allow different orientation

app.py

@@ -46,6 +46,7 @@ from diffusers import FluxPipeline
 import tempfile
 from ovi.utils.io_utils import save_video
 from ovi.utils.processing_utils import clean_text, scale_hw_to_area_divisible
+from gradio_extendedimage import extendedimage
 
 # ----------------------------
 # Parse CLI Args
@@ -126,6 +127,7 @@ def generate_video(
     slg_layer = 11,
     video_negative_prompt = "",
     audio_negative_prompt = "",
+    progress=gr.Progress(track_tqdm=True)
 ):
     try:
         image_path = None
@@ -176,6 +178,19 @@ def generate_image(text_prompt, image_seed, image_height, image_width):
     image.save(tmpfile.name)
     return tmpfile.name
 
+def orientation_changed(evt: gr.EventData):
+
+    detail = getattr(evt, "data", None) or getattr(evt, "_data", {}) or {}
+
+    if detail['value'] == "9:16":
+        orientation_state = [[992, 512]]
+    elif detail['value'] == "1:1":
+        orientation_state = [[992, 992]]
+    elif detail['value'] == "16:9":
+        orientation_state = [[512, 992]]
+
+    return orientation_state
+
 css = """
     #col-container {
         margin: 0 auto;
@@ -193,7 +208,7 @@ css = """
 
 with gr.Blocks(css=css) as demo:
 
-    session_state = gr.State()
+    orientation_state = gr.State([[992, 512]])
 
     with gr.Column(elem_id="col-container"):
         gr.HTML(
@@ -217,7 +232,7 @@ with gr.Blocks(css=css) as demo:
         with gr.Row():
             with gr.Column():
                 # Image section
-                image = gr.Image(type="filepath", label="First Frame Image (upload or generate)")
+                image = extendedimage(type="filepath", label="Image")
 
                 if args.use_image_gen:
                     with gr.Accordion("🖼️ Image Generation Options", visible=True):
@@ -235,7 +250,7 @@ with gr.Blocks(css=css) as demo:
                     value=50,
                     label="Sample Steps",
                     minimum=20,
-                    maximum=100,
+                    maximum=50,
                     step=1.0
                 )
                 run_btn = gr.Button("Generate Video 🚀", variant="primary")
@@ -294,6 +309,8 @@ with gr.Blocks(css=css) as demo:
                     cache_examples=True,
                 )
 
+    image.orientation(fn=orientation_changed, outputs=[orientation_state])
+
     if args.use_image_gen and gen_img_btn is not None:
         gen_img_btn.click(
             fn=generate_image,

requirements.txt

@@ -5,6 +5,7 @@ transformers>=4.49.0,<=4.51.3
 tokenizers>=0.20.3
 accelerate>=1.1.1
 
+gradio_extendedimage @ https://github.com/OutofAi/gradio-extendedimage/releases/download/0.0.2/gradio_extendedimage-0.0.2-py3-none-any.whl
 tqdm
 imageio[ffmpeg]
 easydict