Upload 2 files

app.py

@@ -0,0 +1,156 @@
+import gradio as gr
+import torch
+import numpy as np
+from PIL import Image
+import cv2
+from diffusers import StableDiffusionControlNetPipeline, ControlNetModel, UniPCMultistepScheduler
+
+# ---------------------------------------------------------------------------
+# Model loading (cached – runs once on Space startup)
+# ---------------------------------------------------------------------------
+
+def load_pipeline():
+    controlnet = ControlNetModel.from_pretrained(
+        "lllyasviel/sd-controlnet-canny",
+        torch_dtype=torch.float16,
+    )
+    pipe = StableDiffusionControlNetPipeline.from_pretrained(
+        "runwayml/stable-diffusion-v1-5",
+        controlnet=controlnet,
+        torch_dtype=torch.float16,
+        safety_checker=None,
+    )
+    pipe.scheduler = UniPCMultistepScheduler.from_config(pipe.scheduler.config)
+    pipe.enable_model_cpu_offload()   # keeps VRAM usage low
+    return pipe
+
+pipe = load_pipeline()
+
+# ---------------------------------------------------------------------------
+# Helper: extract Canny edges
+# ---------------------------------------------------------------------------
+
+def extract_canny(image: Image.Image, low: int, high: int) -> Image.Image:
+    img_array = np.array(image.convert("RGB"))
+    edges = cv2.Canny(img_array, low, high)
+    edges_rgb = cv2.cvtColor(edges, cv2.COLOR_GRAY2RGB)
+    return Image.fromarray(edges_rgb)
+
+# ---------------------------------------------------------------------------
+# Main generation function
+# ---------------------------------------------------------------------------
+
+def generate(
+    input_image,
+    prompt,
+    negative_prompt,
+    canny_low,
+    canny_high,
+    guidance_scale,
+    steps,
+    seed,
+):
+    if input_image is None:
+        raise gr.Error("Bitte lade ein Bild hoch.")
+    if not prompt.strip():
+        raise gr.Error("Bitte gib einen Prompt ein.")
+
+    pil_image = Image.fromarray(input_image).resize((512, 512))
+    control_image = extract_canny(pil_image, int(canny_low), int(canny_high))
+
+    generator = torch.manual_seed(int(seed)) if seed >= 0 else None
+
+    result = pipe(
+        prompt=prompt,
+        negative_prompt=negative_prompt or None,
+        image=control_image,
+        num_inference_steps=int(steps),
+        guidance_scale=float(guidance_scale),
+        generator=generator,
+    ).images[0]
+
+    return control_image, result
+
+# ---------------------------------------------------------------------------
+# Gradio UI
+# ---------------------------------------------------------------------------
+
+css = """
+body { font-family: 'Inter', sans-serif; background: #0f0f11; color: #e8e8f0; }
+.gradio-container { max-width: 1100px; margin: 0 auto; }
+#title { text-align: center; padding: 2rem 0 0.5rem; }
+#title h1 { font-size: 2rem; font-weight: 700; letter-spacing: -0.5px;
+             background: linear-gradient(90deg, #a78bfa, #60a5fa); 
+             -webkit-background-clip: text; -webkit-text-fill-color: transparent; }
+#title p  { color: #9090a8; font-size: 0.95rem; margin-top: 0.25rem; }
+.panel    { background: #1a1a22; border: 1px solid #2a2a38; border-radius: 12px; padding: 1.25rem; }
+.generate-btn { background: linear-gradient(135deg, #7c3aed, #2563eb) !important;
+                color: white !important; border: none !important;
+                font-weight: 600 !important; font-size: 1rem !important;
+                border-radius: 8px !important; height: 48px !important; }
+.generate-btn:hover { opacity: 0.9 !important; }
+"""
+
+with gr.Blocks(css=css, title="ControlNet Canny") as demo:
+
+    gr.HTML("""
+        <div id="title">
+            <h1>⚡ ControlNet · Canny Edge</h1>
+            <p>Lade ein Bild hoch, schreib einen Prompt – und erzeuge ein neues Bild, das die Struktur deines Originals übernimmt.</p>
+        </div>
+    """)
+
+    with gr.Row():
+        # ---- Left column: inputs ----
+        with gr.Column(scale=1, elem_classes="panel"):
+            gr.Markdown("### 📥 Eingabe")
+            input_image = gr.Image(label="Referenzbild", type="numpy", height=300)
+
+            prompt = gr.Textbox(
+                label="Prompt",
+                placeholder="a futuristic city at night, neon lights, photorealistic, 8k",
+                lines=3,
+            )
+            negative_prompt = gr.Textbox(
+                label="Negative Prompt (optional)",
+                placeholder="blurry, low quality, watermark, deformed",
+                lines=2,
+            )
+
+            with gr.Accordion("⚙️ Erweiterte Einstellungen", open=False):
+                with gr.Row():
+                    canny_low  = gr.Slider(0, 255, value=100, step=1, label="Canny Low Threshold")
+                    canny_high = gr.Slider(0, 255, value=200, step=1, label="Canny High Threshold")
+                with gr.Row():
+                    guidance_scale = gr.Slider(1, 20, value=7.5, step=0.5, label="Guidance Scale")
+                    steps          = gr.Slider(10, 50, value=20, step=1,   label="Inference Steps")
+                seed = gr.Number(value=42, label="Seed (-1 = zufällig)", precision=0)
+
+            run_btn = gr.Button("🎨  Generieren", elem_classes="generate-btn")
+
+        # ---- Right column: outputs ----
+        with gr.Column(scale=1, elem_classes="panel"):
+            gr.Markdown("### 📤 Ergebnis")
+            canny_out  = gr.Image(label="Canny-Kantenbild", height=250)
+            result_out = gr.Image(label="Generiertes Bild",  height=350)
+
+    # ---- Example images ----
+    gr.Markdown("---")
+    gr.Markdown("### 💡 Beispiel-Prompts")
+    gr.Examples(
+        examples=[
+            [None, "a pencil sketch of a medieval castle, detailed, black and white", "blurry, color", 100, 200, 7.5, 20, 42],
+            [None, "a cyberpunk street scene, neon lights, rain, cinematic", "low quality", 80, 180, 8.0, 25, 0],
+            [None, "oil painting of a mountain landscape, impressionist style", "photo, realistic", 100, 200, 7.0, 20, 7],
+        ],
+        inputs=[input_image, prompt, negative_prompt, canny_low, canny_high, guidance_scale, steps, seed],
+        label="Zum Laden anklicken",
+    )
+
+    run_btn.click(
+        fn=generate,
+        inputs=[input_image, prompt, negative_prompt, canny_low, canny_high, guidance_scale, steps, seed],
+        outputs=[canny_out, result_out],
+    )
+
+demo.launch()

requirements.txt

@@ -0,0 +1,9 @@
+torch==2.1.0
+torchvision==0.16.0
+diffusers==0.25.0
+transformers==4.37.2
+accelerate==0.26.0
+opencv-python-headless==4.9.0.80
+Pillow==10.2.0
+numpy==1.26.3
+gradio==4.19.2