Update app.py

app.py

@@ -10,11 +10,26 @@ import soundfile as sf
 from dotenv import load_dotenv
 from transformers import AutoProcessor, AutoModel
 import torch
+import tempfile
 
 # Load environment variables
 load_dotenv()
 HF_TOKEN = os.getenv("API_KEY")
 
+def cleanup_temp_files():
+    temp_files = [
+        os.path.join(tempfile.gettempdir(), 'speech_output.flac'),
+        os.path.join(tempfile.gettempdir(), 'audio.mp3'),
+        os.path.join(tempfile.gettempdir(), 'my_vid_tmp.mp4'),
+        os.path.join(tempfile.gettempdir(), 'mergedvideo.mp4')
+    ]
+    for file in temp_files:
+        if os.path.exists(file):
+            try:
+                os.remove(file)
+            except:
+                pass
+
 def resize(img_list):
     resize_img_list = []
     for item in img_list:
@@ -24,68 +39,103 @@ def resize(img_list):
     return resize_img_list
 
 def text2speech(text):
-    # Using Microsoft's SpeechT5 model instead of FastSpeech2
-    processor = AutoProcessor.from_pretrained("microsoft/speecht5_tts")
-    model = AutoModel.from_pretrained("microsoft/speecht5_tts")
-    
-    # Preprocessing text input
-    inputs = processor(text=text, return_tensors="pt")
-    
-    # Generate speech with default speaker embedding
-    speaker_embeddings = torch.zeros((1, model.config.speaker_embedding_size))
-    speech = model.generate_speech(inputs["input_ids"], speaker_embeddings)
-    
-    # Save as flac file
-    sf.write("speech_output.flac", speech.numpy(), samplerate=16000)
-    return "speech_output.flac"
+    try:
+        processor = AutoProcessor.from_pretrained("microsoft/speecht5_tts")
+        model = AutoModel.from_pretrained("microsoft/speecht5_tts")
+        
+        inputs = processor(text=text, return_tensors="pt")
+        speaker_embeddings = torch.zeros((1, model.config.speaker_embedding_size))
+        speech = model.generate_speech(inputs["input_ids"], speaker_embeddings)
+        
+        output_path = os.path.join(tempfile.gettempdir(), "speech_output.flac")
+        sf.write(output_path, speech.numpy(), samplerate=16000)
+        return output_path
+    except Exception as e:
+        print(f"Error in text2speech: {str(e)}")
+        raise
 
 def merge_audio_video(entities_num, resize_img_list, text_input):
-    speech = text2speech(text_input)
-    wav_audio = AudioSegment.from_file(speech, "flac")
-    wav_audio.export("audio.mp3", format="mp3")
-    audio_length = int(MP3("audio.mp3").info.length)
-    fps = entities_num / audio_length  
-    fps = float(format(fps, '.5f'))
-    
-    clip = ImageSequenceClip(resize_img_list, fps=fps)
-    clip.write_videofile('my_vid_tmp.mp4')
-    
-    videoclip = VideoFileClip('my_vid_tmp.mp4')
-    audioclip = AudioFileClip('audio.mp3')
-    mergedclip = videoclip.set_audio(audioclip)
-    
-    return mergedclip
+    try:
+        speech = text2speech(text_input)
+        wav_audio = AudioSegment.from_file(speech, "flac")
+        
+        audio_path = os.path.join(tempfile.gettempdir(), "audio.mp3")
+        wav_audio.export(audio_path, format="mp3")
+        
+        audio_length = int(MP3(audio_path).info.length)
+        fps = max(entities_num / audio_length, 1)  # Ensure fps is at least 1
+        fps = float(format(fps, '.5f'))
+        
+        temp_video = os.path.join(tempfile.gettempdir(), "my_vid_tmp.mp4")
+        clip = ImageSequenceClip(resize_img_list, fps=fps)
+        clip.write_videofile(temp_video, codec='libx264', fps=fps)
+        
+        videoclip = VideoFileClip(temp_video)
+        audioclip = AudioFileClip(audio_path)
+        mergedclip = videoclip.set_audio(audioclip)
+        
+        output_path = os.path.join(tempfile.gettempdir(), "mergedvideo.mp4")
+        mergedclip.write_videofile(output_path)
+        
+        # Clean up clips
+        videoclip.close()
+        audioclip.close()
+        mergedclip.close()
+        
+        return output_path
+    except Exception as e:
+        print(f"Error in merge_audio_video: {str(e)}")
+        raise
+    finally:
+        cleanup_temp_files()
 
-with gr.Blocks() as app:
-    # Load models in Blocks context
-    ner = gr.Interface.load("huggingface/flair/ner-english-ontonotes-large")
-    latentdiffusion = gr.Interface.load("spaces/multimodalart/latentdiffusion")
+# Load models outside the Blocks context
+ner = gr.load("huggingface/flair/ner-english-ontonotes-large")
+latentdiffusion = gr.load("spaces/multimodalart/latentdiffusion")
 
-    def engine(text_input):
+def engine(text_input):
+    try:
         entities = ner(text_input)
         entities = [tupl for tupl in entities if None not in tupl]
         entities_num = len(entities)
-        img_list = []
         
+        if entities_num == 0:
+            raise ValueError("No entities found in the input text")
+        
+        img_list = []
         for ent in entities:
             img = latentdiffusion(ent[0], '50', '256', '256', '1', 10)[0]
             img_list.append(img)
         
         resize_img_list = resize(img_list)
-        mergedclip = merge_audio_video(entities_num, resize_img_list, text_input)
-        mergedclip.write_videofile('mergedvideo.mp4')
+        output_path = merge_audio_video(entities_num, resize_img_list, text_input)
         
-        return 'mergedvideo.mp4'
+        return output_path
+    except Exception as e:
+        print(f"Error in engine: {str(e)}")
+        raise gr.Error(f"An error occurred: {str(e)}")
+    finally:
+        cleanup_temp_files()
 
-    interface = gr.Interface(
-        fn=engine,
-        inputs=gr.Textbox(lines=5, label="Input Text"),
-        outputs=gr.Video(label='Final Merged Video'),
-        description="<div>🎭🎞️🍿 AI Movie Maker - Comedy 🎬 🧠 🎨</div>",
+with gr.Blocks() as app:
+    gr.Markdown("# AI Pipeline Multi Model 🎭🎞️🍿 Movie Maker 🎬 🧠 🎨")
+    gr.Markdown("<div>🎭🎞️🍿 AI Movie Maker - Comedy 🎬 🧠 🎨</div>")
+    text_input = gr.Textbox(lines=5, label="Input Text")
+    output_video = gr.Video(label='Final Merged Video')
+    examples = gr.Examples(
         examples=[
             ["Two space marines take up arms to save the planet from an alien invasion. These two dashing strong men play a comedic role in the science fiction movie of the future where even Barnaby bunny is willing to join their wacky gang of space marines to save the planet with good looks and comedy."]
         ],
-        title="AI Pipeline Multi Model 🎭🎞️🍿 Movie Maker 🎬 🧠 🎨",
-        article="<br><div></div>"
+        inputs=text_input
     )
-    interface.launch(debug=True)
+    submit_button = gr.Button("Generate Video")
+    submit_button.click(fn=engine, inputs=text_input, outputs=output_video)
+
+    gr.Markdown("<br><div></div>")
+
+app.launch(
+    debug=True,
+    share=True,  # Enable sharing
+    server_name="0.0.0.0",  # Listen on all interfaces
+    server_port=7860  # Specify port
+)