lenML/ChatTTS-Forge

1import gradio as gr2import torch3from modules.normalization import text_normalize4from modules.webui.webui_utils import (5    get_speakers,6    get_styles,7    split_long_text,8)9from modules.hf import spaces10 11 12# NOTE: 因为 text_normalize 需要使用 tokenizer13@torch.inference_mode()14@spaces.GPU15def merge_dataframe_to_ssml(dataframe, spk, style, seed):16    if style == "*auto":17        style = None18    if spk == "-1" or spk == -1:19        spk = None20    if seed == -1 or seed == "-1":21        seed = None22 23    ssml = ""24    indent = " " * 225 26    for i, row in dataframe.iterrows():27        ssml += f"{indent}<voice"28        if spk:29            ssml += f' spk="{spk}"'30        if style:31            ssml += f' style="{style}"'32        if seed:33            ssml += f' seed="{seed}"'34        ssml += ">\n"35        ssml += f"{indent}{indent}{text_normalize(row.iloc[1])}\n"36        ssml += f"{indent}</voice>\n"37    return f"<speak version='0.1'>\n{ssml}</speak>"38 39 40# 长文本处理41# 可以输入长文本，并选择切割方法，切割之后可以将拼接的SSML发送到SSML tab42# 根据 。 句号切割，切割之后显示到 data table43def create_spliter_tab(ssml_input, tabs):44    speakers = get_speakers()45 46    def get_speaker_show_name(spk):47        if spk.gender == "*" or spk.gender == "":48            return spk.name49        return f"{spk.gender} : {spk.name}"50 51    speaker_names = ["*random"] + [52        get_speaker_show_name(speaker) for speaker in speakers53    ]54 55    styles = ["*auto"] + [s.get("name") for s in get_styles()]56 57    with gr.Row():58        with gr.Column(scale=1):59            # 选择说话人 选择风格 选择seed60            with gr.Group():61                gr.Markdown("🗣️Speaker")62                spk_input_text = gr.Textbox(63                    label="Speaker (Text or Seed)",64                    value="female2",65                    show_label=False,66                )67                spk_input_dropdown = gr.Dropdown(68                    choices=speaker_names,69                    interactive=True,70                    value="female : female2",71                    show_label=False,72                )73                spk_rand_button = gr.Button(74                    value="🎲",75                    variant="secondary",76                )77            with gr.Group():78                gr.Markdown("🎭Style")79                style_input_dropdown = gr.Dropdown(80                    choices=styles,81                    interactive=True,82                    show_label=False,83                    value="*auto",84                )85            with gr.Group():86                gr.Markdown("🗣️Seed")87                infer_seed_input = gr.Number(88                    value=42,89                    label="Inference Seed",90                    show_label=False,91                    minimum=-1,92                    maximum=2**32 - 1,93                )94                infer_seed_rand_button = gr.Button(95                    value="🎲",96                    variant="secondary",97                )98 99            send_btn = gr.Button("📩Send to SSML", variant="primary")100 101        with gr.Column(scale=3):102            with gr.Group():103                gr.Markdown("📝Long Text Input")104                gr.Markdown("- 此页面用于处理超长文本")105                gr.Markdown("- 切割后，可以选择说话人、风格、seed，然后发送到SSML")106                long_text_input = gr.Textbox(107                    label="Long Text Input",108                    lines=10,109                    placeholder="输入长文本",110                    elem_id="long-text-input",111                    show_label=False,112                )113                long_text_split_button = gr.Button("🔪Split Text")114 115    with gr.Row():116        with gr.Column(scale=3):117            with gr.Group():118                gr.Markdown("🎨Output")119                long_text_output = gr.DataFrame(120                    headers=["index", "text", "length"],121                    datatype=["number", "str", "number"],122                    elem_id="long-text-output",123                    interactive=False,124                    wrap=True,125                    value=[],126                )127 128    spk_input_dropdown.change(129        fn=lambda x: x.startswith("*") and "-1" or x.split(":")[-1].strip(),130        inputs=[spk_input_dropdown],131        outputs=[spk_input_text],132    )133    spk_rand_button.click(134        lambda x: int(torch.randint(0, 2**32 - 1, (1,)).item()),135        inputs=[spk_input_text],136        outputs=[spk_input_text],137    )138    infer_seed_rand_button.click(139        lambda x: int(torch.randint(0, 2**32 - 1, (1,)).item()),140        inputs=[infer_seed_input],141        outputs=[infer_seed_input],142    )143    long_text_split_button.click(144        split_long_text,145        inputs=[long_text_input],146        outputs=[long_text_output],147    )148 149    infer_seed_rand_button.click(150        lambda x: int(torch.randint(0, 2**32 - 1, (1,)).item()),151        inputs=[infer_seed_input],152        outputs=[infer_seed_input],153    )154 155    send_btn.click(156        merge_dataframe_to_ssml,157        inputs=[158            long_text_output,159            spk_input_text,160            style_input_dropdown,161            infer_seed_input,162        ],163        outputs=[ssml_input],164    )165 166    def change_tab():167        return gr.Tabs(selected="ssml")168 169    send_btn.click(change_tab, inputs=[], outputs=[tabs])170