Create app.py

app.py

@@ -0,0 +1,439 @@
+import gradio as gr
+import requests
+import base64
+import time
+import json
+
+# ============================================================================
+# REPLACE THESE WITH YOUR MODAL URLS
+# ============================================================================
+MODAL_TRANSCRIBE_URL = "https://vnmalleswar-payyavula--subtitle-generator-test-process-video.modal.run"
+MODAL_BURN_URL = "https://vnmalleswar-payyavula--subtitle-generator-test-burn-edit-099d8b.modal.run"
+
+# ============================================================================
+# STEP 1: TRANSCRIBE VIDEO (Calls Modal Backend)
+# ============================================================================
+
+def transcribe_video(video_file, language="hindi", progress=gr.Progress()):
+    """Call Modal backend to transcribe video"""
+    
+    if not video_file:
+        return None, None, None, "❌ Please upload a video", ""
+    
+    try:
+        progress(0.1, desc="Reading video file...")
+        
+        # Read and encode video
+        with open(video_file, "rb") as f:
+            video_bytes = f.read()
+        
+        video_base64 = base64.b64encode(video_bytes).decode('utf-8')
+        
+        progress(0.2, desc="Uploading to Modal server...")
+        
+        # Prepare request
+        payload = {
+            "video_base64": video_base64,
+            "language": language,
+            "position": "bottom-center"
+        }
+        
+        progress(0.3, desc="Transcribing on GPU server (30-60 seconds)...")
+        
+        # Call Modal transcribe endpoint
+        response = requests.post(
+            MODAL_TRANSCRIBE_URL,
+            json=payload,
+            timeout=600
+        )
+        
+        if response.status_code != 200:
+            return None, None, None, f"❌ Server error: Status {response.status_code}", ""
+        
+        result = response.json()
+        
+        # Check for errors
+        if "error" in result:
+            error_msg = result["error"]
+            if "traceback" in result:
+                error_msg += f"\n\n{result['traceback']}"
+            return None, None, None, f"❌ Error: {error_msg}", ""
+        
+        progress(0.9, desc="Processing results...")
+        
+        # Get results
+        srt_content = result.get("srt_content", "")
+        video_base64_output = result.get("video_base64", "")
+        transcription = result.get("transcription", "")
+        
+        # Save SRT file for download
+        timestamp = int(time.time())
+        srt_path = f"/tmp/subtitles_{timestamp}.srt"
+        with open(srt_path, "w", encoding="utf-8") as f:
+            f.write(srt_content)
+        
+        # Save original video for editing workflow
+        original_video_path = f"/tmp/original_video_{timestamp}.mp4"
+        with open(video_file, "rb") as f:
+            original_bytes = f.read()
+        with open(original_video_path, "wb") as f:
+            f.write(original_bytes)
+        
+        progress(1.0, desc="Complete!")
+        
+        status = result.get("status", "Success")
+        status += f"\n\n📊 **Statistics:**"
+        status += f"\n- Original words: {result.get('original_word_count', 'N/A')}"
+        status += f"\n- Subtitle words: {result.get('subtitle_word_count', 'N/A')}"
+        status += f"\n- Words lost: {result.get('words_lost', 0)}"
+        status += f"\n- Device: {result.get('device_used', 'N/A')}"
+        status += f"\n- Model: {result.get('model_used', 'N/A')}"
+        status += f"\n\n📝 **Transcription Preview:**\n{transcription[:300]}{'...' if len(transcription) > 300 else ''}"
+        
+        return original_video_path, srt_content, srt_path, status, ""
+        
+    except requests.Timeout:
+        return None, None, None, "❌ Timeout: Processing took too long. Try a shorter video.", ""
+    
+    except Exception as e:
+        import traceback
+        error = f"❌ Error: {str(e)}\n\n{traceback.format_exc()}"
+        return None, None, None, error, ""
+
+# ============================================================================
+# STEP 2: BURN EDITED SUBTITLES (Calls Modal Backend)
+# ============================================================================
+
+def burn_edited_subtitles(original_video_path, edited_srt_content, progress=gr.Progress()):
+    """Call Modal backend to burn edited subtitles"""
+    
+    if not original_video_path:
+        return None, "❌ No video found. Please transcribe a video first in Step 1."
+    
+    if not edited_srt_content or not edited_srt_content.strip():
+        return None, "❌ No subtitles provided. Please edit subtitles first."
+    
+    try:
+        progress(0.1, desc="Reading video file...")
+        
+        # Read original video
+        with open(original_video_path, "rb") as f:
+            video_bytes = f.read()
+        
+        video_base64 = base64.b64encode(video_bytes).decode('utf-8')
+        
+        progress(0.2, desc="Parsing edited subtitles...")
+        
+        # Parse edited SRT to subtitles array
+        subtitles = parse_srt_string(edited_srt_content)
+        
+        if not subtitles:
+            return None, "❌ Could not parse subtitles. Please check the format."
+        
+        progress(0.3, desc="Uploading to Modal server...")
+        
+        # Prepare request
+        payload = {
+            "video_base64": video_base64,
+            "subtitles": subtitles,
+            "position": "bottom-center"
+        }
+        
+        progress(0.4, desc="Burning subtitles on server (5-10 seconds)...")
+        
+        # Call Modal burn endpoint
+        response = requests.post(
+            MODAL_BURN_URL,
+            json=payload,
+            timeout=300
+        )
+        
+        if response.status_code != 200:
+            return None, f"❌ Server error: Status {response.status_code}"
+        
+        result = response.json()
+        
+        # Check for errors
+        if "error" in result:
+            error_msg = result["error"]
+            if "traceback" in result:
+                error_msg += f"\n\n{result['traceback']}"
+            return None, f"❌ Error: {error_msg}"
+        
+        progress(0.8, desc="Downloading processed video...")
+        
+        # Get output video
+        video_base64_output = result.get("video_base64", "")
+        
+        # Save final video
+        timestamp = int(time.time())
+        final_video_path = f"/tmp/final_video_{timestamp}.mp4"
+        video_bytes = base64.b64decode(video_base64_output)
+        with open(final_video_path, "wb") as f:
+            f.write(video_bytes)
+        
+        progress(1.0, desc="Complete!")
+        
+        status = result.get("status", "Success")
+        status += f"\n\n✅ Video ready! Burned {len(subtitles)} edited subtitles."
+        
+        return final_video_path, status
+        
+    except requests.Timeout:
+        return None, "❌ Timeout: Processing took too long."
+    
+    except Exception as e:
+        import traceback
+        error = f"❌ Error: {str(e)}\n\n{traceback.format_exc()}"
+        return None, error
+
+# ============================================================================
+# HELPER FUNCTIONS
+# ============================================================================
+
+def parse_srt_string(srt_content):
+    """Parse SRT string to subtitle array for Modal API"""
+    subtitles = []
+    blocks = srt_content.strip().split('\n\n')
+    
+    for block in blocks:
+        lines = block.strip().split('\n')
+        if len(lines) >= 3:
+            try:
+                index = int(lines[0])
+                times = lines[1].split(' --> ')
+                if len(times) == 2:
+                    subtitles.append({
+                        'index': index,
+                        'start': times[0].strip(),
+                        'end': times[1].strip(),
+                        'text': '\n'.join(lines[2:])
+                    })
+            except:
+                continue
+    
+    return subtitles
+
+# ============================================================================
+# GRADIO INTERFACE
+# ============================================================================
+
+css = """
+#subtitle_editor {
+    font-family: 'Courier New', monospace;
+    font-size: 14px;
+}
+.modal-info {
+    background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
+    color: white;
+    padding: 20px;
+    border-radius: 10px;
+    margin: 15px 0;
+    box-shadow: 0 4px 6px rgba(0,0,0,0.1);
+}
+.step-box {
+    border: 2px solid #667eea;
+    border-radius: 10px;
+    padding: 20px;
+    margin: 10px 0;
+}
+"""
+
+with gr.Blocks(title="AI Subtitle Generator & Editor", css=css, theme=gr.themes.Soft()) as demo:
+    
+    # Header
+    gr.Markdown(
+        """
+        # 🎬 AI Subtitle Generator & Editor
+        
+        ### Powered by Whisper Large V3 Turbo on Modal GPU
+        
+        **Two-Step Workflow:**
+        1. 📝 **Transcribe** → Upload video and generate subtitles automatically
+        2. ✏️ **Edit & Burn** → Modify subtitles and get final video (fast, no re-transcription!)
+        """
+    )
+    
+    # Hidden states
+    stored_video = gr.State(None)
+    stored_subtitles_json = gr.State("")
+    
+    gr.Markdown("---")
+    
+    # ========================================================================
+    # STEP 1: TRANSCRIBE
+    # ========================================================================
+    
+    with gr.Group(elem_classes="step-box"):
+        gr.Markdown("## 📝 Step 1: Upload & Transcribe Video")
+        
+        with gr.Row():
+            with gr.Column(scale=1):
+                video_input = gr.Video(label="Upload Video File")
+                language_dropdown = gr.Dropdown(
+                    choices=["hindi", "english", "auto"],
+                    value="hindi",
+                    label="Audio Language",
+                    info="Select the language spoken in the video"
+                )
+                transcribe_btn = gr.Button("🎙️ Transcribe Video", variant="primary", size="lg")
+            
+            with gr.Column(scale=1):
+                transcribe_status = gr.Textbox(
+                    label="Transcription Status", 
+                    lines=12,
+                    interactive=False
+                )
+                srt_download = gr.File(
+                    label="📥 Download SRT File",
+                    interactive=False
+                )
+        
+        gr.Markdown(
+            """
+            <div class="modal-info">
+            ⏱️ <b>Processing Time:</b> 30-60 seconds (depends on video length)<br>
+            🖥️ <b>Processing:</b> Modal GPU Server (T4)<br>
+            🎯 <b>Model:</b> Whisper Large V3 Turbo
+            </div>
+            """
+        )
+    
+    gr.Markdown("---")
+    
+    # ========================================================================
+    # STEP 2: EDIT & BURN
+    # ========================================================================
+    
+    with gr.Group(elem_classes="step-box"):
+        gr.Markdown("## ✏️ Step 2: Edit Subtitles & Burn to Video")
+        gr.Markdown(
+            """
+            **How to Edit:**
+            - Modify subtitle text directly in the box below
+            - Keep the SRT format: `Index` → `Timestamp` → `Text`
+            - Each subtitle separated by blank line
+            - Delete unwanted subtitle blocks entirely
+            """
+        )
+        
+        with gr.Row():
+            with gr.Column(scale=1):
+                srt_editor = gr.Textbox(
+                    label="✏️ Editable Subtitles (SRT Format)",
+                    lines=20,
+                    placeholder="Subtitles will appear here after transcription...\n\nFormat:\n1\n00:00:00,000 --> 00:00:02,000\nSubtitle text\n\n2\n00:00:02,000 --> 00:00:04,000\nNext subtitle...",
+                    elem_id="subtitle_editor"
+                )
+                
+                with gr.Accordion("💡 Editing Tips", open=False):
+                    gr.Markdown(
+                        """
+                        - **Change text:** Just modify the text lines
+                        - **Delete subtitle:** Remove entire block (index + timestamp + text)
+                        - **Adjust timing:** Edit timestamp line (format: `HH:MM:SS,mmm --> HH:MM:SS,mmm`)
+                        - **Add subtitle:** Copy the format and add a new block
+                        - **Re-index:** After deleting, you can renumber indices (1, 2, 3...)
+                        """
+                    )
+            
+            with gr.Column(scale=1):
+                burn_btn = gr.Button("🔥 Burn Edited Subtitles to Video", variant="primary", size="lg")
+                burn_status = gr.Textbox(
+                    label="Burn Status",
+                    lines=5,
+                    interactive=False
+                )
+                final_video = gr.Video(
+                    label="📹 Final Video with Edited Subtitles",
+                    interactive=False
+                )
+        
+        gr.Markdown(
+            """
+            <div class="modal-info">
+            ⏱️ <b>Processing Time:</b> 5-10 seconds (no re-transcription!)<br>
+            🖥️ <b>Processing:</b> Modal GPU Server<br>
+            ⚡ <b>Speed:</b> 5-10x faster than re-transcribing
+            </div>
+            """
+        )
+    
+    gr.Markdown("---")
+    
+    # Footer
+    with gr.Accordion("📚 More Information", open=False):
+        gr.Markdown(
+            """
+            ### 🎯 Features:
+            
+            - ✅ Automatic subtitle generation using Whisper AI
+            - ✅ Word-level timestamp accuracy
+            - ✅ Edit subtitles without re-transcription
+            - ✅ Support for Hindi, English, and auto-detection
+            - ✅ Customizable subtitle appearance
+            - ✅ Fast GPU processing on Modal
+            
+            ### 📊 Technical Details:
+            
+            | Component | Details |
+            |-----------|---------|
+            | Backend | Modal Serverless |
+            | GPU | NVIDIA T4 |
+            | Model | Whisper Large V3 Turbo |
+            | Engine | faster-whisper |
+            | Font | Arial, size 20 |
+            | Position | Bottom center |
+            | Max Words | 3 per subtitle |
+            
+            ### ⚡ Performance Comparison:
+            
+            | Operation | Time |
+            |-----------|------|
+            | Step 1: Transcribe | 30-60 seconds |
+            | Step 2: Burn Only | 5-10 seconds |
+            | **Total (with edit)** | **35-70 seconds** |
+            | Traditional (re-transcribe) | 60-120 seconds |
+            
+            ### 🔧 Setup:
+            
+            This frontend connects to Modal backend endpoints. The backend handles all heavy processing.
+            
+            ### 🐛 Troubleshooting:
+            
+            - **"Server error"**: Check Modal deployment is active
+            - **"Timeout"**: Try a shorter video (under 5 minutes)
+            - **"Parse error"**: Check SRT format in editor
+            """
+        )
+    
+    # ========================================================================
+    # EVENT HANDLERS
+    # ========================================================================
+    
+    # Step 1: Transcribe
+    transcribe_btn.click(
+        fn=transcribe_video,
+        inputs=[video_input, language_dropdown],
+        outputs=[stored_video, srt_editor, srt_download, transcribe_status, stored_subtitles_json]
+    )
+    
+    # Step 2: Burn edited subtitles
+    burn_btn.click(
+        fn=burn_edited_subtitles,
+        inputs=[stored_video, srt_editor],
+        outputs=[final_video, burn_status]
+    )
+
+# ============================================================================
+# LAUNCH
+# ============================================================================
+
+if __name__ == "__main__":
+    demo.queue()
+    demo.launch(
+        share=False,
+        show_error=True,
+        server_name="0.0.0.0",
+        server_port=7860
+    )