Update app.py

app.py

@@ -25,6 +25,8 @@ from torch.utils.data import DataLoader, Dataset
 from tqdm import tqdm
 from clap_wrapper import get_clap_audio_feature, get_clap_text_feature
 
+from tools.sentence import extrac, is_japanese, is_chinese, seconds_to_ass_time, extract_text_from_file, remove_annotations,extract_and_convert
+import re
 
 import gradio as gr
 
@@ -36,13 +38,14 @@ import commons
 from text import cleaned_text_to_sequence, get_bert
 from text.cleaner import clean_text
 import utils
-
+from scipy.io.wavfile import write
 from models import SynthesizerTrn
 from text.symbols import symbols
 import sys
+import shutil
 
 net_g = None
-'''
+
 device = (
         "cuda:0"
         if torch.cuda.is_available()
@@ -52,8 +55,7 @@ device = (
             else "cpu"
         )
     )
-'''
-device = "cpu"
+
 BandList = {
         "PoppinParty":["香澄","有咲","たえ","りみ","沙綾"],
         "Afterglow":["蘭","モカ","ひまり","巴","つぐみ"],
@@ -156,7 +158,6 @@ def infer(
         emo = emo.to(device).unsqueeze(0)
         del phones
         speakers = torch.LongTensor([hps.data.spk2id[sid]]).to(device)
-        print(text)
         audio = (
             net_g.infer(
                 x_tst,
@@ -188,6 +189,161 @@ def infer(
             torch.cuda.empty_cache()
         return (hps.data.sampling_rate,gr.processing_utils.convert_to_16_bit_wav(audio))
 
+
+def generate_audio_and_srt_for_group(group, outputPath, group_index, sampling_rate, speaker, sdp_ratio, noise_scale, noise_scale_w, length_scale,spealerList,silenceTime):
+    audio_fin = []
+    ass_entries = []
+    start_time = 0
+    #speaker = random.choice(cara_list)
+    ass_header = """[Script Info]
+; 我没意见
+Title: Audiobook
+ScriptType: v4.00+
+WrapStyle: 0
+PlayResX: 640
+PlayResY: 360
+ScaledBorderAndShadow: yes
+[V4+ Styles]
+Format: Name, Fontname, Fontsize, PrimaryColour, SecondaryColour, OutlineColour, BackColour, Bold, Italic, Underline, StrikeOut, ScaleX, ScaleY, Spacing, Angle, BorderStyle, Outline, Shadow, Alignment, MarginL, MarginR, MarginV, Encoding
+Style: Default,Arial,20,&H00FFFFFF,&H000000FF,&H00000000,&H00000000,0,0,0,0,100,100,0,0,1,1,1,2,10,10,10,1
+[Events]
+Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text
+"""
+
+    for sentence in group:
+            print(sentence)
+            FakeSpeaker = sentence.split("|")[0]
+            print(FakeSpeaker)
+            SpeakersList = re.split('\n', spealerList)
+            if FakeSpeaker in list(hps.data.spk2id.keys()):
+                speaker = FakeSpeaker
+            for i in SpeakersList:
+                if FakeSpeaker == i.split("|")[1]:
+                    speaker = i.split("|")[0]
+            if sentence != '\n':
+                audio = infer_simple((remove_annotations(sentence.split("|")[-1]).replace(" ","")+"。").replace("，。","。").replace("。。","。"), sdp_ratio, noise_scale, noise_scale_w, length_scale,speaker)
+                silence_frames = int(silenceTime * 44010) if is_chinese(sentence) else int(silenceTime * 44010)
+                silence_data = np.zeros((silence_frames,), dtype=audio.dtype)
+                audio_fin.append(audio)
+                audio_fin.append(silence_data)
+
+                duration = len(audio) / sampling_rate
+                print(duration)
+                end_time = start_time + duration + silenceTime
+                ass_entries.append("Dialogue: 0,{},{},".format(seconds_to_ass_time(start_time), seconds_to_ass_time(end_time)) + "Default,,0,0,0,,{}".format(sentence.replace("|","：")))
+                start_time = end_time
+
+    wav_filename = os.path.join(outputPath, f'audiobook_part_{group_index}.wav')
+    ass_filename = os.path.join(outputPath, f'audiobook_part_{group_index}.ass')
+
+    write(wav_filename, sampling_rate, np.concatenate(audio_fin))
+
+    with open(ass_filename, 'w', encoding='utf-8') as f:
+        f.write(ass_header + '\n'.join(ass_entries))
+    return (hps.data.sampling_rate, np.concatenate(audio_fin))
+
+
+def infer_simple(
+    text,
+    sdp_ratio,
+    noise_scale,
+    noise_scale_w,
+    length_scale,
+    sid,
+    emotion = '',
+    reference_audio=None,
+    skip_start=False,
+    skip_end=False,
+    style_text=None,
+    style_weight=0.7,
+):
+    language = "JP"
+    if isinstance(reference_audio, np.ndarray):
+        emo = get_clap_audio_feature(reference_audio, device)
+    else:
+        emo = get_clap_text_feature(emotion, device)
+    emo = torch.squeeze(emo, dim=1)
+
+    bert, phones, tones, lang_ids = get_text(
+        text,
+        language,
+        hps,
+        device,
+        style_text=style_text,
+        style_weight=style_weight,
+    )
+    if skip_start:
+        phones = phones[3:]
+        tones = tones[3:]
+        lang_ids = lang_ids[3:]
+        bert = bert[:, 3:]
+    if skip_end:
+        phones = phones[:-2]
+        tones = tones[:-2]
+        lang_ids = lang_ids[:-2]
+        bert = bert[:, :-2]
+    with torch.no_grad():
+        x_tst = phones.to(device).unsqueeze(0)
+        tones = tones.to(device).unsqueeze(0)
+        lang_ids = lang_ids.to(device).unsqueeze(0)
+        bert = bert.to(device).unsqueeze(0)
+        x_tst_lengths = torch.LongTensor([phones.size(0)]).to(device)
+        emo = emo.to(device).unsqueeze(0)
+        del phones
+        speakers = torch.LongTensor([hps.data.spk2id[sid]]).to(device)
+        audio = (
+            net_g.infer(
+                x_tst,
+                x_tst_lengths,
+                speakers,
+                tones,
+                lang_ids,
+                bert,
+                emo,
+                sdp_ratio=sdp_ratio,
+                noise_scale=noise_scale,
+                noise_scale_w=noise_scale_w,
+                length_scale=length_scale,
+            )[0][0, 0]
+            .data.cpu()
+            .float()
+            .numpy()
+        )
+        del (
+            x_tst,
+            tones,
+            lang_ids,
+            bert,
+            x_tst_lengths,
+            speakers,
+            emo,
+        )  # , emo
+        if torch.cuda.is_available():
+            torch.cuda.empty_cache()
+    return audio
+
+def audiobook(inputFile, groupsize, speaker, sdp_ratio, noise_scale, noise_scale_w, length_scale,spealerList,silenceTime,filepath,raw_text):
+    directory_path = filepath if torch.cuda.is_available() else "books"
+ 
+    if os.path.exists(directory_path):
+        shutil.rmtree(directory_path)
+
+    os.makedirs(directory_path)
+    if inputFile:
+        text = extract_text_from_file(inputFile.name)
+    else:
+        text = raw_text
+    sentences = extrac(extract_and_convert(text))
+    GROUP_SIZE = groupsize
+    for i in range(0, len(sentences), GROUP_SIZE):
+        group = sentences[i:i+GROUP_SIZE]
+        if spealerList == "":
+            spealerList = "无"
+        result = generate_audio_and_srt_for_group(group,directory_path, i//GROUP_SIZE + 1, 44100, speaker, sdp_ratio, noise_scale, noise_scale_w, length_scale,spealerList,silenceTime)
+        if not torch.cuda.is_available():
+            return result
+    return result
+
 def loadmodel(model):
     _ = net_g.eval()
     _ = utils.load_checkpoint(model, net_g, None, skip_optimizer=True)
@@ -221,7 +377,7 @@ if __name__ == "__main__":
                                         minimum=0.1, maximum=2, value=1, step=0.01, label="语速调节"
                                     )
                                 emotion = gr.Textbox(
-                                        label="情感标注文本t",
+                                        label="情感标注文本",
                                         value = 'なんではるひかげやったの?!!'
                                     )
                                 style_weight = gr.Slider(
@@ -274,6 +430,66 @@ if __name__ == "__main__":
                         ],
                         outputs=[audio_output],
                     )
-
+        with gr.Tab('拓展功能'):
+            with gr.Row():
+                with gr.Column():
+                    gr.Markdown(
+                                    f"从 <a href='https://nijigaku.top/2023/10/03/BangDreamTTS/'>我的博客站点</a> 查看自制galgame使用说明\n</a>"
+                                )
+                    inputFile = gr.UploadButton(label="txt文件输入")
+                    raw_text = gr.TextArea(
+                                        label="文本输入",
+                                        info="输入纯日语或者中文",
+                                        value="つくし|我是来结束这个乐队的。",
+                    )
+                    groupSize = gr.Slider(
+                    minimum=10, maximum=1000 if  torch.cuda.is_available() else 50,value = 50, step=1, label="单个音频文件包含的最大字数"
+                    )
+                    silenceTime = gr.Slider(
+                    minimum=0, maximum=1, value=0.5, step=0.01, label="句子的间隔"
+                    )
+                    filepath = gr.TextArea(
+                                        label="本地合成时的音频存储文件夹(会清空文件夹)",
+                                        value = "D:/audiobook/book1",
+                    )
+                    spealerList = gr.TextArea(
+                                        label="角色对应表,左边是你想要在每一句话合成中用到的speaker(见角色清单)右边是你上传文本时分隔符左边设置的说话人:{ChoseSpeakerFromConfigList}|{SeakerInUploadText}",
+                                        value = "ましろ|真白\n七深|七深\n透子|透子\nつくし|筑紫\n瑠唯|瑠唯\nそよ|素世\n祥子|祥子",
+                    )                  
+                    speaker = gr.Dropdown(
+                        choices=speakers, value = "ましろ", label="选择默认说话人"
+                    )
+                with gr.Column():
+                    sdp_ratio = gr.Slider(
+                    minimum=0, maximum=1, value=0.2, step=0.01, label="SDP/DP混合比"
+                    )
+                    noise_scale = gr.Slider(
+                        minimum=0.1, maximum=2, value=0.6, step=0.01, label="感情调节"
+                    )
+                    noise_scale_w = gr.Slider(
+                        minimum=0.1, maximum=2, value=0.667, step=0.01, label="音素长度"
+                    )
+                    length_scale = gr.Slider(
+                        minimum=0.1, maximum=2, value=1, step=0.01, label="生成长度"
+                    )
+                    LastAudioOutput = gr.Audio(label="当使用cuda时才能在本地文件夹浏览全部文件")
+                    btn2 = gr.Button("点击生成", variant="primary")
+                btn2.click(
+                    audiobook,
+                    inputs=[
+                        inputFile,
+                        groupSize,
+                        speaker,
+                        sdp_ratio,
+                        noise_scale,
+                        noise_scale_w,
+                        length_scale,
+                        spealerList,
+                        silenceTime,
+                        filepath,
+                        raw_text
+                    ],
+                    outputs=[LastAudioOutput],
+                )
     print("推理页面已开启!")
-    app.launch(share=True)
+    app.launch()