sketch - generator

Procfile

@@ -0,0 +1 @@
+web: python app.py

app.py

@@ -1,154 +1,79 @@
 import gradio as gr
-import numpy as np
-import random
-
-# import spaces #[uncomment to use ZeroGPU]
-from diffusers import DiffusionPipeline
+from diffusers import StableDiffusionPipeline
 import torch
 
+# Load models once at the start of the app for efficiency.
+# This prevents reloading the models for every new request, which
+# would be very slow.
 device = "cuda" if torch.cuda.is_available() else "cpu"
-model_repo_id = "stabilityai/sdxl-turbo"  # Replace to the model you would like to use
-
-if torch.cuda.is_available():
-    torch_dtype = torch.float16
-else:
-    torch_dtype = torch.float32
-
-pipe = DiffusionPipeline.from_pretrained(model_repo_id, torch_dtype=torch_dtype)
-pipe = pipe.to(device)
-
-MAX_SEED = np.iinfo(np.int32).max
-MAX_IMAGE_SIZE = 1024
-
-
-# @spaces.GPU #[uncomment to use ZeroGPU]
-def infer(
-    prompt,
-    negative_prompt,
-    seed,
-    randomize_seed,
-    width,
-    height,
-    guidance_scale,
-    num_inference_steps,
-    progress=gr.Progress(track_tqdm=True),
-):
-    if randomize_seed:
-        seed = random.randint(0, MAX_SEED)
-
-    generator = torch.Generator().manual_seed(seed)
-
-    image = pipe(
-        prompt=prompt,
-        negative_prompt=negative_prompt,
-        guidance_scale=guidance_scale,
-        num_inference_steps=num_inference_steps,
-        width=width,
-        height=height,
-        generator=generator,
-    ).images[0]
-
-    return image, seed
-
-
-examples = [
-    "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k",
-    "An astronaut riding a green horse",
-    "A delicious ceviche cheesecake slice",
-]
-
-css = """
-#col-container {
-    margin: 0 auto;
-    max-width: 640px;
-}
-"""
-
-with gr.Blocks(css=css) as demo:
-    with gr.Column(elem_id="col-container"):
-        gr.Markdown(" # Text-to-Image Gradio Template")
-
-        with gr.Row():
-            prompt = gr.Text(
-                label="Prompt",
-                show_label=False,
-                max_lines=1,
-                placeholder="Enter your prompt",
-                container=False,
-            )
-
-            run_button = gr.Button("Run", scale=0, variant="primary")
-
-        result = gr.Image(label="Result", show_label=False)
-
-        with gr.Accordion("Advanced Settings", open=False):
-            negative_prompt = gr.Text(
-                label="Negative prompt",
-                max_lines=1,
-                placeholder="Enter a negative prompt",
-                visible=False,
-            )
-
-            seed = gr.Slider(
-                label="Seed",
-                minimum=0,
-                maximum=MAX_SEED,
-                step=1,
-                value=0,
-            )
-
-            randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
-
-            with gr.Row():
-                width = gr.Slider(
-                    label="Width",
-                    minimum=256,
-                    maximum=MAX_IMAGE_SIZE,
-                    step=32,
-                    value=1024,  # Replace with defaults that work for your model
-                )
-
-                height = gr.Slider(
-                    label="Height",
-                    minimum=256,
-                    maximum=MAX_IMAGE_SIZE,
-                    step=32,
-                    value=1024,  # Replace with defaults that work for your model
-                )
-
-            with gr.Row():
-                guidance_scale = gr.Slider(
-                    label="Guidance scale",
-                    minimum=0.0,
-                    maximum=10.0,
-                    step=0.1,
-                    value=0.0,  # Replace with defaults that work for your model
-                )
-
-                num_inference_steps = gr.Slider(
-                    label="Number of inference steps",
-                    minimum=1,
-                    maximum=50,
-                    step=1,
-                    value=2,  # Replace with defaults that work for your model
-                )
-
-        gr.Examples(examples=examples, inputs=[prompt])
-    gr.on(
-        triggers=[run_button.click, prompt.submit],
-        fn=infer,
-        inputs=[
-            prompt,
-            negative_prompt,
-            seed,
-            randomize_seed,
-            width,
-            height,
-            guidance_scale,
-            num_inference_steps,
-        ],
-        outputs=[result, seed],
+dtype = torch.float16 if torch.cuda.is_available() and device == "cuda" else torch.float32
+
+# Stage 1: Text-to-Sketch model
+# We use a base Stable Diffusion pipeline with a special prompt
+# to generate a line drawing effect.
+try:
+    sketch_pipeline = StableDiffusionPipeline.from_pretrained(
+        "runwayml/stable-diffusion-v1-5",
+        torch_dtype=dtype
+    )
+    sketch_pipeline.to(device)
+except Exception as e:
+    print(f"Error loading sketch pipeline: {e}")
+    sketch_pipeline = None
+
+# Stage 2: Sketch-to-Image model
+# This pipeline is loaded with the Stable Diffusion base and then
+# a LoRA model is attached to handle the sketch-to-image conversion.
+try:
+    image_pipeline = StableDiffusionPipeline.from_pretrained(
+        "runwayml/stable-diffusion-v1-5",
+        torch_dtype=dtype
+    )
+    image_pipeline.load_lora("gokaygokay/Sketch-to-Image-Kontext-Dev-LoRA", lora_weights_name="model.safetensors")
+    image_pipeline.to(device)
+except Exception as e:
+    print(f"Error loading image pipeline or LoRA: {e}")
+    image_pipeline = None
+
+# The main function that connects the two stages
+def generate_full_image(text_prompt):
+    if not sketch_pipeline or not image_pipeline:
+        return None, None
+        
+    # Step 1: Generate the sketch from the text prompt
+    # The "line drawing" prompt helps steer the model's output
+    sketch_prompt = f"line drawing of a {text_prompt}"
+    sketch = sketch_pipeline(sketch_prompt).images[0]
+
+    # Step 2: Generate the final image from the sketch
+    # The 'image' input to the pipeline uses the generated sketch
+    final_image = image_pipeline(image=sketch, prompt="a realistic human portrait").images[0]
+
+    return sketch, final_image
+
+# Define the Gradio UI using Blocks for a custom layout
+with gr.Blocks(title="Sketch-to-Image Pipeline") as demo:
+    gr.Markdown("# Text-to-Sketch-to-Portrait")
+    gr.Markdown("Enter a description to generate a sketch, which is then converted into a realistic human portrait.")
+    
+    with gr.Row():
+        text_input = gr.Textbox(
+            label="Person Description",
+            placeholder="e.g., A middle-aged man with a scar on his right cheek and shaggy hair"
+        )
+    
+    generate_button = gr.Button("Generate Portrait")
+    
+    with gr.Row():
+        sketch_output = gr.Image(label="Generated Sketch", type="pil")
+        final_image_output = gr.Image(label="Generated Portrait", type="pil")
+
+    # Connect the UI components to the Python function
+    generate_button.click(
+        fn=generate_full_image,
+        inputs=text_input,
+        outputs=[sketch_output, final_image_output]
     )
 
-if __name__ == "__main__":
-    demo.launch()
+# Launch the app
+demo.launch()

requirements.txt

@@ -3,4 +3,5 @@ diffusers
 invisible_watermark
 torch
 transformers
-xformers
+xformers
+gradio