| import gradio as gr | |
| def change_choices(): | |
| names = [] | |
| for name in os.listdir(weight_root): | |
| if name.endswith(".pth"): | |
| names.append(name) | |
| index_paths = [] | |
| for root, dirs, files in os.walk(index_root, topdown=False): | |
| for name in files: | |
| if name.endswith(".index") and "trained" not in name: | |
| index_paths.append("%s/%s" % (root, name)) | |
| audios = [os.path.join(audio_root, file) for file in os.listdir(os.path.join(now_dir, "audios"))] | |
| return {"choices": sorted(names), "__type__": "update"}, {"choices": sorted(index_paths),"__type__": "update"},{ | |
| "choices": sorted(audios), "__type__": "update" | |
| } | |
| def paths_for_files(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="rose", neutral_hue="zinc")) as app: | |
| with gr.Tabs(): | |
| with gr.TabItem("Inference"): | |
| voice_model = gr.Dropdown(label="Model Voice", choices=sorted(names)) | |
| 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") | |
| dropbox = gr.File(label="Drop your audio here & hit the Reload button.") | |
| record_button = gr.Audio(source="microphone", label="OR Record audio.", type="filepath") | |
| input_audio0 = gr.Dropdown(label="Input Path", value=paths_for_files('audios')[0] if len(paths_for_files('audios')) > 0 else '', choices=paths_for_files('audios'), allow_custom_value=True) | |
| audio_player = gr.Audio() | |
| input_audio0.change(fn=lambda path: {"value": path, "__type__": "update"} if os.path.exists(path) else None, inputs=[input_audio0], outputs=[audio_player]) | |
| record_button.stop_recording(fn=lambda audio: audio, inputs=[record_button], outputs=[input_audio0]) | |
| dropbox.upload(fn=lambda audio: audio.name, inputs=[dropbox], outputs=[input_audio0]) | |
| with gr.Accordion("Change Index", open=False): | |
| file_index2 = gr.Dropdown(label="Change Index", choices=sorted(index_paths), interactive=True, value=sorted(index_paths)[0] if len(sorted(index_paths)) > 0 else '') | |
| 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) | |
| if voice_model is not None: | |
| vc.get_vc(voice_model.value, protect0, protect0) | |
| file_index1 = gr.Textbox(label="Index Path", interactive=True, visible=False) | |
| refresh_button.click(fn=change_choices, inputs=[], outputs=[voice_model, file_index2], api_name="infer_refresh") | |
| refresh_button.click(fn=lambda: {"choices": paths_for_files('audios'), "__type__": "update"}, inputs=[], outputs=[input_audio0]) | |
| refresh_button.click(fn=lambda: {"value": paths_for_files('audios')[0], "__type__": "update"} if len(paths_for_files('audios')) > 0 else {"value": "", "__type__": "update"}, inputs=[], outputs=[input_audio0]) | |
| f0_file = gr.File(label="F0 Path", visible=False) | |
| 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.TabItem("Download Models"): | |
| 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) | |
| 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.TabItem("Train"): | |
| 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) | |
| easy_uploader.upload(fn=lambda folder: os.makedirs(folder, exist_ok=True), inputs=[dataset_folder], outputs=[]) | |
| easy_uploader.upload(fn=lambda files, folder: [shutil.copy2(f.name, os.path.join(folder, os.path.split(f.name)[1])) for f in files] if folder != "" else gr.Warning('Please enter a folder name for your dataset'), 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) | |
| 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) | |
| f0method8.change(fn=change_f0_method, inputs=[f0method8], outputs=[gpus_rmvpe]) | |
| but1 = gr.Button("1. Process", variant="primary") | |
| info1 = gr.Textbox(label="Information", value="", visible=True) | |
| but1.click(preprocess_dataset, [dataset_folder, training_name, sr2, np7], [info1], api_name="train_preprocess") | |
| but2 = gr.Button("2. Extract Features", variant="primary") | |
| info2 = gr.Textbox(label="Information", value="", max_lines=8) | |
| but2.click(extract_f0_feature, [gpus6, np7, f0method8, if_f0_3, gpus_rmvpe, version19, dataset_folder], [info2], api_name="train_extract_features") | |
| but3 = gr.Button("3. Train", variant="primary") | |
| info3 = gr.Textbox(label="Information", value="", max_lines=8) | |
| but3.click(train_index, [gpus6, np7, f0method8, version19, dataset_folder, spk_id5], [info3], api_name="train_model") | |
| but4 = gr.Button("4. Extract Feature", variant="primary") | |
| info4 = gr.Textbox(label="Information", value="", max_lines=8) | |
| but4.click(extract_feature, [gpus6, np7, f0method8, version19, dataset_folder, spk_id5], [info4], api_name="train_extract_feature") | |
| app.queue(concurrency_count=3, max_size=20).launch() | |