Upload 5 files

README.md

@@ -1,21 +1,21 @@
 ---
-title: SUPIR Image Upscaler
+title: FramePack/HunyuanVideo
+emoji: 🎥
+colorFrom: pink
+colorTo: gray
 sdk: gradio
-emoji: 📷
-sdk_version: 4.38.1
+sdk_version: 5.29.1
 app_file: app.py
-license: mit
-colorFrom: blue
-colorTo: pink
+license: apache-2.0
+short_description: Text-to-Video/Image-to-Video/Video extender (timed prompt)
 tags:
-  - Upscaling
-  - Restoring
-  - Image-to-Image
-  - Image-2-Image
-  - Img-to-Img
-  - Img-2-Img
+  - Image-to-Video
+  - Image-2-Video
+  - Img-to-Vid
+  - Img-2-Vid
   - language models
   - LLMs
-short_description: Restore blurred or small images with prompt
 suggested_hardware: zero-a10g
----
+---
+
+Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app_endframe.py

@@ -108,7 +108,10 @@ stream = AsyncStream()
 outputs_folder = './outputs/'
 os.makedirs(outputs_folder, exist_ok=True)
 
-input_video_debug_value = prompt_debug_value = total_second_length_debug_value = None
+input_video_debug_value = [None]
+end_frame_debug_value = [None]
+prompt_debug_value = [None]
+total_second_length_debug_value = [None]
 
 # 20250506 pftq: Added function to encode input video frames into latents
 @torch.no_grad()
@@ -588,7 +591,8 @@ def worker(input_video, end_frame, end_frame_weight, prompt, n_prompt, seed, bat
                 )
 
                 if is_start_of_video:
-                    generated_latents = torch.cat([video_latents[:, :, -1:].to(generated_latents), generated_latents], dim=2)
+                    #generated_latents = torch.cat([video_latents[:, :, -1:].to(generated_latents), generated_latents], dim=2)
+                    generated_latents = generated_latents[:, :, 2:]
 
                 total_generated_latent_frames += int(generated_latents.shape[2])
                 history_latents = torch.cat([generated_latents.to(history_latents), history_latents], dim=2)
@@ -669,9 +673,8 @@ def get_duration(
     randomize_seed,
     seed, batch, resolution, total_second_length, latent_window_size, steps, cfg, gs, rs, gpu_memory_preservation, use_teacache,
     no_resize, mp4_crf, num_clean_frames, vae_batch):
-    global total_second_length_debug_value
-    if total_second_length_debug_value is not None:
-        return min(total_second_length_debug_value * 60 * 2, 600)
+    if total_second_length_debug_value[0] is not None:
+        return min(total_second_length_debug_value[0] * 60 * 2, 600)
     return total_second_length * 60 * 2
 
 @spaces.GPU(duration=get_duration)
@@ -680,17 +683,18 @@ def process(
             randomize_seed,
             seed, batch, resolution, total_second_length, latent_window_size, steps, cfg, gs, rs, gpu_memory_preservation, use_teacache,
     no_resize, mp4_crf, num_clean_frames, vae_batch):
-    global stream, high_vram, input_video_debug_value, prompt_debug_value, total_second_length_debug_value
+    global stream, high_vram
 
     if torch.cuda.device_count() == 0:
         gr.Warning('Set this space to GPU config to make it work.')
         return None, None, None, None, None, None
 
-    if input_video_debug_value is not None or prompt_debug_value is not None or total_second_length_debug_value is not None:
-        input_video = input_video_debug_value
-        prompt = prompt_debug_value
-        total_second_length = total_second_length_debug_value
-        input_video_debug_value = prompt_debug_value = total_second_length_debug_value = None
+    if input_video_debug_value[0] is not None or end_frame_debug_value[0] is not None or prompt_debug_value[0] is not None or total_second_length_debug_value[0] is not None:
+        input_video = input_video_debug_value[0]
+        end_frame = end_frame_debug_value[0]
+        prompt = prompt_debug_value[0]
+        total_second_length = total_second_length_debug_value[0]
+        allocation_time = min(total_second_length_debug_value[0] * 60 * 100, 600)
 
     if randomize_seed:
         seed = random.randint(0, np.iinfo(np.int32).max)
@@ -813,6 +817,7 @@ with block:
 
             with gr.Accordion("Debug", open=False):
                 input_video_debug = gr.Video(sources='upload', label="Input Video Debug", height=320)
+                end_frame_debug = gr.Image(type="numpy", label="End Image Debug", height=320)
                 prompt_debug = gr.Textbox(label="Prompt Debug", value='')
                 total_second_length_debug = gr.Slider(label="Additional Video Length to Generate (Seconds) Debug", minimum=1, maximum=120, value=5, step=0.1)
 
@@ -885,28 +890,34 @@ with block:
     end_button.click(fn=end_process)
 
     
-    def handle_field_debug_change(input_video_debug_data, prompt_debug_data, total_second_length_debug_data):
-        global input_video_debug_value, prompt_debug_value, total_second_length_debug_value
-        input_video_debug_value = input_video_debug_data
-        prompt_debug_value = prompt_debug_data
-        total_second_length_debug_value = total_second_length_debug_data
+    def handle_field_debug_change(input_video_debug_data, end_frame_debug_data, prompt_debug_data, total_second_length_debug_data):
+        input_video_debug_value[0] = input_video_debug_data
+        end_frame_debug_value[0] = end_frame_debug_data
+        prompt_debug_value[0] = prompt_debug_data
+        total_second_length_debug_value[0] = total_second_length_debug_data
         return []
     
     input_video_debug.upload(
         fn=handle_field_debug_change,
-        inputs=[input_video_debug, prompt_debug, total_second_length_debug],
+        inputs=[input_video_debug, end_frame_debug, prompt_debug, total_second_length_debug],
+        outputs=[]
+    )
+    
+    end_frame_debug.upload(
+        fn=handle_field_debug_change,
+        inputs=[input_video_debug, end_frame_debug, prompt_debug, total_second_length_debug],
         outputs=[]
     )
     
     prompt_debug.change(
         fn=handle_field_debug_change,
-        inputs=[input_video_debug, prompt_debug, total_second_length_debug],
+        inputs=[input_video_debug, end_frame_debug, prompt_debug, total_second_length_debug],
         outputs=[]
     )
     
     total_second_length_debug.change(
         fn=handle_field_debug_change,
-        inputs=[input_video_debug, prompt_debug, total_second_length_debug],
+        inputs=[input_video_debug, end_frame_debug, prompt_debug, total_second_length_debug],
         outputs=[]
     )
 

requirements.txt

@@ -1,41 +1,24 @@
-pydantic==2.10.6
-fastapi==0.115.13
-gradio_imageslider==0.0.20
-gradio_client==1.10.3
-numpy==1.26.4
-requests==2.32.4
+accelerate==1.7.0
+diffusers==0.33.1
+transformers==4.52.4
 sentencepiece==0.2.0
-tokenizers==0.19.1
-torchvision==0.22.0
-uvicorn==0.34.3
-wandb==0.20.1
-httpx==0.28.1
-transformers==4.43.0
-accelerate==1.8.0
-scikit-learn==1.7.0
-einops==0.8.1
-einops-exts==0.0.4
-timm==1.0.15
-openai-clip==1.0.1
-fsspec==2025.5.1
-kornia==0.8.1
-matplotlib==3.10.3
-ninja==1.11.1.4
-omegaconf==2.3.0
-opencv-python==4.11.0.86
-pandas==2.3.0
 pillow==11.2.1
-pytorch-lightning==2.5.1.post0
-PyYAML==6.0.2
-scipy==1.15.3
-tqdm==4.67.1
-triton==3.3.0
-urllib3==2.4.0
-webdataset==0.2.111
-xformers==0.0.30
-facexlib==0.3.0
-k-diffusion==0.1.1.post1
-diffusers==0.33.1
+av==12.1.0
+numpy==1.26.2
+scipy==1.12.0
+requests==2.32.4
+torchsde==0.2.6
+torch>=2.0.0
+torchvision
+torchaudio
+einops
+opencv-contrib-python
+safetensors
+huggingface_hub
+decord
+imageio_ffmpeg==0.6.0
+sageattention==1.0.6
+xformers==0.0.29.post3
+bitsandbytes==0.46.0
 pillow-heif==0.22.0
-
-open-clip-torch==2.24.0
+spaces[security]