second update

app.py

@@ -1,23 +1,22 @@
 import gradio as gr
 import whisperx
-import whisper
 
-def transcribe(audio_file):
-    device = "cuda" if torch.cuda.is_available() else "cpu"
+# 加载WhisperX模型
+model = whisperx.load_model("base", device="cpu")  # 如果有GPU，可将"cpu"改为"cuda"
 
-    # Transcribe with original Whisper
-    model = whisper.load_model("large", device)
-    result = model.transcribe(audio_file)
+def transcribe(audio_path):
+    # 使用WhisperX进行转录
+    result = model.transcribe(audio_path)
+    return result['text']
 
-    # Load alignment model and metadata
-    model_a, metadata = whisperx.load_align_model(language_code=result["language"], device=device)
+# 创建Gradio接口
+iface = gr.Interface(
+    fn=transcribe,
+    inputs=gr.Audio(source="upload", type="filepath"),
+    outputs="text",
+    title="WhisperX 语音转文字",
+    description="上传音频文件，使用WhisperX模型进行转录。"
+)
 
-    # Align Whisper output
-    result_aligned = whisperx.align(result["segments"], model_a, metadata, audio_file, device)
-
-    return {"aligned": result_aligned["segments"], "word_segments": result_aligned["word_segments"]}
-
-inputs = gr.inputs.Audio(source="upload", type="filepath")
-outputs = gr.outputs.JSON()
-
-gr.Interface(fn=transcribe, inputs=inputs, outputs=outputs).launch()
+if __name__ == "__main__":
+    iface.launch()