Update app.py

app.py

@@ -7,6 +7,7 @@ import spaces
 from huggingface_hub import hf_hub_download
 import numpy as np
 import random
+import os
 
 MODEL_ID = "Wan-AI/Wan2.1-T2V-1.3B-Diffusers"
 LORA_REPO_ID = "Kijai/WanVideo_comfy"
@@ -28,8 +29,31 @@ MOD_VALUE = 32
 DEFAULT_H_SLIDER_VALUE = 384 # 512
 DEFAULT_W_SLIDER_VALUE = 640 # 896
 
-SLIDER_MIN_H, SLIDER_MAX_H = 128, 1280
-SLIDER_MIN_W, SLIDER_MAX_W = 128, 1280
+# Environment variable check
+IS_ORIGINAL_SPACE = os.environ.get("IS_ORIGINAL_SPACE", "False") == "True"
+
+# Original limits
+ORIGINAL_SLIDER_MIN_H, ORIGINAL_SLIDER_MAX_H = 128, 1280
+ORIGINAL_SLIDER_MIN_W, ORIGINAL_SLIDER_MAX_W = 128, 1280
+ORIGINAL_MAX_DURATION = round(81/24, 1)  # MAX_FRAMES_MODEL/FIXED_FPS
+
+# Limited space constants
+LIMITED_MAX_RESOLUTION = 640
+LIMITED_MAX_DURATION = 2.0
+LIMITED_MAX_STEPS = 3
+
+# Set limits based on environment variable
+if IS_ORIGINAL_SPACE:
+    SLIDER_MIN_H, SLIDER_MAX_H = ORIGINAL_SLIDER_MIN_H, ORIGINAL_SLIDER_MAX_H
+    SLIDER_MIN_W, SLIDER_MAX_W = ORIGINAL_SLIDER_MIN_W, ORIGINAL_SLIDER_MAX_W
+    MAX_DURATION = ORIGINAL_MAX_DURATION
+    MAX_STEPS = 8
+else:
+    SLIDER_MIN_H, SLIDER_MAX_H = 128, LIMITED_MAX_RESOLUTION
+    SLIDER_MIN_W, SLIDER_MAX_W = 128, LIMITED_MAX_RESOLUTION
+    MAX_DURATION = LIMITED_MAX_DURATION
+    MAX_STEPS = LIMITED_MAX_STEPS
+
 MAX_SEED = np.iinfo(np.int32).max
 
 FIXED_FPS = 24
@@ -62,7 +86,7 @@ def generate_video(prompt, height, width,
     
     This function takes a text prompt and generates a video based on the provided
     prompt and parameters. It uses the Wan 2.1 1.3B Text-to-Video model with CausVid LoRA
-    for fast generation in 4-8 steps.
+    for fast generation in 3-8 steps.
     
     Args:
         prompt (str): Text prompt describing the desired video content.
@@ -97,7 +121,14 @@ def generate_video(prompt, height, width,
         - Generation time varies based on steps and duration (see get_duration function)
     """
     if not prompt or prompt.strip() == "":
-        raise gr.Error("Please enter a text prompt.")
+        raise gr.Error("Please enter a text prompt. Try to use long and precise descriptions.")
+
+    # Apply limits based on environment variable
+    if not IS_ORIGINAL_SPACE:
+        height = min(height, LIMITED_MAX_RESOLUTION)
+        width = min(width, LIMITED_MAX_RESOLUTION)
+        duration_seconds = min(duration_seconds, LIMITED_MAX_DURATION)
+        steps = min(steps, LIMITED_MAX_STEPS)
 
     target_h = max(MOD_VALUE, (int(height) // MOD_VALUE) * MOD_VALUE)
     target_w = max(MOD_VALUE, (int(width) // MOD_VALUE) * MOD_VALUE)
@@ -120,21 +151,45 @@ def generate_video(prompt, height, width,
     return video_path, current_seed
 
 with gr.Blocks() as demo:
-    gr.Markdown("# InstaVideo")
-    gr.Markdown("This 🧨 diffusers demo uses the [Wan2.1 1.3B CausVid LoRA by Kijai](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan21_CausVid_bidirect2_T2V_1_3B_lora_rank32.safetensors).")
+    gr.Markdown("# ⚡ InstaVideo")
+    gr.Markdown("This Gradio space is forked by [wan2-1-fast from multimodalart](https://huggingface.co/spaces/multimodalart/wan2-1-fast), and is powered by the Wan CausVid LoRA [from Kijai](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan21_CausVid_bidirect2_T2V_1_3B_lora_rank32.safetensors).")
+
+    # Add notice for limited spaces
+    if not IS_ORIGINAL_SPACE:
+        gr.Markdown("This free public demo limits the resolution to 640x640, duration to 2s, and inference steps to 3. For full capabilities please duplicate this space.")
+    
     with gr.Row():
         with gr.Column():
             prompt_input = gr.Textbox(label="Prompt", value=default_prompt_t2v, placeholder="Describe the video you want to generate...")
-            duration_seconds_input = gr.Slider(minimum=round(MIN_FRAMES_MODEL/FIXED_FPS,1), maximum=round(MAX_FRAMES_MODEL/FIXED_FPS,1), step=0.1, value=2, label="Duration (seconds)", info=f"Clamped to model's {MIN_FRAMES_MODEL}-{MAX_FRAMES_MODEL} frames at {FIXED_FPS}fps.")
+            duration_seconds_input = gr.Slider(
+                minimum=round(MIN_FRAMES_MODEL/FIXED_FPS,1), 
+                maximum=MAX_DURATION, 
+                step=0.1, 
+                value=2, 
+                label="Duration (seconds)", 
+                info=f"Clamped to model's {MIN_FRAMES_MODEL}-{MAX_FRAMES_MODEL} frames at {FIXED_FPS}fps."
+            )
             
             with gr.Accordion("Advanced Settings", open=False):
                 negative_prompt_input = gr.Textbox(label="Negative Prompt", value=default_negative_prompt, lines=3)
                 seed_input = gr.Slider(label="Seed", minimum=0, maximum=MAX_SEED, step=1, value=42, interactive=True)
                 randomize_seed_checkbox = gr.Checkbox(label="Randomize seed", value=True, interactive=True)
                 with gr.Row():
-                    height_input = gr.Slider(minimum=SLIDER_MIN_H, maximum=SLIDER_MAX_H, step=MOD_VALUE, value=DEFAULT_H_SLIDER_VALUE, label=f"Output Height (multiple of {MOD_VALUE})")
-                    width_input = gr.Slider(minimum=SLIDER_MIN_W, maximum=SLIDER_MAX_W, step=MOD_VALUE, value=DEFAULT_W_SLIDER_VALUE, label=f"Output Width (multiple of {MOD_VALUE})")
-                steps_slider = gr.Slider(minimum=1, maximum=8, step=1, value=3, label="Inference Steps") 
+                    height_input = gr.Slider(
+                        minimum=SLIDER_MIN_H, 
+                        maximum=SLIDER_MAX_H, 
+                        step=MOD_VALUE, 
+                        value=min(DEFAULT_H_SLIDER_VALUE, SLIDER_MAX_H), 
+                        label=f"Output Height (multiple of {MOD_VALUE})"
+                    )
+                    width_input = gr.Slider(
+                        minimum=SLIDER_MIN_W, 
+                        maximum=SLIDER_MAX_W, 
+                        step=MOD_VALUE, 
+                        value=min(DEFAULT_W_SLIDER_VALUE, SLIDER_MAX_W), 
+                        label=f"Output Width (multiple of {MOD_VALUE})"
+                    )
+                steps_slider = gr.Slider(minimum=1, maximum=MAX_STEPS, step=1, value=3, label="Inference Steps") 
                 guidance_scale_input = gr.Slider(minimum=0.0, maximum=20.0, step=0.5, value=1.0, label="Guidance Scale", visible=False)
 
             generate_button = gr.Button("Generate Video", variant="primary")
@@ -148,13 +203,26 @@ with gr.Blocks() as demo:
     ]
     generate_button.click(fn=generate_video, inputs=ui_inputs, outputs=[video_output, seed_input])
 
+    # Adjust examples based on space limits
+    example_configs = [
+        ["a majestic eagle soaring through mountain peaks, cinematic aerial view", 896, 512],
+        ["a serene ocean wave crashing on a sandy beach at sunset", 448, 832],
+        ["a field of flowers swaying in the wind, spring morning light", 512, 896],
+    ]
+    
+    if not IS_ORIGINAL_SPACE:
+        # Limit example resolutions for limited spaces
+        example_configs = [
+            [example[0], min(example[1], LIMITED_MAX_RESOLUTION), min(example[2], LIMITED_MAX_RESOLUTION)]
+            for example in example_configs
+        ]
+    
     gr.Examples(
-        examples=[ 
-            ["a majestic eagle soaring through mountain peaks, cinematic aerial view", 896, 512],
-            ["a serene ocean wave crashing on a sandy beach at sunset", 448, 832],
-            ["a field of flowers swaying in the wind, spring morning light", 512, 896],
-        ],
-        inputs=[prompt_input, height_input, width_input], outputs=[video_output, seed_input], fn=generate_video, cache_examples="lazy"
+        examples=example_configs,
+        inputs=[prompt_input, height_input, width_input], 
+        outputs=[video_output, seed_input], 
+        fn=generate_video, 
+        cache_examples="lazy"
     )
 
 if __name__ == "__main__":