Update app.py

app.py

@@ -117,60 +117,6 @@ os.makedirs("./gradio_tmp", exist_ok=True)
 upscale_model = load_sd_upscale("model_real_esran/RealESRGAN_x4.pth", device)
 frame_interpolation_model = load_rife_model("model_rife")
 
-@spaces.GPU(duration=65)
-def infer(
-    prompt: str,
-    image_input: str,
-    num_inference_steps: int,
-    guidance_scale: float,
-    seed: int = 42,
-    progress=gr.Progress(track_tqdm=True),
-):
-    if seed == -1:
-        seed = random.randint(0, 2**8 - 1)
-
-    id_image = np.array(ImageOps.exif_transpose(Image.fromarray(image_input)).convert("RGB"))
-    id_image = resize_numpy_image_long(id_image, 1024)
-    id_cond, id_vit_hidden, align_crop_face_image, face_kps = process_face_embeddings(face_helper, face_clip_model, handler_ante, 
-                                                                            eva_transform_mean, eva_transform_std, 
-                                                                            face_main_model, device, dtype, id_image, 
-                                                                            original_id_image=id_image, is_align_face=True, 
-                                                                            cal_uncond=False)
-    
-    if is_kps:
-        kps_cond = face_kps
-    else:
-        kps_cond = None
-
-    tensor = align_crop_face_image.cpu().detach()
-    tensor = tensor.squeeze()
-    tensor = tensor.permute(1, 2, 0)
-    tensor = tensor.numpy() * 255
-    tensor = tensor.astype(np.uint8)
-    image  = ImageOps.exif_transpose(Image.fromarray(tensor))
-
-    prompt = prompt.strip('"')
-    
-    generator = torch.Generator(device).manual_seed(seed) if seed else None
-    
-    video_pt = pipe(
-        prompt=prompt,
-        image=image,
-        num_videos_per_prompt=1,
-        num_inference_steps=num_inference_steps,
-        num_frames=49,
-        use_dynamic_cfg=False,
-        guidance_scale=guidance_scale,
-        generator=generator,
-        id_vit_hidden=id_vit_hidden,
-        id_cond=id_cond,
-        kps_cond=kps_cond,
-        output_type="pt",
-    ).frames
-    
-    ##free_memory()
-    return video_pt, seed
-
 
 def convert_to_gif(video_path):
     clip = VideoFileClip(video_path)
@@ -196,7 +142,7 @@ def delete_old_files():
 
 
 ##threading.Thread(target=delete_old_files, daemon=True).start()
-@spaces.GPU
+@spaces.GPU(duration=65)
 def generate(
     prompt,
     image_input,
@@ -205,14 +151,40 @@ def generate(
     rife_status,
     progress=gr.Progress(track_tqdm=True)
 ):
-    latents, seed = infer(
-        prompt,
-        image_input,
-        num_inference_steps=4,
-        guidance_scale=7.0,
-        seed=seed_value,
-        progress=progress,
-    )
+    def infer(prompt: str,image_input: str,num_inference_steps: int,guidance_scale: float,seed: int = 42,progress=gr.Progress(track_tqdm=True),):
+        if seed == -1:
+            seed = random.randint(0, 2**8 - 1)
+
+        id_image = np.array(ImageOps.exif_transpose(Image.fromarray(image_input)).convert("RGB"))
+        id_image = resize_numpy_image_long(id_image, 1024)
+        id_cond, id_vit_hidden, align_crop_face_image, face_kps = process_face_embeddings(face_helper, face_clip_model, handler_ante, 
+                                                                            eva_transform_mean, eva_transform_std, 
+                                                                            face_main_model, device, dtype, id_image, 
+                                                                            original_id_image=id_image, is_align_face=True, 
+                                                                            cal_uncond=False)
+    
+        if is_kps:
+            kps_cond = face_kps
+        else:
+            kps_cond = None
+
+        tensor = align_crop_face_image.cpu().detach()
+        tensor = tensor.squeeze()
+        tensor = tensor.permute(1, 2, 0)
+        tensor = tensor.numpy() * 255
+        tensor = tensor.astype(np.uint8)
+        image  = ImageOps.exif_transpose(Image.fromarray(tensor))
+
+        prompt = prompt.strip('"')
+    
+        generator = torch.Generator(device).manual_seed(seed) if seed else None
+    
+        video_pt = pipe(prompt=prompt,image=image,num_videos_per_prompt=1,num_inference_steps=num_inference_steps,num_frames=49,use_dynamic_cfg=False,guidance_scale=guidance_scale,generator=generator,id_vit_hidden=id_vit_hidden,id_cond=id_cond,kps_cond=kps_cond,output_type="pt",).frames
+    
+        ##free_memory()
+        return video_pt, seed
+
+    latents, seed = infer(prompt,image_input,num_inference_steps=4,guidance_scale=7.0,seed=seed_value,progress=progress,)
     if scale_status:
         latents = upscale_batch_and_concatenate(upscale_model, latents, device)
     if rife_status: