Update app.py

app.py

@@ -1,64 +1,57 @@
 import gradio as gr
 import whisper
+import moviepy.editor as mp
 import tempfile
-from moviepy.editor import VideoFileClip, TextClip, CompositeVideoClip
 import os
 
-def generate_auto_subtitles(video):
-    if video is None:
-        return "Please upload a video first!"
+# Load Whisper (base or small model for performance)
+model = whisper.load_model("small")
 
-    # Temporary save video
-    temp_video = tempfile.NamedTemporaryFile(delete=False, suffix=".mp4")
-    temp_video.write(video.read())
-    temp_video.close()
+def generate_subtitles(video):
+    try:
+        # Save temp video file
+        with tempfile.NamedTemporaryFile(delete=False, suffix=".mp4") as tmp:
+            tmp.write(video.read())
+            tmp_path = tmp.name
 
-    # Load Whisper model
-    model = whisper.load_model("small")
+        # Extract audio
+        clip = mp.VideoFileClip(tmp_path)
+        audio_path = tmp_path.replace(".mp4", ".wav")
+        clip.audio.write_audiofile(audio_path, logger=None)
 
-    # Auto language detection + transcription
-    result = model.transcribe(temp_video.name)
-    detected_lang = result.get("language", "unknown").upper()
+        # Transcribe using Whisper
+        result = model.transcribe(audio_path, task="transcribe")
+        text = result["text"]
+        lang = result["language"]
 
-    # Prepare video
-    video_clip = VideoFileClip(temp_video.name)
-    subtitle_clips = []
+        # Add green background
+        w, h = clip.size
+        bg = mp.ColorClip(size=(w, h), color=(0, 255, 0), duration=clip.duration)
+        final = mp.CompositeVideoClip([bg, clip.set_position("center")])
+        output_path = tmp_path.replace(".mp4", "_subtitled.mp4")
+        final.write_videofile(output_path, codec="libx264", audio_codec="aac", logger=None)
 
-    # Generate subtitle overlays
-    for seg in result["segments"]:
-        txt = seg["text"].strip()
-        if not txt:
-            continue
+        return output_path, lang, text
 
-        subtitle = TextClip(
-            txt,
-            fontsize=60,
-            font="DejaVu-Sans-Bold",
-            color="white",
-            bg_color="green",
-            method="caption",
-            size=(video_clip.w, None)
-        )
-        subtitle = subtitle.set_start(seg["start"]).set_duration(seg["end"] - seg["start"]).set_pos(("center", "bottom"))
-        subtitle_clips.append(subtitle)
+    except Exception as e:
+        return None, "Error", str(e)
 
-    # Combine subtitles with video
-    final = CompositeVideoClip([video_clip, *subtitle_clips])
-    output_path = os.path.join(tempfile.gettempdir(), "vozo_auto_subs.mp4")
-    final.write_videofile(output_path, codec="libx264", audio_codec="aac", verbose=False, logger=None)
+# Gradio UI
+with gr.Blocks(title="Auto Subtitle Generator (VOZO-style)") as app:
+    gr.Markdown("## 🎬 Auto Subtitle Generator (VOZO-style)")
+    gr.Markdown("Upload a short video in Hindi, English, Hinglish, or Indian English — the AI auto-detects the language and adds VOZO-style subtitles with a green screen background.")
 
-    return output_path, f"Detected Language: {detected_lang}"
+    with gr.Row():
+        video_input = gr.Video(label="🎥 Upload your video")
+        video_output = gr.Video(label="📺 Subtitled Output")
 
-# Gradio UI
-interface = gr.Interface(
-    fn=generate_auto_subtitles,
-    inputs=gr.Video(label="🎥 Upload your video (Hindi / English / Hinglish)"),
-    outputs=[
-        gr.Video(label="🎬 Video with Auto Subtitles"),
-        gr.Textbox(label="🌐 Detected Language")
-    ],
-    title="🎞️ Auto Subtitle Generator (VOZO-style)",
-    description="Upload any short Hindi, English, Hinglish, or Indian English video — this AI auto-detects language and adds VOZO-style subtitles with a green background."
-)
+    with gr.Row():
+        detected_lang = gr.Textbox(label="Detected Language")
+        subtitles_text = gr.Textbox(label="Generated Subtitles")
+
+    run_btn = gr.Button("🚀 Generate")
+
+    run_btn.click(fn=generate_subtitles, inputs=[video_input],
+                  outputs=[video_output, detected_lang, subtitles_text])
 
-interface.launch()
+app.launch()