Create app_demo.py

app_demo.py

@@ -0,0 +1,125 @@
+# Copyright (c) 2024 Alibaba Inc (authors: Xiang Lyu)
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#   http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import os
+import sys
+ROOT_DIR = os.path.dirname(os.path.abspath(__file__))
+sys.path.append('{}/third_party/Matcha-TTS'.format(ROOT_DIR))
+
+#os.system("apt-get install sox libsox-dev")
+os.system("pip install pysox")
+os.system("pip install torch==2.1.0+cu121 torchvision==0.16.0+cu121 torchaudio==2.1.0 torchtext==0.16.0+cpu torchdata==0.7.0 --index-url https://download.pytorch.org/whl/cu121")
+
+
+import argparse
+import gradio as gr
+import numpy as np
+import torch
+import torchaudio
+import random
+import librosa
+
+import logging
+logging.getLogger('matplotlib').setLevel(logging.WARNING)
+
+import spaces
+
+logging.basicConfig(level=logging.DEBUG,
+                    format='%(asctime)s %(levelname)s %(message)s')
+
+def generate_seed():
+    seed = random.randint(1, 100000000)
+    return {
+        "__type__": "update",
+        "value": seed
+    }
+
+def set_all_random_seed(seed):
+    random.seed(seed)
+    np.random.seed(seed)
+    torch.manual_seed(seed)
+    torch.cuda.manual_seed_all(seed)
+
+max_val = 0.8
+def postprocess(speech, top_db=60, hop_length=220, win_length=440):
+    speech, _ = librosa.effects.trim(
+        speech, top_db=top_db,
+        frame_length=win_length,
+        hop_length=hop_length
+    )
+    if speech.abs().max() > max_val:
+        speech = speech / speech.abs().max() * max_val
+    speech = torch.concat([speech, torch.zeros(1, int(target_sr * 0.2))], dim=1)
+    return speech
+
+inference_mode_list = ['预训练音色', '3s极速复刻', '跨语种复刻', '自然语言控制']
+instruct_dict = {'预训练音色': '1. 选择预训练音色\n2.点击生成音频按钮',
+                 '3s极速复刻': '1. 选择prompt音频文件，或录入prompt音频，若同时提供，优先选择prompt音频文件\n2. 输入prompt文本\n3.点击生成音频按钮',
+                 '跨语种复刻': '1. 选择prompt音频文件，或录入prompt音频，若同时提供，优先选择prompt音频文件\n2.点击生成音频按钮',
+                 '自然语言控制': '1. 输入instruct文本\n2.点击生成音频按钮'}
+def change_instruction(mode_checkbox_group):
+    return instruct_dict[mode_checkbox_group]
+
+@spaces.GPU
+def generate_audio(tts_text, mode_checkbox_group, sft_dropdown, prompt_text, prompt_wav_upload, prompt_wav_record, instruct_text, seed):
+
+    return "jay_short.wav"
+
+def main():
+    with gr.Blocks() as demo:
+        gr.Markdown("### 代码库 [CosyVoice](https://github.com/FunAudioLLM/CosyVoice) 预训练模型 [CosyVoice-300M](https://www.modelscope.cn/models/iic/CosyVoice-300M) [CosyVoice-300M-Instruct](https://www.modelscope.cn/models/iic/CosyVoice-300M-Instruct) [CosyVoice-300M-SFT](https://www.modelscope.cn/models/iic/CosyVoice-300M-SFT)")
+        gr.Markdown("#### 请输入需要合成的文本，选择推理模式，并按照提示步骤进行操作")
+
+        tts_text = gr.Textbox(label="输入合成文本", lines=1, value="我是通义实验室语音团队全新推出的生成式语音大模型，提供舒适自然的语音合成能力。")
+
+        with gr.Row():
+            mode_checkbox_group = gr.Radio(choices=inference_mode_list, label='选择推理模式', value=inference_mode_list[0])
+            instruction_text = gr.Text(label="操作步骤", value=instruct_dict[inference_mode_list[0]], scale=0.5)
+            sft_dropdown = gr.Dropdown(choices=sft_spk, label='选择预训练音色', value=sft_spk[0], scale=0.25)
+            with gr.Column(scale=0.25):
+                seed_button = gr.Button(value="\U0001F3B2")
+                seed = gr.Number(value=0, label="随机推理种子")
+
+        with gr.Row():
+            prompt_wav_upload = gr.Audio(sources='upload', type='filepath', label='选择prompt音频文件，注意采样率不低于16khz')
+            prompt_wav_record = gr.Audio(sources='microphone', type='filepath', label='录制prompt音频文件')
+        prompt_text = gr.Textbox(label="输入prompt文本", lines=1, placeholder="请输入prompt文本，需与prompt音频内容一致，暂时不支持自动识别...", value='')
+        instruct_text = gr.Textbox(label="输入instruct文本", lines=1, placeholder="请输入instruct文本.", value='')
+
+        generate_button = gr.Button("生成音频")
+
+        audio_output = gr.Audio(label="合成音频")
+
+        seed_button.click(generate_seed, inputs=[], outputs=seed)
+        generate_button.click(generate_audio,
+                              inputs=[tts_text, mode_checkbox_group, sft_dropdown, prompt_text, prompt_wav_upload, prompt_wav_record, instruct_text, seed],
+                              outputs=[audio_output])
+        mode_checkbox_group.change(fn=change_instruction, inputs=[mode_checkbox_group], outputs=[instruction_text])
+    demo.queue()
+    demo.launch(share=False, show_error=True)
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--port',
+                        type=int,
+                        default=8000)
+    parser.add_argument('--model_dir',
+                        type=str,
+                        default='iic/CosyVoice-300M',
+                        help='local path or modelscope repo id')
+    args = parser.parse_args()
+    cosyvoice = CosyVoice(args.model_dir)
+    sft_spk = cosyvoice.list_avaliable_spks()
+    prompt_sr, target_sr = 16000, 22050
+    default_data = np.zeros(target_sr)
+    main()