Upload folder using huggingface_hub

app.py

@@ -0,0 +1,317 @@
+import torch
+from diffusers import AutoencoderKLWan, WanImageToVideoPipeline, UniPCMultistepScheduler
+from diffusers.utils import export_to_video
+from transformers import CLIPVisionModel
+import gradio as gr
+import tempfile
+import spaces
+from huggingface_hub import hf_hub_download, list_repo_files
+import numpy as np
+from PIL import Image
+import random
+import os
+
+# ----------------------
+#
+# CONFIGURATION
+#
+# ----------------------
+
+MODEL_ID = "gaalos/Wan2.1-I2V-14B-720P-Diffusers-scaled"
+LORA_REPO_ID = "hotdogs/wan_nsfw_lora"
+MOD_VALUE = 32
+FIXED_FPS = 24
+MIN_FRAMES_MODEL = 8
+MAX_FRAMES_MODEL = 30 * FIXED_FPS  # 30s max
+MAX_SEED = np.iinfo(np.int32).max
+
+DEFAULT_H_SLIDER_VALUE = 640
+DEFAULT_W_SLIDER_VALUE = 1024
+NEW_FORMULA_MAX_AREA = 640.0 * 1024.0
+SLIDER_MIN_H, SLIDER_MAX_H = 128, 1024
+SLIDER_MIN_W, SLIDER_MAX_W = 128, 1024
+
+default_prompt_i2v = "make this image come alive, cinematic motion, smooth animation"
+default_negative_prompt = "Bright tones, overexposed, static, blurred details, subtitles, style, works, paintings, images, static, overall gray, worst quality, low quality, JPEG compression residue, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn faces, deformed, disfigured, misshapen limbs, fused fingers, still picture, messy background, three legs, many people in the background, walking backwards, watermark, text, signature"
+
+# ----------------------
+# LOAD BASE MODEL
+# ----------------------
+
+print("Loading models...")
+image_encoder = CLIPVisionModel.from_pretrained(MODEL_ID, subfolder="image_encoder", torch_dtype=torch.float32)
+vae = AutoencoderKLWan.from_pretrained(MODEL_ID, subfolder="vae", torch_dtype=torch.float32)
+pipe = WanImageToVideoPipeline.from_pretrained(
+    MODEL_ID, vae=vae, image_encoder=image_encoder, torch_dtype=torch.bfloat16
+)
+pipe.scheduler = UniPCMultistepScheduler.from_config(pipe.scheduler.config, flow_shift=8.0)
+pipe.to("cuda")
+print("Models loaded successfully!")
+
+# ----------------------
+# LOAD ALL LORAS FROM "hotdogs/wan_nsfw_lora"
+# ----------------------
+
+print("Loading LoRAs...")
+lora_files = [f for f in list_repo_files(LORA_REPO_ID) if f.endswith(".safetensors")]
+lora_paths = {}
+for file in lora_files:
+    local_path = hf_hub_download(repo_id=LORA_REPO_ID, filename=file)
+    adapter_name = os.path.splitext(os.path.basename(file))[0]
+    lora_paths[adapter_name] = local_path
+    pipe.load_lora_weights(local_path, adapter_name=adapter_name)
+print(f"Loaded {len(lora_paths)} LoRA adapters: {list(lora_paths.keys())}")
+
+# ----------------------
+# DIMENSION HELPERS
+# ----------------------
+
+def _calculate_new_dimensions(pil_image):
+    """Calculate optimal dimensions based on image aspect ratio"""
+    orig_w, orig_h = pil_image.size
+    if orig_w <= 0 or orig_h <= 0:
+        return DEFAULT_H_SLIDER_VALUE, DEFAULT_W_SLIDER_VALUE
+
+    aspect_ratio = orig_h / orig_w
+    calc_h = round(np.sqrt(NEW_FORMULA_MAX_AREA * aspect_ratio))
+    calc_w = round(np.sqrt(NEW_FORMULA_MAX_AREA / aspect_ratio))
+
+    calc_h = max(MOD_VALUE, (calc_h // MOD_VALUE) * MOD_VALUE)
+    calc_w = max(MOD_VALUE, (calc_w // MOD_VALUE) * MOD_VALUE)
+
+    new_h = int(np.clip(calc_h, SLIDER_MIN_H, (SLIDER_MAX_H // MOD_VALUE) * MOD_VALUE))
+    new_w = int(np.clip(calc_w, SLIDER_MIN_W, (SLIDER_MAX_W // MOD_VALUE) * MOD_VALUE))
+
+    return new_h, new_w
+
+def handle_image_upload(image, current_h, current_w):
+    """Update height/width sliders when image is uploaded"""
+    if image is None:
+        return gr.update(value=DEFAULT_H_SLIDER_VALUE), gr.update(value=DEFAULT_W_SLIDER_VALUE)
+    try:
+        new_h, new_w = _calculate_new_dimensions(image)
+        return gr.update(value=new_h), gr.update(value=new_w)
+    except Exception as e:
+        print(f"Error calculating dimensions: {e}")
+        return gr.update(value=DEFAULT_H_SLIDER_VALUE), gr.update(value=DEFAULT_W_SLIDER_VALUE)
+
+# ----------------------
+# DURATION CONFIGURATION
+# ----------------------
+
+def get_duration(input_image, prompt, height, width, negative_prompt, duration_seconds,
+                 guidance_scale, steps, seed, randomize_seed, lora_list, progress):
+    """Calculate GPU duration for spaces.GPU decorator"""
+    base = 60
+    if duration_seconds > 10: base += 30
+    if duration_seconds > 20: base += 30
+    if steps > 8: base += 20
+    return base
+
+# ----------------------
+# MAIN GENERATION FUNCTION
+# ----------------------
+
+@spaces.GPU(duration=get_duration)
+def generate_video(input_image, prompt, height, width,
+                   negative_prompt=default_negative_prompt, duration_seconds=2,
+                   guidance_scale=1, steps=4,
+                   seed=42, randomize_seed=False,
+                   lora_list=None, progress=gr.Progress(track_tqdm=True)):
+    """
+    Generate video from image using Wan2.1 I2V pipeline with optional LoRAs
+    """
+    if input_image is None:
+        raise gr.Error("Please upload an image.")
+
+    current_seed = random.randint(0, MAX_SEED) if randomize_seed else int(seed)
+    target_h = max(MOD_VALUE, (int(height) // MOD_VALUE) * MOD_VALUE)
+    target_w = max(MOD_VALUE, (int(width) // MOD_VALUE) * MOD_VALUE)
+    num_frames = np.clip(int(round(duration_seconds * FIXED_FPS)), MIN_FRAMES_MODEL, MAX_FRAMES_MODEL)
+    resized_image = input_image.resize((target_w, target_h))
+
+    selected_adapters = lora_list or []
+    pipe.set_adapters(selected_adapters, adapter_weights=[0.95] * len(selected_adapters))
+    pipe.fuse_lora()
+
+    with torch.inference_mode():
+        output_frames_list = pipe(
+            image=resized_image, prompt=prompt, negative_prompt=negative_prompt,
+            height=target_h, width=target_w, num_frames=num_frames,
+            guidance_scale=float(guidance_scale), num_inference_steps=int(steps),
+            generator=torch.Generator(device="cuda").manual_seed(current_seed)
+        ).frames[0]
+
+    with tempfile.NamedTemporaryFile(suffix=".mp4", delete=False) as tmpfile:
+        video_path = tmpfile.name
+    export_to_video(output_frames_list, video_path, fps=FIXED_FPS)
+    return video_path, current_seed
+
+# ----------------------
+# GRADIO UI - GRADIO 6 SYNTAX
+# ----------------------
+
+with gr.Blocks() as demo:
+    # Header with anycoder link
+    gr.Markdown(
+        """
+        # Wan2.1 I2V + Multi-LoRA Generator (NSFW LoRA repo)
+        Generate videos using LoRAs from `hotdogs/wan_nsfw_lora` with Wan2.1. Up to **30s** per video.
+        [Built with anycoder](https://huggingface.co/spaces/akhaliq/anycoder)
+        """
+    )
+
+    with gr.Row():
+        with gr.Column(scale=1):
+            gr.Markdown("### Input & Settings")
+            image_input = gr.Image(
+                type="pil", 
+                label="Input Image",
+                sources=["upload", "webcam", "clipboard"],
+                height=300
+            )
+            prompt = gr.Textbox(
+                label="Prompt", 
+                value=default_prompt_i2v,
+                placeholder="Describe the motion you want to see...",
+                lines=3
+            )
+            duration_slider = gr.Slider(
+                minimum=round(MIN_FRAMES_MODEL / FIXED_FPS, 1), 
+                maximum=30.0, 
+                step=0.1, 
+                value=2, 
+                label="Video Duration (seconds)"
+            )
+
+            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
+                )
+                random_seed_checkbox = gr.Checkbox(
+                    label="Randomize seed", 
+                    value=True
+                )
+                
+                with gr.Row():
+                    height_input = gr.Slider(
+                        SLIDER_MIN_H, 
+                        SLIDER_MAX_H, 
+                        MOD_VALUE, 
+                        value=DEFAULT_H_SLIDER_VALUE, 
+                        label="Output Height"
+                    )
+                    width_input = gr.Slider(
+                        SLIDER_MIN_W, 
+                        SLIDER_MAX_W, 
+                        MOD_VALUE, 
+                        value=DEFAULT_W_SLIDER_VALUE, 
+                        label="Output Width"
+                    )
+                
+                steps_slider = gr.Slider(
+                    1, 
+                    30, 
+                    step=1, 
+                    value=4, 
+                    label="Inference Steps"
+                )
+                guidance_slider = gr.Slider(
+                    0.0, 
+                    20.0, 
+                    step=0.5, 
+                    value=1.0, 
+                    label="Guidance Scale", 
+                    visible=False
+                )
+                lora_selector = gr.CheckboxGroup(
+                    choices=list(lora_paths.keys()), 
+                    label="Activate LoRA(s)",
+                    info="Select one or more LoRAs to apply"
+                )
+
+            generate_button = gr.Button(
+                "Generate Video", 
+                variant="primary",
+                size="lg"
+            )
+        
+        with gr.Column(scale=1):
+            gr.Markdown("### Output")
+            video_output = gr.Video(
+                label="Generated Video", 
+                autoplay=True,
+                height=400
+            )
+            seed_output = gr.Textbox(
+                label="Seed used", 
+                interactive=False,
+                info="Use this seed to reproduce the same video"
+            )
+
+    # Event handlers
+    image_input.upload(
+        fn=handle_image_upload, 
+        inputs=[image_input, height_input, width_input], 
+        outputs=[height_input, width_input],
+        api_visibility="private"
+    )
+    image_input.clear(
+        fn=handle_image_upload, 
+        inputs=[image_input, height_input, width_input], 
+        outputs=[height_input, width_input],
+        api_visibility="private"
+    )
+
+    inputs = [
+        image_input, prompt, height_input, width_input,
+        negative_prompt_input, duration_slider, guidance_slider,
+        steps_slider, seed_input, random_seed_checkbox, lora_selector
+    ]
+
+    generate_button.click(
+        fn=generate_video, 
+        inputs=inputs, 
+        outputs=[video_output, seed_output],
+        api_visibility="public",
+        concurrency_limit=1
+    )
+
+# ----------------------
+# LAUNCH - GRADIO 6 SYNTAX
+# ----------------------
+
+if __name__ == "__main__":
+    demo.queue().launch(
+        theme=gr.themes.Soft(
+            primary_hue="blue",
+            secondary_hue="indigo",
+            neutral_hue="slate",
+            font=gr.themes.GoogleFont("Inter"),
+            text_size="lg",
+            spacing_size="lg",
+            radius_size="md"
+        ).set(
+            button_primary_background_fill="*primary_600",
+            button_primary_background_fill_hover="*primary_700",
+            block_title_text_weight="600",
+            block_background_fill="*neutral_50"
+        ),
+        footer_links=[
+            {"label": "Built with anycoder", "url": "https://huggingface.co/spaces/akhaliq/anycoder"},
+            {"label": "Model", "url": f"https://huggingface.co/{MODEL_ID}"},
+            {"label": "LoRAs", "url": f"https://huggingface.co/{LORA_REPO_ID}"}
+        ],
+        show_error=True,
+        max_threads=40,
+        share=False
+    )

requirements.txt

@@ -0,0 +1,8 @@
+Pillow
+diffusers
+gradio
+huggingface_hub
+numpy
+spaces
+torch
+transformers