Update app.py

app.py

@@ -10,6 +10,7 @@ import gradio as gr
 import torchaudio
 import whisper.tokenizer
 import ffmpeg
+import time
 
 # -----------------------------
 # Helper Functions
@@ -58,6 +59,21 @@ def generate_srt(segments):
             subs.append(srt.Subtitle(index=i+1, start=start, end=end, content=text))
     return srt.compose(subs)
 
+# -----------------------------
+# Model Loading Functions
+# -----------------------------
+
+def load_kotani_model():
+    status = "📥 Loading Kotani Whisper Small model..."
+    whisper.load_model("small", download_root=".")
+    return status
+
+def load_khaiii_model():
+    status = "📥 Loading Khaiii Wav2Vec2 model..."
+    Wav2Vec2Processor.from_pretrained("khaiii/wav2vec2-xls1r-aishell-korean")
+    Wav2Vec2ForCTC.from_pretrained("khaiii/wav2vec2-xls1r-aishell-korean")
+    return status
+
 # -----------------------------
 # Transcription Functions
 # -----------------------------
@@ -106,6 +122,16 @@ def process_video(youtube_url, video_file, selected_model, translate, target_lan
             yield status, "", None
             audio_path = download_youtube_audio(youtube_url)
         elif video_file:
+            status = "📼 Waiting for upload to complete..."
+            yield status, "", None
+
+            # Wait until file exists
+            start_time = time.time()
+            while not os.path.exists(video_file.name):
+                if time.time() - start_time > 30:
+                    raise RuntimeError("Timeout: File upload took too long.")
+                time.sleep(1)
+
             status = "📼 Extracting audio from video..."
             yield status, "", None
             audio_path = extract_audio_from_video(video_file.name)
@@ -173,6 +199,8 @@ model_desc_khaiii = """
 """
 
 with gr.Blocks(theme=gr.themes.Soft()) as demo:
+    status_box = gr.Textbox(label="Status", interactive=False)
+
     gr.Markdown("## 🌍 Multilingual Subtitle Generator")
     gr.Markdown("Upload a video or paste a YouTube link. Automatically detect language and optionally translate subtitles.")
 
@@ -187,21 +215,27 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
             khaiii_btn = gr.Button("✅ Select Khaiii Wav2Vec2")
             gr.HTML(model_desc_khaiii)
 
-    selected_model = gr.State(value="kotani")
-
     def select_kotani():
-        return "kotani"
+        msg = load_kotani_model()
+        return "kotani", msg
 
     def select_khaiii():
-        return "khaiii"
+        msg = load_khaiii_model()
+        return "khaiii", msg
 
-    kotani_btn.click(fn=select_kotani, outputs=selected_model)
-    khaiii_btn.click(fn=select_khaiii, outputs=selected_model)
+    kotani_btn.click(fn=select_kotani, outputs=[selected_model, status_box])
+    khaiii_btn.click(fn=select_khaiii, outputs=[selected_model, status_box])
 
     gr.Markdown("### 📥 Input Source")
     with gr.Row():
         youtube_url = gr.Textbox(label="YouTube URL", scale=2)
-        video_upload = gr.File(label="Upload Video", type="filepath", scale=1)  # ✅ FIXED HERE
+        video_upload = gr.File(
+            label="Upload Video",
+            type="filepath",
+            file_types=["video"],
+            elem_props={"accept": "video/*"},
+            scale=1
+        )
 
     gr.Markdown("### 🌍 Translation Options")
     with gr.Row():
@@ -213,7 +247,6 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
 
     translate_checkbox.change(fn=toggle_translate, inputs=translate_checkbox, outputs=target_lang)
 
-    status_box = gr.Textbox(label="Status", interactive=False)
     subtitle_preview = gr.Textbox(label="Generated Subtitles", lines=10)
     download_file = gr.File(label="Download .srt File")