End stillness

app.py

@@ -1338,11 +1338,11 @@ def process(input_image,
             fps_number
            )
 
-def get_duration_video(input_video, end_frame, 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):
+def get_duration_video(input_video, end_frame, end_stillness, 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
 
 @spaces.GPU(duration=get_duration_video)
-def process_video_on_gpu(input_video, end_frame, 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):
+def process_video_on_gpu(input_video, end_frame, end_stillness, 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()
     global stream
     stream = AsyncStream()
@@ -1378,7 +1378,7 @@ def process_video_on_gpu(input_video, end_frame, prompts, n_prompt, seed, batch,
             " You can upscale the result with https://huggingface.co/spaces/Nick088/Real-ESRGAN_Pytorch. To make all your generated scenes consistent, you can then apply a face swap on the main character. If you do not see the generated video above, the process may have failed. See the logs for more information. If you see an error like ''NVML_SUCCESS == r INTERNAL ASSERT FAILED'', you probably haven't enough VRAM. Test an example or other options to compare. You can share your inputs to the original space or set your space in public for a peer review.", '', gr.update(interactive=True), gr.update(interactive=False), gr.update(visible = False)
             break
 
-def process_video(input_video, end_frame, 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):
+def process_video(input_video, end_frame, end_stillness, 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):
     global high_vram
     if auto_allocation:
         allocation_time = min(total_second_length * 60 * (2.5 if use_teacache else 3.5) * (1 + ((steps - 25) / 25))**2, 600)
@@ -1419,7 +1419,7 @@ def process_video(input_video, end_frame, prompt, n_prompt, randomize_seed, seed
     if cfg > 1:
         gs = 1
 
-    yield from process_video_on_gpu(input_video, end_frame, prompt, 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)
+    yield from process_video_on_gpu(input_video, end_frame, end_stillness, prompt, 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)
 
 def end_process():
     stream.input_queue.push('end')
@@ -1495,6 +1495,7 @@ with block:
             image_position = gr.Slider(label="Image position", minimum=0, maximum=100, value=0, step=1, info='0=Video start; 100=Video end (lower quality)')
             input_video = gr.Video(sources='upload', label="Input Video", height=320)
             end_image = gr.Image(sources='upload', type="numpy", label="End Frame (optional; more a guidance than the exact end frame)", height=320)
+            end_stillness = gr.Slider(label="End stillness", minimum=0, maximum=100, value=0, step=1, info='0=Realistic end; 100=Matches exactly the end image (but the time seems to freeze)')
             timeless_prompt = gr.Textbox(label="Timeless prompt", info='Used on the whole duration of the generation', value='', placeholder="The creature starts to move, fast motion, fixed camera, focus motion, consistent arm, consistent position, mute colors, insanely detailed")
             prompt_number = gr.Slider(label="Timed prompt number", minimum=0, maximum=1000, value=0, step=1, info='Prompts will automatically appear')
 
@@ -1581,7 +1582,7 @@ with block:
             progress_bar = gr.HTML('', elem_classes='no-generating-animation')
 
     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, end_image, 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]
+    ips_video = [input_video, end_image, end_stillness, 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):
         gr.Examples(
@@ -1794,6 +1795,7 @@ with block:
                     [
                         "./img_examples/Example1.mp4", # input_video
                         None, # end_image
+                        0, # end_stillness
                         "View of the sea as far as the eye can see, from the seaside, a piece of land is barely visible on the horizon at the middle, the sky is radiant, reflections of the sun in the water, photorealistic, realistic, intricate details, 8k, insanely detailed",
                         "Missing arm, long hand, unrealistic position, impossible contortion, visible bone, muscle contraction, poorly framed, blurred, blurry, over-smooth", # n_prompt
                         True, # randomize_seed
@@ -1819,6 +1821,7 @@ with block:
                     [
                         "./img_examples/Example1.mp4", # input_video
                         None, # end_image
+                        0, # end_stillness
                         "View of the sea as far as the eye can see, from the seaside, a piece of land is barely visible on the horizon at the middle, the sky is radiant, reflections of the sun in the water, photorealistic, realistic, intricate details, 8k, insanely detailed",
                         "Missing arm, long hand, unrealistic position, impossible contortion, visible bone, muscle contraction, poorly framed, blurred, blurry, over-smooth", # n_prompt
                         True, # randomize_seed
@@ -2056,6 +2059,7 @@ with block:
                 [
                     "./img_examples/Example1.mp4", # input_video
                     None, # end_image
+                    0, # end_stillness
                     "View of the sea as far as the eye can see, from the seaside, a piece of land is barely visible on the horizon at the middle, the sky is radiant, reflections of the sun in the water, photorealistic, realistic, intricate details, 8k, insanely detailed",
                     "Missing arm, long hand, unrealistic position, impossible contortion, visible bone, muscle contraction, poorly framed, blurred, blurry, over-smooth", # n_prompt
                     True, # randomize_seed
@@ -2081,6 +2085,7 @@ with block:
                 [
                     "./img_examples/Example1.mp4", # input_video
                     "./img_examples/Example1.png", # end_image
+                    0, # end_stillness
                     "View of the sea as far as the eye can see, from the seaside, a piece of land is barely visible on the horizon at the middle, the sky is radiant, reflections of the sun in the water, photorealistic, realistic, intricate details, 8k, insanely detailed",
                     "Missing arm, long hand, unrealistic position, impossible contortion, visible bone, muscle contraction, poorly framed, blurred, blurry, over-smooth", # n_prompt
                     True, # randomize_seed
@@ -2140,6 +2145,7 @@ with block:
             gr.update(visible = False), # image_position
             gr.update(visible = False), # input_image
             gr.update(visible = False), # end_image
+            gr.update(visible = False), # end_stillness
             gr.update(visible = False), # input_video
             gr.update(visible = True),  # start_button
             gr.update(visible = False), # start_button_video
@@ -2156,6 +2162,7 @@ with block:
             gr.update(visible = True),  # image_position
             gr.update(visible = True),  # input_image
             gr.update(visible = False), # end_image
+            gr.update(visible = False), # end_stillness
             gr.update(visible = False), # input_video
             gr.update(visible = True),  # start_button
             gr.update(visible = False), # start_button_video
@@ -2172,6 +2179,7 @@ with block:
             gr.update(visible = False), # image_position
             gr.update(visible = True),  # input_image
             gr.update(visible = True),  # end_image
+            gr.update(visible = True), # end_stillness
             gr.update(visible = False), # input_video
             gr.update(visible = True),  # start_button
             gr.update(visible = False), # start_button_video
@@ -2188,6 +2196,7 @@ with block:
             gr.update(visible = False), # image_position
             gr.update(visible = False), # input_image
             gr.update(visible = True), # end_image
+            gr.update(visible = True), # end_stillness
             gr.update(visible = True),  # input_video
             gr.update(visible = False), # start_button
             gr.update(visible = True),  # start_button_video
@@ -2258,7 +2267,7 @@ with block:
     generation_mode.change(
         fn=handle_generation_mode_change,
         inputs=[generation_mode],
-        outputs=[text_to_video_hint, image_position, input_image, end_image, input_video, start_button, start_button_video, no_resize, batch, num_clean_frames, vae_batch, prompt_hint, fps_number]
+        outputs=[text_to_video_hint, image_position, input_image, end_image, end_stillness, input_video, start_button, start_button_video, no_resize, batch, num_clean_frames, vae_batch, prompt_hint, fps_number]
     )
 
     # Update display when the page loads
@@ -2266,7 +2275,7 @@ with block:
         fn=handle_generation_mode_change, inputs = [
         generation_mode
     ], outputs = [
-       text_to_video_hint, image_position, input_image, end_image, input_video, start_button, start_button_video, no_resize, batch, num_clean_frames, vae_batch, prompt_hint, fps_number
+       text_to_video_hint, image_position, input_image, end_image, end_stillness, input_video, start_button, start_button_video, no_resize, batch, num_clean_frames, vae_batch, prompt_hint, fps_number
     ]
     )