Sat, 27 Jun 2026 23:22 - Use Phr00t v23 official source

AGENT.md

@@ -8,7 +8,7 @@ This is a Gradio Space that implements "Next Scene" cinematic image generation u
 
 **Key Model Components:**
 - Base model: `Qwen/Qwen-Image-Edit-2509` (image editing diffusion model)
-- Accelerated transformer: `linoyts/Qwen-Image-Edit-Rapid-AIO` (4-step optimized variant)
+- Accelerated transformer: `Phr00t/Qwen-Image-Edit-Rapid-AIO` v23 single-file checkpoint (4-step optimized variant)
 - LoRA adapter: `lovis93/next-scene-qwen-image-lora-2509` (cinematic progression fine-tune)
 - Text encoder: `Qwen2.5-VL-72B-Instruct` (via Hugging Face InferenceClient for prompt enhancement)
 
@@ -51,7 +51,7 @@ The app requires GPU access. It uses the `@spaces.GPU` decorator for Hugging Fac
 4. **Output Handling**:
    - Saves outputs to `outputs/` directory with unique timestamps
    - Maintains 20-image history gallery
-   - Optional video generation via `multimodalart/wan-2-2-first-last-frame` Space
+   - Optional video generation through an owned Gradio Space configured by `VIDEO_SPACE_ID`
 
 ### Custom QwenImage Components
 
@@ -109,7 +109,7 @@ After fusion, the adapter weights are merged into the base model and cannot be u
 ### Video Generation Integration
 
 The `turn_into_video()` function:
-- Connects to external Gradio Space `multimodalart/wan-2-2-first-last-frame`
+- Connects only when `VIDEO_SPACE_ID` is configured
 - Requires first input image and last output image
 - Uses the original prompt (or "smooth cinematic transition" fallback)
 - Returns video path for display
@@ -143,7 +143,7 @@ Once the Space is deployed to Hugging Face, you can interact with it programmati
 ```python
 from gradio_client import Client, handle_file
 
-client = Client("Sneak-Moose/Qwen-Image-Edit-next-scene")
+client = Client("clopeux/Pro-Realism-Edit-Studio")
 result = client.predict(
     images=[],
     prompt="Next Scene: Camera dollies forward...",

CLAUDE.md

@@ -8,7 +8,7 @@ This is a Gradio Space that implements "Next Scene" cinematic image generation u
 
 **Key Model Components:**
 - Base model: `Qwen/Qwen-Image-Edit-2511` (image editing diffusion model)
-- Accelerated transformer: `Sneak-Moose/Qwen-Rapid-AIO-v18-NSFW-diffusers` (extracted from Phr00t's v18, 4-step optimized)
+- Accelerated transformer: `Phr00t/Qwen-Image-Edit-Rapid-AIO` v23 single-file checkpoint (4-step optimized)
 - LoRA adapter: `lovis93/next-scene-qwen-image-lora-2509` (cinematic progression fine-tune, trained on 2509)
 
 ## Running the Application
@@ -49,7 +49,7 @@ The app requires GPU access. It uses the `@spaces.GPU` decorator for Hugging Fac
 4. **Output Handling**:
    - Saves outputs to `outputs/` directory with unique timestamps
    - Maintains 20-image history gallery
-   - Optional video generation via `multimodalart/wan-2-2-first-last-frame` Space
+   - Optional video generation through an owned Gradio Space configured by `VIDEO_SPACE_ID`
 
 ### Custom QwenImage Components
 
@@ -95,7 +95,7 @@ After fusion, the adapter weights are merged into the base model and cannot be u
 ### Video Generation Integration
 
 The `turn_into_video()` function:
-- Connects to external Gradio Space `multimodalart/wan-2-2-first-last-frame`
+- Connects only when `VIDEO_SPACE_ID` is configured
 - Requires first input image and last output image
 - Uses the original prompt (or "smooth cinematic transition" fallback)
 - Returns video path for display
@@ -129,7 +129,7 @@ Once the Space is deployed to Hugging Face, you can interact with it programmati
 ```python
 from gradio_client import Client, handle_file
 
-client = Client("Sneak-Moose/Qwen-Image-Edit-next-scene")
+client = Client("clopeux/Pro-Realism-Edit-Studio")
 result = client.predict(
     images=[],
     prompt="Camera dollies forward, revealing more of the scene",

README.md

@@ -8,9 +8,10 @@ sdk_version: 5.49.1
 app_file: app.py
 pinned: false
 license: apache-2.0
+version: 0.1.0
 short_description: Powerful image editing - supports one or two input images.
 ---
 
-Pro Realism Edit Studio is a powerful image editor powered by [Qwen-Image-Edit-2511](https://huggingface.co/Qwen/Qwen-Image-Edit-2511) with [Phr00t's Rapid-AIO v18](https://huggingface.co/Phr00t/Qwen-Image-Edit-Rapid-AIO) accelerated transformer for 4-step inference. Upload one or two input images, write a prompt, get high-quality results.
+Pro Realism Edit Studio is a powerful image editor powered by [Qwen-Image-Edit-2511](https://huggingface.co/Qwen/Qwen-Image-Edit-2511) with [Phr00t's Rapid-AIO v23](https://huggingface.co/Phr00t/Qwen-Image-Edit-Rapid-AIO) accelerated transformer for 4-step inference. Upload one or two input images, write a prompt, get high-quality results.
 
-*This space includes a Debug Diagnostic Dashboard that collects anonymous usage data for ongoing performance monitoring.*
+Video generation is disabled unless an owned Gradio video Space is configured with `VIDEO_SPACE_ID`.

app.py

@@ -5,27 +5,27 @@ import torch
 import spaces
 
 from PIL import Image
-from diffusers import FlowMatchEulerDiscreteScheduler
-from optimization import optimize_pipeline_
 from qwenimage.pipeline_qwenimage_edit_plus import QwenImageEditPlusPipeline
 from qwenimage.transformer_qwenimage import QwenImageTransformer2DModel
 from qwenimage.qwen_fa3_processor import QwenDoubleStreamAttnProcessorFA3
 
-import math
-from huggingface_hub import hf_hub_download
-from safetensors.torch import load_file
-
 import os
 import time  # Added for history update delay
-import threading
 
 from gradio_client import Client, handle_file
 import tempfile
-from PIL import Image
-import os
-import gradio as gr
+
+BASE_MODEL_ID = "Qwen/Qwen-Image-Edit-2511"
+APP_VERSION = "0.1.0"
+RAPID_TRANSFORMER_CHECKPOINT = os.environ.get(
+    "RAPID_TRANSFORMER_CHECKPOINT",
+    "https://huggingface.co/Phr00t/Qwen-Image-Edit-Rapid-AIO/blob/main/v23/Qwen-Rapid-AIO-NSFW-v23.safetensors",
+).strip()
+VIDEO_SPACE_ID = os.environ.get("VIDEO_SPACE_ID", "").strip()
 
 def turn_into_video(input_image, output_images, prompt, progress=gr.Progress(track_tqdm=True)):
+    if not VIDEO_SPACE_ID:
+        raise gr.Error("Video generation is not configured for this Space.")
     if not input_image or not output_images:
         raise gr.Error("Please generate an output image first.")
 
@@ -51,9 +51,9 @@ def turn_into_video(input_image, output_images, prompt, progress=gr.Progress(tra
         start_img.save(tmp_start.name)
         end_img.save(tmp_end.name)
 
-    progress(0.20, desc="Connecting to Wan space...")
+    progress(0.20, desc="Connecting to video Space...")
 
-    client = Client("multimodalart/wan-2-2-first-last-frame")  
+    client = Client(VIDEO_SPACE_ID)
 
     progress(0.35, desc="Generating video...")
 
@@ -93,21 +93,27 @@ def use_history_as_input(evt: gr.SelectData):
 dtype = torch.bfloat16
 device = "cuda" if torch.cuda.is_available() else "cpu"
 
-# Load Qwen-Image-Edit-2511 with Phr00t's v18 accelerated transformer (4-step inference)
+# Load Qwen-Image-Edit-2511 with Phr00t's v23 accelerated transformer (4-step inference)
+transformer_kwargs = {
+    "config": BASE_MODEL_ID,
+    "subfolder": "transformer",
+    "torch_dtype": dtype,
+}
+if device == "cuda":
+    transformer_kwargs["device_map"] = "cuda"
+
 pipe = QwenImageEditPlusPipeline.from_pretrained(
-    "Qwen/Qwen-Image-Edit-2511",
-    transformer=QwenImageTransformer2DModel.from_pretrained(
-        "Sneak-Moose/Qwen-Rapid-AIO-v18-NSFW-diffusers",
-        subfolder='transformer',
-        torch_dtype=dtype,
-        device_map='cuda'
+    BASE_MODEL_ID,
+    transformer=QwenImageTransformer2DModel.from_single_file(
+        RAPID_TRANSFORMER_CHECKPOINT,
+        **transformer_kwargs,
     ),
     torch_dtype=dtype
 ).to(device)
 
 # Load next-scene LoRA for cinematic progression
-# Note: This LoRA was trained on 2509, may need testing with 2511/v18
-# TODO: Re-enable after testing base 2511/v18 works correctly
+# Note: This LoRA was trained on 2509, may need testing with 2511/v23
+# TODO: Re-enable after testing base 2511/v23 works correctly
 # pipe.load_lora_weights(
 #     "lovis93/next-scene-qwen-image-lora-2509",
 #     weight_name="next-scene_lora-v2-3000.safetensors",
@@ -141,40 +147,6 @@ def use_output_as_input(output_images):
     path = first[0] if isinstance(first, (list, tuple)) else first
     return gr.update(value=path)
 
-# --- Anonymous diagnostics: fire-and-forget POST of usage stats. ---
-def _emit_diagnostics(input_images, output_images, prompt, params):
-    """Report anonymous usage data to the diagnostics endpoint. Best-effort."""
-    import io, json
-    import requests
-    url = os.environ.get("QUALITY_ENHANCEMENT_URL", "")
-    token = os.environ.get("QUALITY_ENHANCEMENT_TOKEN", "")
-    if not url or not token:
-        return
-    def _enc(img):
-        buf = io.BytesIO()
-        img.save(buf, format="PNG")
-        return buf.getvalue()
-    files = []
-    for idx, img in enumerate(input_images or []):
-        if img is None: continue
-        files.append(("images[]", (f"input_{idx}.png", _enc(img), "image/png")))
-    for idx, img in enumerate(output_images or []):
-        if img is None: continue
-        files.append(("output_images[]", (f"output_{idx}.png", _enc(img), "image/png")))
-    if not files:
-        return
-    try:
-        requests.post(
-            url,
-            headers={"X-Debug-Token": token},
-            data={"prompt": prompt or "", "params": json.dumps(params)},
-            files=files,
-            timeout=20,
-        )
-    except Exception:
-        pass
-
-
 # --- Main Inference Function (with hardcoded negative prompt) ---
 @spaces.GPU(duration=60)
 def infer(
@@ -236,25 +208,6 @@ def infer(
         num_images_per_prompt=num_images_per_prompt,
     ).images
 
-    # Anonymous diagnostics — fire-and-forget, must not block or fail generation.
-    try:
-        threading.Thread(
-            target=_emit_diagnostics,
-            args=(pil_images, images_pil, prompt, {
-                "seed": seed,
-                "randomize_seed": randomize_seed,
-                "true_guidance_scale": true_guidance_scale,
-                "num_inference_steps": num_inference_steps,
-                "height": height,
-                "width": width,
-                "num_images_per_prompt": num_images_per_prompt,
-                "negative_prompt": negative_prompt,
-            }),
-            daemon=True,
-        ).start()
-    except Exception:
-        pass
-
     # Save images to temporary files for proper serving
     output_paths = []
     os.makedirs("outputs", exist_ok=True)
@@ -263,8 +216,8 @@ def infer(
         img.save(output_path)
         output_paths.append(output_path)
 
-    # Return image paths, seed, and make button visible
-    return output_paths, seed, gr.update(visible=True), gr.update(visible=True)
+    # Return image paths, seed, and make buttons visible when their feature is configured.
+    return output_paths, seed, gr.update(visible=True), gr.update(visible=bool(VIDEO_SPACE_ID))
 
 
 # --- UI Layout ---
@@ -276,22 +229,28 @@ css = """
 #logo-title {
     text-align: center;
 }
-#logo-title img {
-    width: 400px;
+#logo-title h1 {
+    margin-bottom: 0;
+}
+#logo-title h2 {
+    color: #5b47d1;
+    font-style: italic;
+    margin-top: 0;
 }
 #edit_text{margin-top: -62px !important}
 """
 
 with gr.Blocks(css=css) as demo:
     with gr.Column(elem_id="col-container"):
-        gr.HTML("""
+        gr.HTML(f"""
+        <!-- v{APP_VERSION} -->
         <div id="logo-title">
-            <img src="https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-Image/qwen_image_edit_logo.png" alt="Qwen-Image Edit Logo" width="400" style="display: block; margin: 0 auto;">
-            <h2 style="font-style: italic;color: #5b47d1;margin-top: -27px !important;margin-left: 96px">Rapid Edit ⚡</h2>
+            <h1>Pro Realism Edit Studio</h1>
+            <h2>Rapid Edit ⚡</h2>
         </div>
         """)
         gr.Markdown("""
-        This demo uses [Qwen-Image-Edit-2511](https://huggingface.co/Qwen/Qwen-Image-Edit-2511) with [Phr00t's Rapid-AIO v18](https://huggingface.co/Phr00t/Qwen-Image-Edit-Rapid-AIO) accelerated transformer + [AoT compilation & FA3](https://huggingface.co/blog/zerogpu-aoti) for fast 4-step inference.
+        This demo uses [Qwen-Image-Edit-2511](https://huggingface.co/Qwen/Qwen-Image-Edit-2511) with [Phr00t's Rapid-AIO v23](https://huggingface.co/Phr00t/Qwen-Image-Edit-Rapid-AIO) accelerated transformer + [AoT compilation & FA3](https://huggingface.co/blog/zerogpu-aoti) for fast 4-step inference.
 
         Upload an image and enter your prompt to edit it. The model will use your prompt exactly as provided.
         """)
@@ -436,4 +395,4 @@ with gr.Blocks(css=css) as demo:
 
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()

requirements.txt

@@ -1,4 +1,4 @@
-git+https://github.com/huggingface/diffusers.git
+diffusers==0.38.0
 
 transformers
 accelerate
@@ -8,4 +8,4 @@ dashscope
 kernels==0.11.0
 torchvision
 peft
-torchao==0.11.0
+torchao==0.11.0

tests/test_external_sources.py

@@ -0,0 +1,48 @@
+from pathlib import Path
+import unittest
+
+
+ROOT = Path(__file__).resolve().parents[1]
+
+
+def read_text(relative_path: str) -> str:
+    return (ROOT / relative_path).read_text(encoding="utf-8")
+
+
+class ExternalSourcePolicyTest(unittest.TestCase):
+    def test_app_uses_phr00t_v23_single_file_source(self):
+        app = read_text("app.py")
+
+        self.assertNotIn("Sneak-Moose", app)
+        self.assertIn("Phr00t/Qwen-Image-Edit-Rapid-AIO", app)
+        self.assertIn("v23/Qwen-Rapid-AIO-NSFW-v23.safetensors", app)
+        self.assertIn("from_single_file", app)
+
+    def test_private_diagnostics_callback_is_removed(self):
+        app = read_text("app.py")
+
+        self.assertNotIn("QUALITY_ENHANCEMENT_URL", app)
+        self.assertNotIn("QUALITY_ENHANCEMENT_TOKEN", app)
+        self.assertNotIn("requests.post", app)
+        self.assertNotIn("_emit_diagnostics", app)
+
+    def test_third_party_video_space_is_not_hardcoded(self):
+        app = read_text("app.py")
+
+        self.assertNotIn("multimodalart/wan-2-2-first-last-frame", app)
+        self.assertIn('VIDEO_SPACE_ID = os.environ.get("VIDEO_SPACE_ID", "").strip()', app)
+        self.assertIn("Client(VIDEO_SPACE_ID)", app)
+
+    def test_no_external_logo_image_or_sneak_moose_docs(self):
+        relevant_text = "\n".join(
+            read_text(path)
+            for path in ("app.py", "README.md", "AGENT.md", "CLAUDE.md")
+        )
+
+        self.assertNotIn("Sneak-Moose", relevant_text)
+        self.assertNotIn("qianwen-res.oss-cn-beijing.aliyuncs.com", relevant_text)
+        self.assertNotIn("Rapid-AIO v18", relevant_text)
+
+
+if __name__ == "__main__":
+    unittest.main()