Update app.py

app.py

@@ -1,41 +1,110 @@
 import gradio as gr
-import torch
 import numpy as np
 import modin.pandas as pd
+import torch
+import os
+from glob import glob
+from pathlib import Path
+from typing import Optional
+from diffusers import LTXImageToVideoPipeline
+from diffusers.utils import load_image, export_to_video
 from PIL import Image
-from diffusers import DiffusionPipeline, StableDiffusion3Pipeline
+
+import uuid
+import random
 from huggingface_hub import hf_hub_download
 
 device = 'cuda' if torch.cuda.is_available() else 'cpu'
 torch.cuda.max_memory_allocated(device=device)
 torch.cuda.empty_cache()
-pipe = DiffusionPipeline.from_pretrained("circulus/canvers-fusionXL-v1").to(device)
-#pipe.enable_xformers_memory_efficient_attention()
+pipe = LTXImageToVideoPipeline.from_pretrained("Lightricks/LTX-Video", torch_dtype=torch.bfloat16)
+pipe.to(device)
 torch.cuda.empty_cache()
+max_64_bit_int = 2**63 - 1
 
-torch.cuda.max_memory_allocated(device=device)
-refiner = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-xl-refiner-1.0").to(device)
-#refiner.enable_xformers_memory_efficient_attention()
-torch.cuda.empty_cache()
+def sample(
+    image: Image, prompt, negative_prompt,
+    seed: Optional[int] = 42,
+    randomize_seed: bool = True,
+    num_inference_steps: int = 25,
+    num_frames: int = 60,
+    fps_id: int = 12,
+    decode_noise_scale: float = 0.3,
+    max_sequence_length: int = 512,
+    decoding_t: int = 3,  # Number of frames decoded at a time! This eats most VRAM. Reduce if necessary.
+    device: str = "cuda",
+    output_folder: str = "outputs",
+):
+    if image.mode == "RGBA":
+        image = image.convert("RGB")
+        
+    if(randomize_seed):
+        seed = random.randint(0, max_64_bit_int)
+    torch.cuda.max_memory_allocated(device=device)
+    torch.cuda.empty_cache()
+    generator = torch.manual_seed(seed)
+    
+    os.makedirs(output_folder, exist_ok=True)
+    base_count = len(glob(os.path.join(output_folder, "*.mp4")))
+    video_path = os.path.join(output_folder, f"{base_count:06d}.mp4")
+    video = pipe(image=image, prompt=prompt, negative_prompt=negative_prompt, width=768, height=512, num_frames=num_frames, num_inference_steps=num_inference_steps, guidance_scale=3.5, decode_noise_scale=0.3, max_sequence_length=512).frames[0]
+    export_to_video(video, video_path, fps=fps_id)
+    torch.manual_seed(seed)
+    torch.cuda.empty_cache()
+    return video_path, seed
+
+def resize_image(image, output_size=(768, 512)):
+    # Calculate aspect ratios
+    target_aspect = output_size[0] / output_size[1]  # Aspect ratio of the desired size
+    image_aspect = image.width / image.height  # Aspect ratio of the original image
 
+    # Resize then crop if the original image is larger
+    if image_aspect > target_aspect:
+        # Resize the image to match the target height, maintaining aspect ratio
+        new_height = output_size[1]
+        new_width = int(new_height * image_aspect)
+        resized_image = image.resize((new_width, new_height), Image.LANCZOS)
+        # Calculate coordinates for cropping
+        left = (new_width - output_size[0]) / 2
+        top = 0
+        right = (new_width + output_size[0]) / 2
+        bottom = output_size[1]
+    else:
+        # Resize the image to match the target width, maintaining aspect ratio
+        new_width = output_size[0]
+        new_height = int(new_width / image_aspect)
+        resized_image = image.resize((new_width, new_height), Image.LANCZOS)
+        # Calculate coordinates for cropping
+        left = 0
+        top = (new_height - output_size[1]) / 2
+        right = output_size[0]
+        bottom = (new_height + output_size[1]) / 2
 
-def genie (Prompt, negative_prompt, height, width, scale, steps, seed):
-    generator = np.random.seed(0) if seed == 0 else torch.manual_seed(seed)
-    #generator=np.random.seed(0)
-    int_image = pipe(Prompt, negative_prompt=negative_prompt, height=height, width=width, num_inference_steps=steps, guidance_scale=scale, output_type="latent").images
-    image = refiner(Prompt, negative_prompt=negative_prompt, image=int_image, denoising_start=.99).images[0]
     torch.cuda.empty_cache()
+    cropped_image = resized_image.crop((left, top, right, bottom))
+    return cropped_image
+
+with gr.Blocks() as demo:
+  
+  with gr.Row():
+    with gr.Column():
+        image = gr.Image(label="Upload your image", type="pil")
+        generate_btn = gr.Button("Generate")
         
-    return image
-    
-gr.Interface(fn=genie, inputs=[gr.Textbox(label='What you want the AI to generate. 77 Token Limit.'), 
-                               gr.Textbox(label='What you Do Not want the AI to generate. 77 Token Limit'),
-                               gr.Slider(512, 1536, 1024, step=128, label='Height'),
-                               gr.Slider(512, 1536, 1024, step=128, label='Width'),
-                               gr.Slider(.5, maximum=15, value=7, step=.25, label='Guidance Scale'), 
-                               gr.Slider(10, maximum=50, value=25, step=5, label='Number of Prior Iterations'),
-                               gr.Slider(minimum=0, step=1, maximum=9999999999999999, randomize=True, label='Seed: 0 is Random')],
-             outputs=gr.Image(label='Generated Image'), 
-             title="Manju Dream Booth V2.5 with Fusion XL - GPU", 
-             description="<br><br><b/>Warning: This Demo is capable of producing NSFW content.", 
-             article = "If You Enjoyed this Demo and would like to Donate, you can send any amount to any of these Wallets. <br><br>SHIB (BEP20): 0xbE8f2f3B71DFEB84E5F7E3aae1909d60658aB891 <br>PayPal: https://www.paypal.me/ManjushriBodhisattva <br>ETH: 0xbE8f2f3B71DFEB84E5F7E3aae1909d60658aB891 <br>DOGE: D9QdVPtcU1EFH8jDC8jhU9uBcSTqUiA8h6<br><br>Code Monkey: <a href=\"https://huggingface.co/Manjushri\">Manjushri</a>").launch(debug=True)
+    video = gr.Video()
+  with gr.Accordion("Advanced options", open=False):
+      seed = gr.Slider(label="Seed", value=42, randomize=True, minimum=0, maximum=max_64_bit_int, step=1)
+      randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
+      prompt=gr.Textbox(label="Prompt")
+      negative_prompt=gr.Textbox(label="Negative Prompt", value="Blur, Motion Blur, Artifacts, Motion Artifacts, Poor Quality, Low Resolution, Smudging, Streaks, Motion Streaks, Pixels, Pixelated, Ugly, Mutation, Mutated")
+      num_inference_steps=gr.Slider(label="Number of Inference Steps", value=25, minimum=25, maximum=100, step=25)
+      num_frames=gr.Slider(label = "Number of Frames", value=60, minimum=30, maximum=120)
+      fps_id = gr.Slider(label="Frames per second", info="The length of your video in seconds will be frames/fps", value=12, minimum=5, maximum=30, step=1)
+      
+  image.upload(fn=resize_image, inputs=image, outputs=image, queue=False)
+  generate_btn.click(fn=sample, inputs=[image, prompt, negative_prompt, seed, randomize_seed, num_inference_steps, num_frames, fps_id], outputs=[video, seed], api_name="video")
+  
+
+if __name__ == "__main__":
+    demo.queue(max_size=20, api_open=False)
+    demo.launch(show_api=False)