Update app.py

app.py

@@ -33,9 +33,9 @@ pipe.scheduler = EulerDiscreteScheduler.from_config(pipe.scheduler.config, times
 # Safety checkers
 from transformers import CLIPFeatureExtractor
 
-feature_extractor = CLIPFeatureExtractor.from_pretrained("openai/clip-vit-base-patch32")
+feature_extractor = CLIPFeatureExtractor.from_pretrained("openai/clip-vit-base-patch32") # change for open-source model
 
-# Function 
+# Function: we are using Gradio server to queue calls. However this is open for different architectures 
 @spaces.GPU(duration=15,enable_queue=True)
 def generate_image(prompt, base, motion, step, progress=gr.Progress()):
     global step_loaded
@@ -44,8 +44,8 @@ def generate_image(prompt, base, motion, step, progress=gr.Progress()):
     print(prompt, base, step)
 
     if step_loaded != step:
-        repo = "ByteDance/AnimateDiff-Lightning"
-        ckpt = f"animatediff_lightning_{step}step_diffusers.safetensors"
+        repo = "ByteDance/AnimateDiff-Lightning" # we can change to other Diffusion models...
+        ckpt = f"animatediff_lightning_{step}step_diffusers.safetensors" #...but you must change the implementation at this point to match with the checkpoint
         pipe.unet.load_state_dict(load_file(hf_hub_download(repo, ckpt), device=device), strict=False)
         step_loaded = step
 
@@ -61,10 +61,11 @@ def generate_image(prompt, base, motion, step, progress=gr.Progress()):
         motion_loaded = motion
 
     progress((0, step))
+    
     def progress_callback(i, t, z):
         progress((i+1, step))
 
-    output = pipe(prompt=prompt, guidance_scale=1.2, num_inference_steps=step, callback=progress_callback, callback_steps=1)
+    output = pipe(prompt=prompt, guidance_scale=1.2, num_inference_steps=step, callback=progress_callback, callback_steps=1) #providing visibility to progress. Useful if using gradio interface
 
     name = str(uuid.uuid4()).replace("-", "")
     path = f"/tmp/{name}.mp4"
@@ -73,13 +74,15 @@ def generate_image(prompt, base, motion, step, progress=gr.Progress()):
 
 
 # Gradio Interface
-with gr.Blocks(css="style.css") as demo:
+with gr.Blocks(css="style.css", theme='sudeepshouche/minimalist') as syntvideo:
     gr.HTML(
-        "<h1><center>Instant⚡Video</center></h1>" +
-        "<p><center>Lightning-fast text-to-video generation</center></p>" +
-        "<p><center><span style='color: red;'>You may change the steps from 4 to 8, if you didn't get satisfied results.</center></p>" +
-        "<p><center><strong> First Video Generating takes time then Videos generate faster.</p>" +
-        "<p><center>Write prompts in style as Given in Example</p>"
+        "<h1><center>MAGIC Demo: synthetic video generation application</center></h1>" +
+        "<p><center><span style='color: red;'>Change the steps from 4 to 8 to get better results.</center></p>" +
+        "<p><center>Write prompts in style as given in the examples below:</center></p>" +
+        "<p><center>Focus: Group of Birds in sky (Animate:  Birds Moving) (Shot From distance)</center></p>" +
+        "<p><center>Focus: Trees In forest (Animate: Lion running)</center></p>" +
+        "<p><center>Focus: Kids Playing (Season: Winter)</center></p>" +
+        "<p><center>Focus: Cars in Street (Season: Rain, Daytime) (Shot from Distance) (Movement: Cars running)</center></p>"
     )
     with gr.Group():
         with gr.Row():
@@ -130,7 +133,7 @@ with gr.Blocks(css="style.css") as demo:
                 variant='primary'
             )
     video = gr.Video(
-        label='AnimateDiff-Lightning',
+        label='Generate Synthetic Video',
         autoplay=True,
         height=512,
         width=512,
@@ -148,21 +151,4 @@ with gr.Blocks(css="style.css") as demo:
         outputs=video,
     )
 
-    gr.Examples(
-        examples=[
-        ["Focus: Eiffel Tower (Animate: Clouds moving)"], #Atmosphere Movement Example
-        ["Focus: Trees In forest (Animate: Lion running)"], #Object Movement Example
-        ["Focus: Astronaut in Space"], #Normal
-        ["Focus: Group of Birds in sky (Animate:  Birds Moving) (Shot From distance)"], #Camera distance
-        ["Focus:  Statue of liberty (Shot from Drone) (Animate: Drone coming toward statue)"], #Camera Movement
-        ["Focus: Panda in Forest (Animate: Drinking Tea)"], #Doing Something
-        ["Focus: Kids Playing (Season: Winter)"], #Atmosphere or Season
-        {"Focus: Cars in Street (Season: Rain, Daytime) (Shot from Distance) (Movement: Cars running)"} #Mixture
-    ], 
-        fn=generate_image,
-        inputs=[prompt, select_base, select_motion, select_step],
-        outputs=video,
-        cache_examples=False,
-)
-
-demo.queue().launch()
+syntvideo.queue().launch()