Update app.py

app.py

@@ -15,12 +15,11 @@ model = AutoModelForCausalLM.from_pretrained("gpt2")
 # Initialize TTS model
 tts = TTS(model_name="tts_models/en/ljspeech/tacotron2-DDC")
 
-# Initialize Stable Diffusion pipeline
-pipe = StableDiffusionPipeline.from_pretrained("CompVis/stable-diffusion-v1-4", torch_dtype=torch.float16)
+# Initialize Stable Diffusion pipeline for CPU
+pipe = StableDiffusionPipeline.from_pretrained("CompVis/stable-diffusion-v1-4", torch_dtype=torch.float32)
 pipe.scheduler = DPMSolverMultistepScheduler.from_config(pipe.scheduler.config)
 pipe = pipe.to("cpu")
 
-
 def generate_text(prompt, max_length=200):
     input_ids = tokenizer.encode(prompt, return_tensors="pt")
     attention_mask = torch.ones_like(input_ids)
@@ -32,16 +31,16 @@ def generate_speech(text):
     tts.tts_to_file(text=text, file_path=output_path)
     return output_path
 
-def generate_video_frames(prompt, num_frames=30):
+def generate_video_frames(prompt, num_frames=10):
     frames = []
     for i in range(num_frames):
-        # Add some variation to the prompt for each frame
         frame_prompt = f"{prompt}, frame {i+1} of {num_frames}"
-        image = pipe(frame_prompt).images[0]
+        with torch.no_grad():
+            image = pipe(frame_prompt, num_inference_steps=20).images[0]
         frames.append(np.array(image))
     return frames
 
-def create_video_from_frames(frames, output_path="output_video.mp4", fps=10):
+def create_video_from_frames(frames, output_path="output_video.mp4", fps=5):
     frames_tensor = torch.from_numpy(np.array(frames)).permute(0, 3, 1, 2)
     write_video(output_path, frames_tensor, fps=fps)
     return output_path
@@ -64,7 +63,7 @@ def generate_kids_music_animation(theme):
 
 # Gradio Interface
 with gr.Blocks() as app:
-    gr.Markdown("## AI-Generated Video and Audio Content")
+    gr.Markdown("## AI-Generated Video and Audio Content (CPU Version)")
     
     with gr.Tab("Comedy Animation"):
         comedy_prompt = gr.Textbox(label="Enter comedy prompt")