first real test

app.py

@@ -1,7 +1,28 @@
 import gradio as gr
+import whisperx
+import whisper
 
-def greet(name):
-    return "Hello " + name + "!!"
+def transcribe(audio_file):
+    device = "cuda"
 
-iface = gr.Interface(fn=greet, inputs="text", outputs="text")
+    # Transcribe with original Whisper
+    model = whisper.load_model("large", device)
+    result = model.transcribe(audio_file)
+
+    # Load alignment model and metadata
+    model_a, metadata = whisperx.load_align_model(language_code=result["language"], device=device)
+
+    # Align Whisper output
+    result_aligned = whisperx.align(result["segments"], model_a, metadata, audio_file, device)
+
+    return result_aligned["segments"], result_aligned["word_segments"]
+
+# Define Gradio interface
+inputs = gr.inputs.Audio(source="upload", type="file")
+outputs = [
+    gr.outputs.Textbox(label="Segments (before alignment)"),
+    gr.outputs.Textbox(label="Segments (after alignment)"),
+]
+
+iface = gr.Interface(fn=transcribe, inputs=inputs, outputs=outputs, title="WhisperX Transcription")
 iface.launch()

requirements.txt

@@ -0,0 +1,2 @@
+gradio
+git+https://github.com/m-bain/whisperx.git