Add end_image

app.py

@@ -120,6 +120,7 @@ outputs_folder = './outputs/'
 os.makedirs(outputs_folder, exist_ok=True)
 
 input_image_debug_value = [None]
+end_image_debug_value = [None]
 input_video_debug_value = [None]
 prompt_debug_value = [None]
 total_second_length_debug_value = [None]
@@ -315,7 +316,7 @@ def set_mp4_comments_imageio_ffmpeg(input_file, comments):
         return False
 
 @torch.no_grad()
-def worker(input_image, image_position, prompts, n_prompt, seed, resolution, total_second_length, latent_window_size, steps, cfg, gs, rs, gpu_memory_preservation, enable_preview, use_teacache, mp4_crf, fps_number):
+def worker(input_image, end_image, image_position, prompts, n_prompt, seed, resolution, total_second_length, latent_window_size, steps, cfg, gs, rs, gpu_memory_preservation, enable_preview, use_teacache, mp4_crf, fps_number):
     def encode_prompt(prompt, n_prompt):
         llama_vec, clip_l_pooler = encode_prompt_conds(prompt, text_encoder, text_encoder_2, tokenizer, tokenizer_2)
 
@@ -864,18 +865,17 @@ def worker_video(input_video, prompts, n_prompt, seed, batch, resolution, total_
     stream.output_queue.push(('end', None))
     return
 
-def get_duration(input_image, image_position, prompts, generation_mode, n_prompt, seed, resolution, total_second_length, allocation_time, latent_window_size, steps, cfg, gs, rs, gpu_memory_preservation, enable_preview, use_teacache, mp4_crf, fps_number):
+def get_duration(input_image, end_image, image_position, prompts, generation_mode, n_prompt, seed, resolution, total_second_length, allocation_time, latent_window_size, steps, cfg, gs, rs, gpu_memory_preservation, enable_preview, use_teacache, mp4_crf, fps_number):
     return allocation_time
 
-# Remove this decorator if you run on local
 @spaces.GPU(duration=get_duration)
-def process_on_gpu(input_image, image_position, prompts, generation_mode, n_prompt, seed, resolution, total_second_length, allocation_time, latent_window_size, steps, cfg, gs, rs, gpu_memory_preservation, enable_preview, use_teacache, mp4_crf, fps_number
+def process_on_gpu(input_image, end_image, image_position, prompts, generation_mode, n_prompt, seed, resolution, total_second_length, allocation_time, latent_window_size, steps, cfg, gs, rs, gpu_memory_preservation, enable_preview, use_teacache, mp4_crf, fps_number
            ):
     start = time.time()
     global stream
     stream = AsyncStream()
 
-    async_run(worker, input_image, image_position, prompts, n_prompt, seed, resolution, total_second_length, latent_window_size, steps, cfg, gs, rs, gpu_memory_preservation, enable_preview, use_teacache, mp4_crf, fps_number)
+    async_run(worker, input_image, end_image, image_position, prompts, n_prompt, seed, resolution, total_second_length, latent_window_size, steps, cfg, gs, rs, gpu_memory_preservation, enable_preview, use_teacache, mp4_crf, fps_number)
 
     output_filename = None
 
@@ -905,6 +905,7 @@ def process_on_gpu(input_image, image_position, prompts, generation_mode, n_prom
             break
 
 def process(input_image,
+            end_image,
             image_position=0,
             prompt="",
             generation_mode="image",
@@ -955,6 +956,7 @@ def process(input_image,
     yield gr.update(label="Previewed Frames"), None, '', '', gr.update(interactive=False), gr.update(interactive=True), gr.skip()
 
     yield from process_on_gpu(input_image,
+                              end_image,
             image_position,
             prompts,
             generation_mode,
@@ -978,7 +980,6 @@ def process(input_image,
 def get_duration_video(input_video, prompts, n_prompt, seed, batch, resolution, total_second_length, allocation_time, latent_window_size, steps, cfg, gs, rs, gpu_memory_preservation, enable_preview, use_teacache, no_resize, mp4_crf, num_clean_frames, vae_batch):
     return allocation_time
 
-# Remove this decorator if you run on local
 @spaces.GPU(duration=get_duration_video)
 def process_video_on_gpu(input_video, prompts, n_prompt, seed, batch, resolution, total_second_length, allocation_time, latent_window_size, steps, cfg, gs, rs, gpu_memory_preservation, enable_preview, use_teacache, no_resize, mp4_crf, num_clean_frames, vae_batch):
     start = time.time()
@@ -1216,7 +1217,7 @@ with block:
             progress_bar = gr.HTML('', elem_classes='no-generating-animation')
 
     # 20250506 pftq: Updated inputs to include num_clean_frames
-    ips = [input_image, image_position, final_prompt, generation_mode, n_prompt, randomize_seed, seed, auto_allocation, allocation_time, resolution, total_second_length, latent_window_size, steps, cfg, gs, rs, gpu_memory_preservation, enable_preview, use_teacache, mp4_crf, fps_number]
+    ips = [input_image, end_image, image_position, final_prompt, generation_mode, n_prompt, randomize_seed, seed, auto_allocation, allocation_time, resolution, total_second_length, latent_window_size, steps, cfg, gs, rs, gpu_memory_preservation, enable_preview, use_teacache, mp4_crf, fps_number]
     ips_video = [input_video, final_prompt, n_prompt, randomize_seed, seed, auto_allocation, allocation_time, batch, resolution, total_second_length, latent_window_size, steps, cfg, gs, rs, gpu_memory_preservation, enable_preview, use_teacache, no_resize, mp4_crf, num_clean_frames, vae_batch]
 
     with gr.Row(elem_id="text_examples", visible=False):
@@ -1225,6 +1226,7 @@ with block:
             examples = [
                     [
                         None, # input_image
+                        None, # end_image
                         0, # image_position
                         "Overcrowed street in Japan, photorealistic, realistic, intricate details, 8k, insanely detailed",
                         "text", # generation_mode
@@ -1260,6 +1262,7 @@ with block:
             examples = [
                     [
                         "./img_examples/Example2.webp", # input_image
+                        None, # end_image
                         0, # image_position
                         "A man on the left and a woman on the right face each other ready to start a conversation, large space between the persons, full view, full-length view, 3D, pixar, 3D render, CGI. The man talks and the woman listens; A man on the left and a woman on the right face each other ready to start a conversation, large space between the persons, full view, full-length view, 3D, pixar, 3D render, CGI. The woman talks, the man stops talking and the man listens; A man on the left and a woman on the right face each other ready to start a conversation, large space between the persons, full view, full-length view, 3D, pixar, 3D render, CGI. The woman talks and the man listens",
                         "image", # generation_mode
@@ -1283,6 +1286,7 @@ with block:
                     ],
                     [
                         "./img_examples/Example1.png", # input_image
+                        None, # end_image
                         0, # image_position
                         "A dolphin emerges from the water, photorealistic, realistic, intricate details, 8k, insanely detailed",
                         "image", # generation_mode
@@ -1306,6 +1310,7 @@ with block:
                     ],
                     [
                         "./img_examples/Example4.webp", # input_image
+                        None, # end_image
                         1, # image_position
                         "A building starting to explode, photorealistic, realisitc, 8k, insanely detailed",
                         "image", # generation_mode
@@ -1329,6 +1334,7 @@ with block:
                     ],
                     [
                         "./img_examples/Example4.webp", # input_image
+                        None, # end_image
                         50, # image_position
                         "A building starting to explode, photorealistic, realisitc, 8k, insanely detailed",
                         "image", # generation_mode
@@ -1352,6 +1358,7 @@ with block:
                     ],
                     [
                         "./img_examples/Example4.webp", # input_image
+                        None, # end_image
                         100, # image_position
                         "A building starting to explode, photorealistic, realisitc, 8k, insanely detailed",
                         "image", # generation_mode
@@ -1446,6 +1453,7 @@ with block:
         examples = [
                 [
                     None, # input_image
+                    None, # end_image
                     0, # image_position
                     "Overcrowed street in Japan, photorealistic, realistic, intricate details, 8k, insanely detailed",
                     "text", # generation_mode
@@ -1480,6 +1488,7 @@ with block:
         examples = [
                 [
                     "./img_examples/Example1.png", # input_image
+                    None, # end_image
                     0, # image_position
                     "A dolphin emerges from the water, photorealistic, realistic, intricate details, 8k, insanely detailed",
                     "image", # generation_mode
@@ -1503,6 +1512,7 @@ with block:
                 ],
                 [
                     "./img_examples/Example2.webp", # input_image
+                    None, # end_image
                     0, # image_position
                     "A man on the left and a woman on the right face each other ready to start a conversation, large space between the persons, full view, full-length view, 3D, pixar, 3D render, CGI. The man talks and the woman listens; A man on the left and a woman on the right face each other ready to start a conversation, large space between the persons, full view, full-length view, 3D, pixar, 3D render, CGI. The woman talks, the man stops talking and the man listens; A man on the left and a woman on the right face each other ready to start a conversation, large space between the persons, full view, full-length view, 3D, pixar, 3D render, CGI. The woman talks and the man listens",
                     "image", # generation_mode
@@ -1526,6 +1536,7 @@ with block:
                 ],
                 [
                     "./img_examples/Example2.webp", # input_image
+                    None, # end_image
                     0, # image_position
                     "A man on the left and a woman on the right face each other ready to start a conversation, large space between the persons, full view, full-length view, 3D, pixar, 3D render, CGI. The woman talks and the man listens; A man on the left and a woman on the right face each other ready to start a conversation, large space between the persons, full view, full-length view, 3D, pixar, 3D render, CGI. The man talks, the woman stops talking and the woman listens A man on the left and a woman on the right face each other ready to start a conversation, large space between the persons, full view, full-length view, 3D, pixar, 3D render, CGI. The man talks and the woman listens",
                     "image", # generation_mode
@@ -1549,6 +1560,7 @@ with block:
                 ],
                 [
                     "./img_examples/Example3.jpg", # input_image
+                    None, # end_image
                     0, # image_position
                     "A boy is walking to the right, full view, full-length view, cartoon",
                     "image", # generation_mode
@@ -1572,6 +1584,7 @@ with block:
                 ],
                 [
                     "./img_examples/Example4.webp", # input_image
+                    None, # end_image
                     100, # image_position
                     "A building starting to explode, photorealistic, realisitc, 8k, insanely detailed",
                     "image", # generation_mode