flux-kontext with variable resolution

Signed-off-by: Vladimir Mandic <mandic00@live.com>

CHANGELOG.md

@@ -4,7 +4,8 @@
 
 ### Highlights for 2025-08-08
 
-Several new and updated models: [Qwen-Image](https://qwenlm.github.io/blog/qwen-image/), [FLUX.1-Krea-Dev](https://www.krea.ai/blog/flux-krea-open-source-release), [Chroma](https://huggingface.co/lodestones/Chroma), [SkyReels-V2](https://huggingface.co/Skywork/SkyReels-V2-DF-14B-720P-Diffusers), [Wan-VACE](https://huggingface.co/Wan-AI/Wan2.1-VACE-14B-diffusers)  
+Several new models: [Qwen-Image](https://qwenlm.github.io/blog/qwen-image/), [FLUX.1-Krea-Dev](https://www.krea.ai/blog/flux-krea-open-source-release)  
+And several updated models: [Chroma](https://huggingface.co/lodestones/Chroma), [SkyReels-V2](https://huggingface.co/Skywork/SkyReels-V2-DF-14B-720P-Diffusers), [Wan-VACE](https://huggingface.co/Wan-AI/Wan2.1-VACE-14B-diffusers)  
 Plus continuing with major **UI** work, there is new embedded **Docs/Wiki** search, redesigned real-time **hints**, built-in **GPU monitor**, **CivitAI** integration and more!  
 On the compute side, new profiles for high-vram GPUs and offloading improvements  
 And (*as always*) many bugfixes and improvements to existing features!  
@@ -91,7 +92,7 @@ And (*as always*) many bugfixes and improvements to existing features!
 - **Fixes**  
   - refactor legacy processing loop  
   - fix settings components mismatch  
-  - fix Wan 2.2-5B I2V workflow  
+  - fix *Wan 2.2-5B I2V* workflow  
   - fix OpenVINO  
   - fix video model vs pipeline mismatch  
   - fix video generic save frames  
@@ -100,8 +101,9 @@ And (*as always*) many bugfixes and improvements to existing features!
   - fix progress bar with refine/detailer  
   - fix api progress reporting endpoint  
   - fix openvino backend failing to compile  
-  - fix nunchaku fallback on unsupported model  
-  - fix nunchaku windows download links  
+  - fix `nunchaku` fallback on unsupported model  
+  - fix `nunchaku` windows download links  
+  - fix *Flux.1-Kontext-Dev* with variable resolution  
   - use `utf_16_be` as primary metadata decoding  
   - reapply offloading on ipadapter load  
   - api set default script-name  

modules/processing_args.py

@@ -99,6 +99,8 @@ def task_specific_kwargs(p, model):
             'height': height,
             'width': width,
         }
+
+    # model specific args
     if model.__class__.__name__ == 'LatentConsistencyModelPipeline' and hasattr(p, 'init_images') and len(p.init_images) > 0:
         p.ops.append('lcm')
         init_latents = [processing_vae.vae_encode(image, model=shared.sd_model, vae_type=p.vae_type).squeeze(dim=0) for image in p.init_images]
@@ -120,6 +122,8 @@ def task_specific_kwargs(p, model):
             'target_subject_category': getattr(p, 'prompt', '').split()[-1],
             'output_type': 'pil',
         }
+
+# TODO
     if debug_enabled:
         debug_log(f'Process task specific args: {task_args}')
     return task_args
@@ -388,6 +392,8 @@ def set_pipeline_args(p, model, prompts:list, negative_prompts:list, prompts_2:t
             else:
                 args['width'] = 8 * math.ceil(args['image'][0].width / 8)
                 args['height'] = 8 * math.ceil(args['image'][0].height / 8)
+    if 'max_area' in possible and 'width' in args and 'height' in args and 'max_area' not in args:
+        args['max_area'] = args['width'] * args['height']
 
     # handle implicit controlnet
     if 'control_image' in possible and 'control_image' not in args and 'image' in args: