Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+examples/input_0.png filter=lfs diff=lfs merge=lfs -text
+examples/input_1.png filter=lfs diff=lfs merge=lfs -text
+examples/output.png filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,13 +1,28 @@
 ---
-title: Joyai Image Edit Plus
-emoji: 🌍
-colorFrom: purple
-colorTo: indigo
+title: JoyAI Image Edit Plus
+emoji: 🎨
+colorFrom: indigo
+colorTo: green
 sdk: gradio
-sdk_version: 6.19.0
-python_version: '3.12'
+sdk_version: 6.15.1
 app_file: app.py
-pinned: false
+short_description: Multi-image instruction-guided image editing
+python_version: "3.12"
+startup_duration_timeout: 1h
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+JoyAI-Image Edit Plus is a multi-image instruction-guided editing model from the JoyAI-Image family.
+Provide one or more reference images and a text instruction to generate a new image that combines
+elements from the references according to the instruction.
+
+## Model
+
+- **Model**: [jdopensource/JoyAI-Image-Edit-Plus-Diffusers](https://huggingface.co/jdopensource/JoyAI-Image-Edit-Plus-Diffusers)
+- **Architecture**: Qwen3-VL-8B text encoder + JoyImageEditPlusTransformer3D (16B) + AutoencoderKLWan VAE
+- **License**: Apache-2.0
+
+## How it works
+
+1. Upload 1–6 reference images in the gallery.
+2. Write an instruction describing how to combine or edit the images.
+3. The model generates a new image that follows your instruction using the references as context.

app.py

@@ -0,0 +1,197 @@
+"""JoyAI-Image Edit Plus — multi-image instruction-guided editing demo.
+
+Loads the jdopensource/JoyAI-Image-Edit-Plus-Diffusers model and exposes a
+Gradio interface where visitors provide one or more reference images and a
+text instruction; the model generates a new image that combines elements
+from the references according to the instruction.
+"""
+
+import os
+
+os.environ.setdefault("PYTORCH_CUDA_ALLOC_CONF", "expandable_segments:True")
+
+import spaces  # noqa: E402  -- must precede torch import
+
+import random  # noqa: E402
+
+import gradio as gr  # noqa: E402
+import torch  # noqa: E402
+from diffusers import JoyImageEditPlusPipeline  # noqa: E402
+
+MODEL_ID = "jdopensource/JoyAI-Image-Edit-Plus-Diffusers"
+
+pipe = JoyImageEditPlusPipeline.from_pretrained(
+    MODEL_ID,
+    torch_dtype=torch.bfloat16,
+)
+pipe.enable_model_cpu_offload()
+
+
+def _estimate_duration(images, prompt, num_inference_steps, guidance_scale, seed, negative_prompt, progress):
+    """Estimate the GPU duration needed based on step count and image count."""
+    n_imgs = len(images) if images else 0
+    # ~1s per step for 16B transformer + 8B text encoder + VAE, plus overhead
+    return min(300, 60 + int(num_inference_steps * 3.5) + n_imgs * 10)
+
+
+@spaces.GPU(duration=_estimate_duration)
+def generate(
+    images,
+    prompt: str,
+    negative_prompt: str,
+    num_inference_steps: int,
+    guidance_scale: float,
+    seed: int,
+    randomize_seed: bool,
+    progress=gr.Progress(track_tqdm=True),
+):
+    """Edit/generate an image from multiple reference images and a text instruction.
+
+    Args:
+        images: One or more reference images (1-6 supported).
+        prompt: Text instruction describing the desired edit or composition.
+        negative_prompt: Negative prompt to guide what to avoid.
+        num_inference_steps: Number of denoising steps (30 recommended).
+        guidance_scale: Classifier-free guidance scale (4.0 recommended).
+        seed: RNG seed for reproducibility.
+        randomize_seed: If True, pick a random seed each run.
+    """
+    if images is None or (isinstance(images, list) and len(images) == 0):
+        raise gr.Error("Please provide at least one reference image.")
+    if not prompt or not prompt.strip():
+        raise gr.Error("Please provide a text instruction.")
+
+    # Gradio Gallery returns list of file paths or dicts; normalize to PIL list
+    from PIL import Image
+
+    if isinstance(images, list):
+        pil_images = []
+        for img in images:
+            if isinstance(img, dict) and "path" in img:
+                pil_images.append(Image.open(img["path"]).convert("RGB"))
+            elif isinstance(img, str):
+                pil_images.append(Image.open(img).convert("RGB"))
+            elif isinstance(img, Image.Image):
+                pil_images.append(img.convert("RGB"))
+            else:
+                pil_images.append(Image.open(img).convert("RGB"))
+    elif isinstance(images, str):
+        pil_images = [Image.open(images).convert("RGB")]
+    else:
+        pil_images = [images.convert("RGB")]
+
+    if len(pil_images) > 6:
+        pil_images = pil_images[:6]
+
+    if randomize_seed:
+        seed = random.randint(0, 2**31 - 1)
+
+    # Determine output resolution from the last reference image
+    target_h, target_w = pipe.vae_image_processor.get_default_height_width(pil_images[-1])
+
+    generator = torch.Generator(device="cpu").manual_seed(int(seed))
+
+    result = pipe(
+        images=pil_images,
+        prompt=prompt,
+        negative_prompt=negative_prompt,
+        height=target_h,
+        width=target_w,
+        num_inference_steps=int(num_inference_steps),
+        guidance_scale=float(guidance_scale),
+        generator=generator,
+    )
+
+    output_image = result.images[0]
+    return output_image, gr.update(value=int(seed))
+
+
+CSS = """
+#col-container { max-width: 1100px; margin: 0 auto; }
+.dark .gradio-container { color: var(--body-text-color); }
+"""
+
+with gr.Blocks(theme=gr.themes.Citrus(), css=CSS) as demo:
+    gr.Markdown(
+        """
+        # JoyAI-Image Edit Plus
+        Multi-image instruction-guided editing — provide reference images and a text instruction
+        to generate a new image combining elements from the references.
+        [Model card](https://huggingface.co/jdopensource/JoyAI-Image-Edit-Plus-Diffusers)
+        """
+    )
+
+    with gr.Row(elem_id="col-container"):
+        with gr.Column(scale=1):
+            input_gallery = gr.Gallery(
+                label="Reference images",
+                show_label=True,
+                columns=3,
+                height=240,
+                object_fit="contain",
+                file_types=["image"],
+                type="filepath",
+            )
+            prompt = gr.Textbox(
+                label="Edit instruction",
+                placeholder="e.g. The woman is lovingly holding the cute puppy in her arms",
+                lines=2,
+            )
+            run_btn = gr.Button("Generate", variant="primary")
+
+            with gr.Accordion("Advanced settings", open=False):
+                negative_prompt = gr.Textbox(
+                    label="Negative prompt",
+                    value="low quality, blurry, deformed",
+                    lines=2,
+                )
+                num_inference_steps = gr.Slider(
+                    label="Inference steps",
+                    minimum=1,
+                    maximum=100,
+                    step=1,
+                    value=30,
+                )
+                guidance_scale = gr.Slider(
+                    label="Guidance scale",
+                    minimum=1.0,
+                    maximum=20.0,
+                    step=0.1,
+                    value=4.0,
+                )
+                randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
+                seed = gr.Number(label="Seed", value=42, precision=0)
+
+        with gr.Column(scale=1):
+            output_image = gr.Image(label="Result", show_label=True, height=420)
+
+    run_btn.click(
+        fn=generate,
+        inputs=[
+            input_gallery,
+            prompt,
+            negative_prompt,
+            num_inference_steps,
+            guidance_scale,
+            seed,
+            randomize_seed,
+        ],
+        outputs=[output_image, seed],
+        api_name="generate",
+    )
+
+    gr.Examples(
+        examples=[
+            {
+                "images": ["examples/input_0.png", "examples/input_1.png"],
+                "prompt": "The woman is lovingly holding the cute puppy in her arms",
+            },
+        ],
+        inputs=[input_gallery, prompt],
+        outputs=[output_image, seed],
+        fn=generate,
+        cache_examples=True,
+        cache_mode="lazy",
+    )
+
+demo.launch(mcp_server=True)

requirements.txt

@@ -0,0 +1,6 @@
+git+https://github.com/tangyanf/diffusers.git@add-joyimage-edit-plus
+transformers>=4.57.0
+accelerate
+sentencepiece
+pillow
+numpy