
                    gpus_rmvpe = gr.Textbox(
                        label="GPU numbers to use separated by -, (e.g. 0-1-2)",
                        value="%s-%s" % (gpus, gpus),
                        interactive=True,
                        visible=F0GPUVisible,
                    )
                    but2 = gr.Button("2. Extract Features", variant="primary")
                    info2 = gr.Textbox(label="Information", value="", max_lines=8)
                    f0method8.change(
                        fn=change_f0_method,
                        inputs=[f0method8],
                        outputs=[gpus_rmvpe],
                    )
                    but2.click(
                        extract_f0_feature,
                        [
                            gpus6,
                            np7,
                            f0method8,
                            if_f0_3,
                            training_name,
                            version19,
                            gpus_rmvpe,
                        ],
                        [info2],
                        api_name="train_extract_f0_feature",
                    )
                
                with gr.Column():
                    total_epoch11 = gr.Slider(
                        minimum=2,
                        maximum=1000,
                        step=1,
                        label="Epochs (more epochs may improve quality but takes longer)",
                        value=150,
                        interactive=True,
                    )
                    but4 = gr.Button("3. Train Index", variant="primary")
                    but3 = gr.Button("4. Train Model", variant="primary")
                    info3 = gr.Textbox(label="Information", value="", max_lines=10)
                    with gr.Accordion(label="General Settings", open=False):
                        gpus16 = gr.Textbox(
                            label="GPUs separated by -, (e.g. 0-1-2)",
                            value="0",
                            interactive=True,
                            visible=True
                        )
                        save_epoch10 = gr.Slider(
                            minimum=1,
                            maximum=50,
                            step=1,
                            label="Weight Saving Frequency",
                            value=25,
                            interactive=True,
                        )
                        batch_size12 = gr.Slider(
                            minimum=1,
                            maximum=40,
                            step=1,
                            label="Batch Size",
                            value=default_batch_size,
                            interactive=True,
                        )
                        if_save_latest13 = gr.Radio(
                            label="Only save the latest model",
                            choices=["yes", "no"],
                            value="yes",
                            interactive=True,
                            visible=False
                        )
                        if_cache_gpu17 = gr.Radio(
                            label="If your dataset is UNDER 10 minutes, cache it to train faster",
                            choices=["yes", "no"],
                            value="no",
                            interactive=True,
                        )
                        if_save_every_weights18 = gr.Radio(
                            label="Save small model at every save point",
                            choices=["yes", "no"],
                            value="yes",
                            interactive=True,
                        )
                        with gr.Accordion(label="Change pretrains", open=False):
                            # Fixed: Get pretrained choices
                            def get_pretrained_choices(sr, letter):
                                pretrained_dir = 'assets/pretrained_v2'
                                if not os.path.exists(pretrained_dir):
                                    return []
                                return [os.path.abspath(os.path.join(pretrained_dir, file)) 
                                       for file in os.listdir(pretrained_dir) 
                                       if file.endswith('.pth') and sr in file and letter in file]

                            pretrained_G14 = gr.Dropdown(
                                label="pretrained G",
                                choices=[],
                                value="",
                                interactive=True,
                                visible=True
                            )
                            pretrained_D15 = gr.Dropdown(
                                label="pretrained D",
                                choices=[],
                                value="",
                                visible=True,
                                interactive=True
                            )

                            def update_pretrained_dropdowns(sr, f0, ver):
                                sr_str = sr if isinstance(sr, str) else str(sr)
                                g_choices = get_pretrained_choices(sr_str, 'G')
                                d_choices = get_pretrained_choices(sr_str, 'D')
                                return (
                                    gr.update(choices=g_choices, value=g_choices[0] if g_choices else ""),
                                    gr.update(choices=d_choices, value=d_choices[0] if d_choices else "")
                                )
                            
                            # Bind update function to changes
                            sr2.change(fn=update_pretrained_dropdowns, inputs=[sr2, if_f0_3, version19], outputs=[pretrained_G14, pretrained_D15])
                            version19.change(fn=update_pretrained_dropdowns, inputs=[sr2, if_f0_3, version19], outputs=[pretrained_G14, pretrained_D15])

                    with gr.Row():
                        download_model = gr.Button('5.Download Model')
                    with gr.Row():
                        # Fixed: File handling for model download
                        model_files = gr.File(label='Your Model and Index file can be downloaded here:')
                        
                        def download_model_files(name):
                            if not name or name.strip() == "":
                                return [], "Please enter a model name"
                            
                            model_path = f'assets/weights/{name}'
                            index_pattern = f'logs/{name.split(".")[0]}/added_*.index'
                            
                            files = []
                            if os.path.exists(model_path):
                                files.extend([os.path.join(model_path, f) for f in os.listdir(model_path)])
                            files.extend(glob.glob(index_pattern))
                            
                            return files, f"Found {len(files)} files"
                        
                        download_model.click(
                            fn=download_model_files,
                            inputs=[training_name], 
                            outputs=[model_files, info3]
                        )
                    
                    if_f0_3.change(
                        change_f0,
                        [if_f0_3, sr2, version19],
                        [f0method8, pretrained_G14, pretrained_D15],
                    )
                    
                    but5 = gr.Button("1 Click Training", variant="primary", visible=False)
                    but3.click(
                        click_train,
                        [
                            training_name,
                            sr2,
                            if_f0_3,
                            spk_id5,
                            save_epoch10,
                            total_epoch11,
                            batch_size12,
                            if_save_latest13,
                            pretrained_G14,
                            pretrained_D15,
                            gpus16,
                            if_cache_gpu17,
                            if_save_every_weights18,
                            version19,
                        ],
                        info3,
                        api_name="train_start",
                    )
                    but4.click(train_index, [training_name, version19], info3)
                    but5.click(
                        train1key,
                        [
                            training_name,
                            sr2,
                            if_f0_3,
                            dataset_folder,
                            spk_id5,
                            np7,
                            f0method8,
                            save_epoch10,
                            total_epoch11,
                            batch_size12,
                            if_save_latest13,
                            pretrained_G14,
                            pretrained_D15,
                            gpus16,
                            if_cache_gpu17,
                            if_save_every_weights18,
                            version19,
                            gpus_rmvpe,
                        ],
                        info3,
                        api_name="train_start_all",
                    )

    # Populate UI on load
    app.load(
        fn=refresh_ui,
        inputs=[],
        outputs=[voice_model, file_index2, input_audio0]
    )

    if config.iscolab:
        app.launch(share=True, quiet=False)
    else:
        app.launch(
            server_name="0.0.0.0",
            inbrowser=not config.noautoopen,
            server_port=config.listen_port,
            quiet=True,
        )