Update app.py

app.py

@@ -118,10 +118,11 @@ def format_str_v3(s):
 
 
 def model_inference(
-    audio_input, language, output_format, use_itn, merge_vad, merge_length, ban_emo_unk
+    audio_input, language, use_itn, merge_vad, merge_length, ban_emo_unk
 ):
     if audio_input is None:
-        return "错误：请上传或录制音频。"
+        error_msg = "错误：请上传或录制音频。"
+        return error_msg, error_msg
     fs, input_wav = audio_input
     input_wav = input_wav.astype(np.float32) / np.iinfo(np.int16).max
     if len(input_wav.shape) > 1:
@@ -140,6 +141,11 @@ def model_inference(
         ban_emo_unk=ban_emo_unk,
     )
     raw_text = res[0].get("text", "未能识别出文本。")
+    plain_text = " ".join(re.sub(r"<\|[^>]*>", "", raw_text).split())
+    return raw_text, plain_text
+
+
+def format_output(raw_text, output_format):
     if output_format == "原始富文本":
         return raw_text
     elif output_format == "纯净文本":
@@ -148,7 +154,7 @@ def model_inference(
         return format_str_v3(raw_text)
 
 
-html_intro = """<div style="text-align: center; font-family: var(--font-sans);"><h1 style="font-size: 28px;">SenseVoice-Small 语音识别</h1><p>SenseVoice 是具有音频理解能力的音频基础模型，包括语音识别（ASR）、语种识别（LID）、语音情感识别（SER）和声学事件分类（AEC）或声学事件检测（AED）。</p><p style="font-size: small; color: #888;">支持 MP3, WAV, FLAC, M4A 等常见音频格式。</p></div>"""
+html_intro = """<div style="text-align: center; font-family: var(--font-sans);"><h1 style="font-size: 28px;">SenseVoice-Small 语音识别模型</h1><p>SenseVoice 是具有音频理解能力的音频基础模型，包括语音识别（ASR）、语种识别（LID）、语音情感识别（SER）和声学事件分类（AEC）或声学事件检测（AED）。</p><p style="font-size: small; color: #888;">支持 MP3, WAV, FLAC, M4A 等常见音频格式。</p></div>"""
 custom_css = "body, .gradio-container {font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, Oxygen, Ubuntu, Cantarell, 'Open Sans', 'Helvetica Neue', sans-serif;}"
 lang_options = [
     ("自动检测", "auto"),
@@ -162,17 +168,13 @@ lang_options = [
 
 with gr.Blocks(theme=gr.themes.Soft(), css=custom_css) as demo:
     gr.HTML(html_intro)
+    raw_text_state = gr.State("")
     with gr.Row():
         with gr.Column(scale=2):
             audio_inputs = gr.Audio(label="上传/录制音频")
             language_inputs = gr.Dropdown(
                 choices=lang_options, value="auto", label="源语言"
             )
-            output_format_dropdown = gr.Dropdown(
-                choices=["纯净文本", "原始富文本", "Emoji 格式"],
-                value="纯净文本",
-                label="输出格式",
-            )
             with gr.Accordion("高级设置", open=False):
                 use_itn_checkbox = gr.Checkbox(value=True, label="自动添加标点与格式化")
                 merge_vad_checkbox = gr.Checkbox(value=True, label="优化长音频断句")
@@ -182,20 +184,32 @@ with gr.Blocks(theme=gr.themes.Soft(), css=custom_css) as demo:
                 ban_emo_unk_checkbox = gr.Checkbox(value=False, label="强制情感分类")
             fn_button = gr.Button("开始识别", variant="primary")
         with gr.Column(scale=3):
+            output_format_dropdown = gr.Dropdown(
+                choices=["纯净文本", "原始富文本", "Emoji 格式"],
+                value="纯净文本",
+                label="输出格式",
+                interactive=True,
+            )
             text_outputs = gr.Textbox(label="识别结果", lines=25, show_copy_button=True)
+
     fn_button.click(
         fn=model_inference,
         inputs=[
             audio_inputs,
             language_inputs,
-            output_format_dropdown,
             use_itn_checkbox,
             merge_vad_checkbox,
             merge_length_slider,
             ban_emo_unk_checkbox,
         ],
-        outputs=text_outputs,
+        outputs=[raw_text_state, text_outputs],
         api_name="model_inference",
     )
 
+    output_format_dropdown.change(
+        fn=format_output,
+        inputs=[raw_text_state, output_format_dropdown],
+        outputs=text_outputs,
+    )
+
 demo.launch()