Update app.py

app.py

@@ -32,7 +32,7 @@ from src.audio2vid import get_headpose_temp, smooth_pose_seq
 from src.utils.frame_interpolation import init_frame_interpolation_model, batch_images_interpolation_tool
 
 
-config = OmegaConf.load('./configs/prompts/animation_audio.yaml')
+config = OmegaConf.load('/content/AniPortrait_official/configs/prompts/animation_audio.yaml')
 if config.weight_dtype == "fp16":
     weight_dtype = torch.float16
 else:
@@ -167,7 +167,7 @@ def audio2video(input_audio, ref_img, headpose_video=None, size=512, steps=25, l
     #     [transforms.Resize((height, width)), transforms.ToTensor()]
     # )
     args_L = len(pose_images) if length==0 or length > len(pose_images) else length
-    args_L = min(args_L, 90)
+    #args_L = min(args_L, 9999)
     for pose_image_np in pose_images[: args_L : fi_step]:
         # pose_image_pil = Image.fromarray(cv2.cvtColor(pose_image_np, cv2.COLOR_BGR2RGB))
         # pose_tensor_list.append(pose_transform(pose_image_pil))
@@ -263,7 +263,7 @@ def video2video(ref_img, source_video, size=512, steps=25, length=60, seed=42):
     verts_list = []
     bs_list = []
     args_L = len(source_images) if length==0 or length*step > len(source_images) else length*step
-    args_L = min(args_L, 90*step)
+    #args_L = min(args_L, 90*step)
     for src_image_pil in source_images[: args_L : step*fi_step]:
         src_img_np = cv2.cvtColor(np.array(src_image_pil), cv2.COLOR_RGB2BGR)
         frame_height, frame_width, _ = src_img_np.shape
@@ -389,20 +389,20 @@ with gr.Blocks() as demo:
                     a2v_step_slider = gr.Slider(minimum=5, maximum=20, step=1, value=15, label="Steps (--steps)")
                 
                 with gr.Row():
-                    a2v_length = gr.Slider(minimum=0, maximum=90, step=1, value=30, label="Length (-L)")
+                    a2v_length = gr.Slider(minimum=0, maximum=9999, step=1, value=30, label="Length (-L) (Set to 0 to automatically calculate length)")
                     a2v_seed = gr.Number(value=42, label="Seed (--seed)")
                 
                 a2v_botton = gr.Button("Generate", variant="primary")
             a2v_output_video = gr.PlayableVideo(label="Result", interactive=False)
         
-        gr.Examples(
-            examples=[
-                ["configs/inference/audio/lyl.wav", "configs/inference/ref_images/Aragaki.png", None],
-                ["configs/inference/audio/lyl.wav", "configs/inference/ref_images/solo.png", None],
-                ["configs/inference/audio/lyl.wav", "configs/inference/ref_images/lyl.png", "configs/inference/head_pose_temp/pose_ref_video.mp4"],
-                ],
-            inputs=[a2v_input_audio, a2v_ref_img, a2v_headpose_video],
-        )
+        #gr.Examples(
+            #examples=[
+            #    ["configs/inference/audio/lyl.wav", "configs/inference/ref_images/Aragaki.png", None],
+            #    ["configs/inference/audio/lyl.wav", "configs/inference/ref_images/solo.png", None],
+            #    ["configs/inference/audio/lyl.wav", "configs/inference/ref_images/lyl.png", "configs/inference/head_pose_temp/pose_ref_video.mp4"],
+            #    ],
+           # inputs=[a2v_input_audio, a2v_ref_img, a2v_headpose_video],
+        #)
             
     
     with gr.Tab("Video2video"):
@@ -417,20 +417,20 @@ with gr.Blocks() as demo:
                     v2v_step_slider = gr.Slider(minimum=5, maximum=20, step=1, value=15, label="Steps (--steps)")
                 
                 with gr.Row():
-                    v2v_length = gr.Slider(minimum=0, maximum=90, step=1, value=30, label="Length (-L)")
+                    v2v_length = gr.Slider(minimum=0, maximum=999, step=1, value=30, label="Length (-L) (Set to 0 to automatically calculate length)")
                     v2v_seed = gr.Number(value=42, label="Seed (--seed)")
                 
                 v2v_botton = gr.Button("Generate", variant="primary")
             v2v_output_video = gr.PlayableVideo(label="Result", interactive=False)
             
-        gr.Examples(
-            examples=[
-                ["configs/inference/ref_images/Aragaki.png", "configs/inference/video/Aragaki_song.mp4"],
-                ["configs/inference/ref_images/solo.png", "configs/inference/video/Aragaki_song.mp4"],
-                ["configs/inference/ref_images/lyl.png", "configs/inference/head_pose_temp/pose_ref_video.mp4"],
-                ],
-            inputs=[v2v_ref_img, v2v_source_video, a2v_headpose_video],
-        )
+        #gr.Examples(
+           # examples=[
+            #    ["configs/inference/ref_images/Aragaki.png", "configs/inference/video/Aragaki_song.mp4"],
+            #    ["configs/inference/ref_images/solo.png", "configs/inference/video/Aragaki_song.mp4"],
+            #   ["configs/inference/ref_images/lyl.png", "configs/inference/head_pose_temp/pose_ref_video.mp4"],
+            #    ],
+           # inputs=[v2v_ref_img, v2v_source_video, a2v_headpose_video],
+        #)
             
     a2v_botton.click(
         fn=audio2video,
@@ -445,4 +445,4 @@ with gr.Blocks() as demo:
         outputs=[v2v_output_video, v2v_ref_img]
     )
     
-demo.launch()
+demo.launch(share=True)