Update app.py

app.py

@@ -1,50 +1,200 @@
 import gradio as gr
-from funasr import AutoModel
+import numpy as np
 import torch
+import torchaudio
+from funasr import AutoModel
 
-model_name = "iic/SenseVoiceSmall"
+print("正在加载SenseVoice-Small模型，请稍候...")
+model = AutoModel(
+    model="iic/SenseVoiceSmall",
+    vad_model="iic/speech_fsmn_vad_zh-cn-16k-common-pytorch",
+    vad_kwargs={"max_single_segment_time": 30000},
+    trust_remote_code=True,
+)
+print("模型加载成功！")
 
-device = "cuda:0" if torch.cuda.is_available() else "cpu"
+emo_dict = {
+    "<|HAPPY|>": "😊",
+    "<|SAD|>": "😔",
+    "<|ANGRY|>": "😡",
+    "<|NEUTRAL|>": "",
+    "<|FEARFUL|>": "😰",
+    "<|DISGUSTED|>": "🤢",
+    "<|SURPRISED|>": "😮",
+}
+
+event_dict = {
+    "<|BGM|>": "🎼",
+    "<|Speech|>": "",
+    "<|Applause|>": "👏",
+    "<|Laughter|>": "😀",
+    "<|Cry|>": "😭",
+    "<|Sneeze|>": "🤧",
+    "<|Breath|>": "",
+    "<|Cough|>": "🤧",
+}
+
+emoji_dict = {
+    "<|nospeech|><|Event_UNK|>": "❓",
+    "<|zh|>": "",
+    "<|en|>": "",
+    "<|yue|>": "",
+    "<|ja|>": "",
+    "<|ko|>": "",
+    "<|nospeech|>": "",
+    "<|HAPPY|>": "😊",
+    "<|SAD|>": "😔",
+    "<|ANGRY|>": "😡",
+    "<|NEUTRAL|>": "",
+    "<|BGM|>": "🎼",
+    "<|Speech|>": "",
+    "<|Applause|>": "👏",
+    "<|Laughter|>": "😀",
+    "<|FEARFUL|>": "😰",
+    "<|DISGUSTED|>": "🤢",
+    "<|SURPRISED|>": "😮",
+    "<|Cry|>": "😭",
+    "<|EMO_UNKNOWN|>": "",
+    "<|Sneeze|>": "🤧",
+    "<|Breath|>": "",
+    "<|Cough|>": "😷",
+    "<|Sing|>": "",
+    "<|Speech_Noise|>": "",
+    "<|withitn|>": "",
+    "<|woitn|>": "",
+    "<|GBG|>": "",
+    "<|Event_UNK|>": "",
+}
+
+lang_dict = {
+    "<|zh|>": "<|lang|>",
+    "<|en|>": "<|lang|>",
+    "<|yue|>": "<|lang|>",
+    "<|ja|>": "<|lang|>",
+    "<|ko|>": "<|lang|>",
+    "<|nospeech|>": "<|lang|>",
+}
+
+emo_set = {"😊", "😔", "😡", "😰", "🤢", "😮"}
+event_set = {"🎼", "👏", "😀", "😭", "🤧", "😷"}
+
+
+def format_str_v2(s):
+    sptk_dict = {}
+    for sptk in emoji_dict:
+        sptk_dict[sptk] = s.count(sptk)
+        s = s.replace(sptk, "")
+    emo = "<|NEUTRAL|>"
+    for e in emo_dict:
+        if sptk_dict.get(e, 0) > sptk_dict.get(emo, 0):
+            emo = e
+    for e in event_dict:
+        if sptk_dict.get(e, 0) > 0:
+            s = event_dict[e] + s
+    s = s + emo_dict.get(emo, "")
+
+    for emoji in emo_set.union(event_set):
+        s = s.replace(" " + emoji, emoji)
+        s = s.replace(emoji + " ", emoji)
+    return s.strip()
+
+
+def format_str_v3(s):
+    def get_emo(s):
+        return s[-1] if s and s[-1] in emo_set else None
+
+    def get_event(s):
+        return s[0] if s and s[0] in event_set else None
+
+    s = s.replace("<|nospeech|><|Event_UNK|>", "❓")
+    for lang in lang_dict:
+        s = s.replace(lang, "<|lang|>")
+    s_list = [format_str_v2(s_i).strip(" ") for s_i in s.split("<|lang|>")]
+    new_s = " " + (s_list[0] if s_list else "")
+    cur_ent_event = get_event(new_s)
+    for i in range(1, len(s_list)):
+        if not s_list[i]:
+            continue
+        if get_event(s_list[i]) == cur_ent_event and cur_ent_event is not None:
+            s_list[i] = s_list[i][1:]
+        cur_ent_event = get_event(s_list[i])
+        if get_emo(s_list[i]) is not None and get_emo(s_list[i]) == get_emo(new_s):
+            new_s = new_s[:-1]
+        new_s += s_list[i].strip().lstrip()
+    new_s = new_s.replace("The.", " ")
+    return new_s.strip()
 
-try:
-    model = AutoModel(
-        model=model_name,
-        vad_model="fsmn-vad",
-        vad_kwargs={"max_single_segment_time": 30000},
-        device=device,
-    )
-    model_loaded_successfully = True
-except Exception:
-    model = None
-    model_loaded_successfully = False
 
+def model_inference(audio_input, language):
+    if audio_input is None:
+        return "错误：请先上传或录制音频。"
 
-def transcribe_audio(audio_file_path):
-    if not model_loaded_successfully:
-        return "模型加载失败。很可能是免费硬件的内存不足，请按教程切换到 'iic/SenseVoiceSmall' 模型。"
+    fs, input_wav = audio_input
+    input_wav = input_wav.astype(np.float32) / 32768.0
 
-    if audio_file_path is None:
-        return "错误：请上传或录制音频。"
+    if len(input_wav.shape) > 1:
+        input_wav = input_wav.mean(-1)
+
+    if fs != 16000:
+        print(f"音频采样率不是16k，正在重采样: {fs}")
+        resampler = torchaudio.transforms.Resample(orig_freq=fs, new_freq=16000)
+        input_wav = resampler(torch.from_numpy(input_wav)).numpy()
+
+    print(f"开始识别... 语言: {language}")
 
     res = model.generate(
-        input=audio_file_path,
-        batch_size_s=60,
-        language="auto",
+        input=input_wav,
+        cache={},
+        language=language,
         use_itn=True,
+        batch_size_s=60,
+        merge_vad=True,
     )
 
-    return res[0]["text"]
+    raw_text = res[0].get("text", "未能识别出文本。")
+    print(f"原始输出: {raw_text}")
 
+    # 使用格式化函数处理文本
+    formatted_text = format_str_v3(raw_text)
+    print(f"格式化输出: {formatted_text}")
 
-iface = gr.Interface(
-    fn=transcribe_audio,
-    inputs=gr.Audio(
-        sources=["upload", "microphone"], type="filepath", label="上传音频或录音"
-    ),
-    outputs=gr.Textbox(label="识别结果"),
-    title="SenseVoice 在线体验版",
-    description="应用启动时需要下载和加载模型，请耐心等待。",
-    allow_flagging="never",
-)
+    return formatted_text
+
+
+html_content = """
+<div style="text-align: center;">
+    <h2 style="font-size: 22px;">语音理解模型： SenseVoice-Small</h2>
+    <p style="font-size: 16px;">
+        这是一个部署在Hugging Face上的SenseVoice-Small语音识别模型。<br>
+        它支持中、英、粤、日、韩等多语言识别，并能识别情感和声音事件。
+    </p>
+    <p style="font-size: 16px;">
+        <b>用法:</b> 上传音频或使用麦克风录音，在下方“配置”中选择识别语言，然后点击“开始”。
+    </p>
+</div>
+"""
+
+with gr.Blocks(theme=gr.themes.Soft()) as demo:
+    gr.HTML(html_content)
+
+    with gr.Row():
+        with gr.Column():
+            audio_inputs = gr.Audio(label="上传音频或使用麦克风录音")
+
+            with gr.Accordion("配置", open=True):
+                language_inputs = gr.Dropdown(
+                    choices=["auto", "zh", "en", "yue", "ja", "ko", "nospeech"],
+                    value="auto",
+                    label="选择语言 (Select Language)",
+                )
+
+            fn_button = gr.Button("开始识别", variant="primary")
+
+    with gr.Row():
+        text_outputs = gr.Textbox(label="识别结果", lines=20)
+
+    fn_button.click(
+        fn=model_inference, inputs=[audio_inputs, language_inputs], outputs=text_outputs
+    )
 
-iface.launch()
+demo.launch()