Upload app.py

app.py

@@ -280,213 +280,320 @@ def create_interface():
         
 @spaces.GPU(timeout=120)  # 2 minute timeout for ZeroGPU
 def generate_commercial(
-            brand_name: str,
-            structure_text: str,
-            script_text: str,
-            duration_val: int,
-            voice_style: str,
-            music_enabled: bool,
-            video_model: str,
-            audio_model: str,
-            vhs_intensity: float,
-            seed_val: int
-        ) -> Tuple[str, str, str]:
+    brand_name: str,
+    structure_text: str,
+    script_text: str,
+    duration_val: int,
+    voice_style: str,
+    music_enabled: bool,
+    video_model: str,
+    audio_model: str,
+    vhs_intensity: float,
+    seed_val: int
+) -> Tuple[str, str, str]:
+    """
+    Generate a complete retro commercial with perfect sync.
+    """
+    try:
+        # Update progress
+        progress_html = """
+            <div class="progress-info">
+                <strong>🎬 Generating Commercial...</strong><br>
+                <div style="margin-top: 0.5rem;">
+                    <div>📝 Generating script with AI...</div>
+                </div>
+            </div>
+        """
+        yield progress_html, None, None, None
+        
+        # Generate script using LLM
+        generated_script = script_generator.generate_script(
+            brand=brand_name or "Brand",
+            structure=structure_text or "Montage → Close-up → Logo",
+            script_prompt=script_text or "Back to '87",
+            duration=duration_val,
+            voice_style=voice_style,
+            seed=seed_val
+        )
+        
+        # Update progress
+        progress_html = """
+            <div class="progress-info">
+                <strong>🎬 Generating Commercial...</strong><br>
+                <div style="margin-top: 0.5rem;">
+                    <div>✅ Script generated</div>
+                    <div>🎥 Generating video...</div>
+                </div>
+            </div>
+        """
+        yield progress_html, None, None, None
+        
+        # Create temporary directory
+        with tempfile.TemporaryDirectory() as tmpdir:
+            # Generate video
+            video_prompt = f"{structure_text}. {script_text}. 1980s commercial, VHS texture, soft lighting, bold retro titles, 4:3, brand {brand_name}"
+            
+            # Calculate optimal frame count
+            num_frames = min(sync_manager.get_optimal_frame_count(duration_val, DEFAULT_FPS), 16)  # Cap for ZeroGPU
+            
+            clip = synth_t2v(
+                prompt=video_prompt,
+                seed=seed_val,
+                num_frames=num_frames,
+                fps=DEFAULT_FPS,
+                device=DEVICE,
+                model_name=video_model
+            )
+            
+            # Save raw video
+            raw_video_path = os.path.join(tmpdir, "raw.mp4")
+            clip.write_videofile(
+                raw_video_path,
+                fps=DEFAULT_FPS,
+                codec='libx264',
+                audio=False,
+                verbose=False,
+                logger=None
+            )
+            
+            # Apply retro filters
+            retro_video_path = os.path.join(tmpdir, "retro.mp4")
+            apply_retro_filters(raw_video_path, retro_video_path, intensity=vhs_intensity)
+            
+            # Update progress
+            progress_html = """
+                <div class="progress-info">
+                    <strong>🎬 Generating Commercial...</strong><br>
+                    <div style="margin-top: 0.5rem;">
+                        <div>✅ Script generated</div>
+                        <div>✅ Video generated</div>
+                        <div>🎤 Generating audio...</div>
+                    </div>
+                </div>
             """
-            Generate a complete retro commercial with perfect sync.
+            yield progress_html, None, None, None
+            
+            # Generate audio
+            voiceover_text = " ".join([seg.text for seg in generated_script.segments])
+            sr_voice, wav_voice = synth_voice(
+                text=voiceover_text,
+                voice_prompt=voice_style,
+                model_name=audio_model,
+                device=DEVICE
+            )
+            
+            # Add background music if requested
+            if music_enabled:
+                sr_music, wav_music = retro_bed(clip.duration)
+                sr_final, stereo_audio = mix_to_stereo(
+                    sr_voice, wav_voice, sr_music, wav_music, bed_gain=0.3
+                )
+            else:
+                sr_final = sr_voice
+                stereo_audio = np.stack([wav_voice, wav_voice], axis=1)
+            
+            # Save audio
+            audio_path = os.path.join(tmpdir, "audio.wav")
+            write_wav(audio_path, sr_final, stereo_audio)
+            
+            # Update progress
+            progress_html = """
+                <div class="progress-info">
+                    <strong>🎬 Generating Commercial...</strong><br>
+                    <div style="margin-top: 0.5rem;">
+                        <div>✅ Script generated</div>
+                        <div>✅ Video generated</div>
+                        <div>✅ Audio generated</div>
+                        <div>🔄 Synchronizing audio and video...</div>
+                    </div>
+                </div>
             """
+            yield progress_html, None, None, None
+            
+            # Synchronize audio and video
+            # Clean brand name for filename (remove spaces and special characters)
+            clean_brand = "".join(c for c in brand_name if c.isalnum() or c in (' ', '-', '_')).rstrip()
+            clean_brand = clean_brand.replace(' ', '_')
+            if not clean_brand:
+                clean_brand = "commercial"
+            final_video_path = os.path.join(tmpdir, f"{clean_brand}_commercial.mp4")
+            
             try:
-                # Update progress
-                progress_html = """
-                    <div class="progress-info">
-                        <strong>🎬 Generating Commercial...</strong><br>
-                        <div style="margin-top: 0.5rem;">
-                            <div>📝 Generating script with AI...</div>
-                        </div>
+                sync_manager.synchronize_media(
+                    video_path=retro_video_path,
+                    audio_path=audio_path,
+                    output_path=final_video_path,
+                    prefer_audio_duration=True
+                )
+            except Exception as e:
+                logger.error(f"Sync failed, using simple mux: {e}")
+                # Fallback to simple muxing
+                from utils_video import mux_audio
+                mux_audio(retro_video_path, audio_path, final_video_path)
+            
+            # Validate sync and file existence
+            if os.path.exists(final_video_path):
+                is_synced, sync_diff = sync_manager.validate_sync(final_video_path, final_video_path)
+            else:
+                logger.error("Final video file was not created")
+                is_synced, sync_diff = False, float('inf')
+            
+            # Format script output
+            script_lines = []
+            for i, segment in enumerate(generated_script.segments, 1):
+                script_lines.append(f"{i}. {segment.timing_marker} {segment.text}")
+            
+            script_output = "\n".join(script_lines) + f"\n\nTAGLINE: {generated_script.tagline}"
+            
+            # Final progress
+            sync_status = "✅ Perfect sync" if is_synced else f"⚠️ Sync diff: {sync_diff:.3f}s"
+            progress_html = f"""
+                <div class="progress-info">
+                    <strong>🎉 Commercial Complete!</strong><br>
+                    <div style="margin-top: 0.5rem;">
+                        <div>✅ Script generated ({generated_script.word_count} words)</div>
+                        <div>✅ Video generated ({num_frames} frames)</div>
+                        <div>✅ Audio generated ({len(stereo_audio)/sr_final:.1f}s)</div>
+                        <div>{sync_status}</div>
                     </div>
-                """
-                yield progress_html, None, None, None
+                </div>
+            """
+            
+            yield progress_html, final_video_path, script_output, final_video_path
+            
+    except Exception as e:
+        logger.error(f"Commercial generation failed: {e}")
+        logger.error(f"Traceback: {traceback.format_exc()}")
+        error_html = f"""
+            <div class="error-info">
+                <strong>❌ Generation Failed</strong><br>
+                <div style="margin-top: 0.5rem; color: #666;">
+                    Error: {str(e)}<br>
+                    Please try again with different parameters or check the logs.
+                </div>
+            </div>
+        """
+        yield error_html, None, None, None
+
+def create_interface():
+    """Create the Gradio interface."""
+    with gr.Blocks(
+        title="EceMotion Pictures",
+        theme=gr.themes.Soft(),
+        css="""
+            .progress-info { 
+                background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
+                color: white; 
+                padding: 1rem; 
+                border-radius: 10px; 
+                margin: 1rem 0;
+            }
+            .error-info { 
+                background: linear-gradient(135deg, #ff6b6b 0%, #ee5a24 100%);
+                color: white; 
+                padding: 1rem; 
+                border-radius: 10px; 
+                margin: 1rem 0;
+            }
+        """
+    ) as demo:
+        gr.Markdown("""
+        # 🎬 EceMotion Pictures
+        ## AI-Powered 1980s Style Commercial Generator
+        
+        Create authentic retro commercials with perfect audio-video synchronization!
+        """)
+        
+        with gr.Row():
+            with gr.Column(scale=1):
+                gr.Markdown("### Commercial Setup")
                 
-                # Generate script using LLM
-                generated_script = script_generator.generate_script(
-                    brand=brand_name or "Brand",
-                    structure=structure_text or "Montage → Close-up → Logo",
-                    script_prompt=script_text or "Back to '87",
-                    duration=duration_val,
-                    voice_style=voice_style,
-                    seed=seed_val
+                brand = gr.Textbox(
+                    label="Brand Name",
+                    placeholder="Enter your brand name...",
+                    value="EceMotion Pictures"
                 )
                 
-                # Update progress
-                progress_html = """
-                    <div class="progress-info">
-                        <strong>🎬 Generating Commercial...</strong><br>
-                        <div style="margin-top: 0.5rem;">
-                            <div>✅ Script generated</div>
-                            <div>🎥 Generating video...</div>
-                        </div>
-                    </div>
-                """
-                yield progress_html, None, None, None
+                structure = gr.Dropdown(
+                    label="Commercial Structure",
+                    choices=[
+                        "Montage → Close-up → Logo stinger",
+                        "Problem → Solution → Call to action",
+                        "Story → Product reveal → Tagline",
+                        "Before/After → Benefits → Brand"
+                    ],
+                    value="Montage → Close-up → Logo stinger"
+                )
                 
-                # Create temporary directory
-                with tempfile.TemporaryDirectory() as tmpdir:
-                    # Generate video
-                    video_prompt = f"{structure_text}. {script_text}. 1980s commercial, VHS texture, soft lighting, bold retro titles, 4:3, brand {brand_name}"
-                    
-                    # Calculate optimal frame count
-                    num_frames = min(sync_manager.get_optimal_frame_count(duration_val, DEFAULT_FPS), 16)  # Cap for ZeroGPU
-                    
-                    clip = synth_t2v(
-                        prompt=video_prompt,
-                        seed=seed_val,
-                        num_frames=num_frames,
-                        fps=DEFAULT_FPS,
-                        device=DEVICE,
-                        model_name=video_model
+                script_prompt = gr.Textbox(
+                    label="Script Hook",
+                    placeholder="Enter your script prompt...",
+                    value="Remember when technology was simple?"
+                )
+                
+                duration = gr.Slider(
+                    label="Duration (seconds)",
+                    minimum=5,
+                    maximum=15,
+                    value=10,
+                    step=1
+                )
+                
+                with gr.Row():
+                    voice = gr.Dropdown(
+                        label="Voice Style",
+                        choices=["Announcer '80s", "Friendly '80s", "Dramatic '80s", "Upbeat '80s"],
+                        value="Announcer '80s"
                     )
-                    
-                    # Save raw video
-                    raw_video_path = os.path.join(tmpdir, "raw.mp4")
-                    clip.write_videofile(
-                        raw_video_path,
-                        fps=DEFAULT_FPS,
-                        codec='libx264',
-                        audio=False,
-                        verbose=False,
-                        logger=None
+                    music = gr.Checkbox(label="Background Music", value=True)
+                
+                with gr.Accordion("Advanced Settings", open=False):
+                    model_video = gr.Dropdown(
+                        label="Video Model",
+                        choices=["damo-vilab/text-to-video-ms-1.7b", "THUDM/CogVideoX-5b"],
+                        value="damo-vilab/text-to-video-ms-1.7b"
                     )
                     
-                    # Apply retro filters
-                    retro_video_path = os.path.join(tmpdir, "retro.mp4")
-                    apply_retro_filters(raw_video_path, retro_video_path, intensity=vhs_intensity)
-                    
-                    # Update progress
-                    progress_html = """
-                        <div class="progress-info">
-                            <strong>🎬 Generating Commercial...</strong><br>
-                            <div style="margin-top: 0.5rem;">
-                                <div>✅ Script generated</div>
-                                <div>✅ Video generated</div>
-                                <div>🎤 Generating audio...</div>
-                            </div>
-                        </div>
-                    """
-                    yield progress_html, None, None, None
-                    
-                    # Generate audio
-                    voiceover_text = " ".join([seg.text for seg in generated_script.segments])
-                    sr_voice, wav_voice = synth_voice(
-                        text=voiceover_text,
-                        voice_prompt=voice_style,
-                        model_name=audio_model,
-                        device=DEVICE
+                    model_audio = gr.Dropdown(
+                        label="Audio Model",
+                        choices=["parler-tts/parler-tts-mini-v1", "SWivid/F5-TTS"],
+                        value="parler-tts/parler-tts-mini-v1"
                     )
                     
-                    # Add background music if requested
-                    if music_enabled:
-                        sr_music, wav_music = retro_bed(clip.duration)
-                        sr_final, stereo_audio = mix_to_stereo(
-                            sr_voice, wav_voice, sr_music, wav_music, bed_gain=0.3
-                        )
-                    else:
-                        sr_final = sr_voice
-                        stereo_audio = np.stack([wav_voice, wav_voice], axis=1)
-                    
-                    # Save audio
-                    audio_path = os.path.join(tmpdir, "audio.wav")
-                    write_wav(audio_path, sr_final, stereo_audio)
-                    
-                    # Update progress
-                    progress_html = """
-                        <div class="progress-info">
-                            <strong>🎬 Generating Commercial...</strong><br>
-                            <div style="margin-top: 0.5rem;">
-                                <div>✅ Script generated</div>
-                                <div>✅ Video generated</div>
-                                <div>✅ Audio generated</div>
-                                <div>🔄 Synchronizing audio and video...</div>
-                            </div>
-                        </div>
-                    """
-                    yield progress_html, None, None, None
-                    
-                    # Synchronize audio and video
-                    # Clean brand name for filename (remove spaces and special characters)
-                    clean_brand = "".join(c for c in brand_name if c.isalnum() or c in (' ', '-', '_')).rstrip()
-                    clean_brand = clean_brand.replace(' ', '_')
-                    if not clean_brand:
-                        clean_brand = "commercial"
-                    final_video_path = os.path.join(tmpdir, f"{clean_brand}_commercial.mp4")
-                    
-                    try:
-                        sync_manager.synchronize_media(
-                            video_path=retro_video_path,
-                            audio_path=audio_path,
-                            output_path=final_video_path,
-                            prefer_audio_duration=True
-                        )
-                    except Exception as e:
-                        logger.error(f"Sync failed, using simple mux: {e}")
-                        # Fallback to simple muxing
-                        from utils_video import mux_audio
-                        mux_audio(retro_video_path, audio_path, final_video_path)
-                    
-                    # Validate sync and file existence
-                    if os.path.exists(final_video_path):
-                        is_synced, sync_diff = sync_manager.validate_sync(final_video_path, final_video_path)
-                    else:
-                        logger.error("Final video file was not created")
-                        is_synced, sync_diff = False, float('inf')
-                    
-                    # Format script output
-                    script_lines = []
-                    for i, segment in enumerate(generated_script.segments, 1):
-                        script_lines.append(f"{i}. {segment.timing_marker} {segment.text}")
-                    
-                    script_output = "\n".join(script_lines) + f"\n\nTAGLINE: {generated_script.tagline}"
-                    
-                    # Final progress
-                    sync_status = "✅ Perfect sync" if is_synced else f"⚠️ Sync diff: {sync_diff:.3f}s"
-                    progress_html = f"""
-                        <div class="progress-info">
-                            <strong>🎉 Commercial Complete!</strong><br>
-                            <div style="margin-top: 0.5rem;">
-                                <div>✅ Script generated ({generated_script.word_count} words)</div>
-                                <div>✅ Video generated ({num_frames} frames)</div>
-                                <div>✅ Audio generated ({len(stereo_audio)/sr_final:.1f}s)</div>
-                                <div>{sync_status}</div>
-                            </div>
-                        </div>
-                    """
-                    
-                    yield progress_html, final_video_path, script_output, final_video_path
+                    vhs_intensity = gr.Slider(
+                        label="VHS Effect Intensity",
+                        minimum=0.0,
+                        maximum=1.0,
+                        value=0.5,
+                        step=0.1
+                    )
                     
-            except Exception as e:
-                logger.error(f"Commercial generation failed: {e}")
-                logger.error(f"Traceback: {traceback.format_exc()}")
-                error_html = f"""
-                    <div class="error-info">
-                        <strong>❌ Generation Failed</strong><br>
-                        <div style="margin-top: 0.5rem; color: #666;">
-                            Error: {str(e)}<br>
-                            Please try again with different parameters or check the logs.
-                        </div>
-                    </div>
-                """
-                yield error_html, None, None, None
+                    seed = gr.Number(label="Random Seed", value=42, precision=0)
+                
+                with gr.Row():
+                    roll_btn = gr.Button("🎲 Roll Script", variant="secondary")
+                    generate_btn = gr.Button("🎬 Generate Commercial", variant="primary")
+            
+            with gr.Column(scale=2):
+                gr.Markdown("### Generated Commercial")
+                
+                progress_info = gr.HTML()
+                output_video = gr.Video(label="Commercial Preview")
+                output_script = gr.Textbox(label="Generated Script", lines=10)
+                download_btn = gr.DownloadButton("📥 Download Commercial", visible=False)
         
-    # Connect event handlers
-    roll_btn.click(
-        roll_script_suggestion,
-        inputs=[structure, seed],
-        outputs=[script_prompt]
-    )
-    
-    generate_btn.click(
-        generate_commercial,
-        inputs=[
-            brand, structure, script_prompt, duration, voice, music,
-            model_video, model_audio, vhs_intensity, seed
+        # Connect event handlers
+        roll_btn.click(
+            roll_script_suggestion,
+            inputs=[structure, seed],
+            outputs=[script_prompt]
+        )
+        
+        generate_btn.click(
+            generate_commercial,
+            inputs=[
+                brand, structure, script_prompt, duration, voice, music,
+                model_video, model_audio, vhs_intensity, seed
             ],
             outputs=[progress_info, output_video, output_script, download_btn]
         )