add ovis-image

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

CHANGELOG.md

@@ -1,13 +1,16 @@
 # Change Log for SD.Next
 
-## Known issues
+## Update for 2025-12-26
 
-- z-image-turbo controlnet device mismatch: <https://github.com/huggingface/diffusers/pull/12886>
-- z-image-turbo safetensors loader: <https://github.com/huggingface/diffusers/issues/12887>
-- kandinsky-image-5 hardcoded cuda: <https://github.com/huggingface/diffusers/pull/12814>
-- peft lora with torch-rocm-windows: <https://github.com/huggingface/peft/pull/2963>
+### Highlights for 2025-12-26
 
-## Update for 2025-12-25
+End of year release with:
+- Several new models including highly anticipated **Qwen-Image-Edit 2511** as well as **Qwen-Image-Layered**, **LongCat Image** and **Ovis Image**  
+- New features including support for **Z-Image** *ControlNets* and *fine-tunes* and **Detailer** segmentation support  
+
+[ReadMe](https://github.com/vladmandic/automatic/blob/master/README.md) | [ChangeLog](https://github.com/vladmandic/automatic/blob/master/CHANGELOG.md) | [Docs](https://vladmandic.github.io/sdnext-docs/) | [WiKi](https://github.com/vladmandic/automatic/wiki) | [Discord](https://discord.com/invite/sd-next-federal-batch-inspectors-1101998836328697867) | [Sponsor](https://github.com/sponsors/vladmandic)  
+
+### Details for 2025-12-26
 
 - **Models**
   - [LongCat Image](https://github.com/meituan-longcat/LongCat-Image) in *Image* and *Image Edit* variants  
@@ -17,6 +20,8 @@
   - [Qwen-Image-Layered](https://huggingface.co/Qwen/Qwen-Image-Layered) in *base* and *pre-quantized* variants  
     Qwen-Image-Layered, a model capable of decomposing an image into multiple RGBA layers  
     *note*: set number of desired output layers in *settings -> model options*  
+  - [Ovis Image 7B](https://huggingface.co/AIDC-AI/Ovis-Image-7B)  
+    Ovis Image is a new text-to-image base model based on Qwen3 text-encoder and optimized for text-rendering  
 - **Features**
   - Google **Gemini** and **Veo** models support for both *Dev* and *Vertex* access methods  
     see [docs](https://vladmandic.github.io/sdnext-docs/Google-GenAI/) for details  
@@ -45,6 +50,7 @@
   - control input media with non-english locales  
   - handle embeds when on meta device  
   - improve offloading when model has manual modules  
+  - fix ui section colapsible state, thanks @awsr  
 
 ## Update for 2025-12-11
 

TODO.md

@@ -1,5 +1,12 @@
 # TODO
 
+## Known issues
+
+- z-image-turbo controlnet device mismatch: <https://github.com/huggingface/diffusers/pull/12886>
+- z-image-turbo safetensors loader: <https://github.com/huggingface/diffusers/issues/12887>
+- kandinsky-image-5 hardcoded cuda: <https://github.com/huggingface/diffusers/pull/12814>
+- peft lora with torch-rocm-windows: <https://github.com/huggingface/peft/pull/2963>
+
 ## Project Board
 
 - <https://github.com/users/vladmandic/projects>

extensions-builtin/sd-extension-chainner

@@ -1 +1 @@
-Subproject commit c6dc85eb28a02bc7af268497b7a5a596770c5d7b
+Subproject commit 2a7005fbcf8985644b66121365fa7228a65f34b0

extensions-builtin/sdnext-modernui

@@ -1 +1 @@
-Subproject commit af99fbab29e9a424c4e79fa8e4ae194481cb5f75
+Subproject commit 50f7613276ecb37015cc4dd270a033d32d038415

html/reference.json

@@ -938,6 +938,16 @@
     "date": "2024 January"
   },
 
+  "AIDC Ovis-Image 7B": {
+    "path": "AIDC-AI/Ovis-Image-7B",
+    "skip": true,
+    "desc": "Built upon Ovis-U1, Ovis-Image is a 7B text-to-image model specifically optimized for high-quality text rendering, designed to operate efficiently under stringent computational constraints.",
+    "preview": "AIDC-AI--Ovis-Image-7B.jpg",
+    "size": 23.38,
+    "date": "2025 December",
+    "extras": ""
+  },
+
   "HDM-XUT 340M Anime": {
     "path": "KBlueLeaf/HDM-xut-340M-anime",
     "skip": true,

installer.py

@@ -648,7 +648,7 @@ def check_diffusers():
     t_start = time.time()
     if args.skip_all:
         return
-    sha = '52766e6a6939ac6e74375bde5e19c5e0b90d24c1' # diffusers commit hash
+    sha = 'f6b6a7181eb44f0120b29cd897c129275f366c2a' # diffusers commit hash
     # if args.use_rocm or args.use_zluda or args.use_directml:
     #     sha = '043ab2520f6a19fce78e6e060a68dbc947edb9f9' # lock diffusers versions for now
     pkg = pkg_resources.working_set.by_key.get('diffusers', None)

modules/modeldata.py

@@ -78,6 +78,8 @@ def get_model_type(pipe):
         model_type = 'prx'
     elif 'LongCat' in name:
         model_type = 'longcat'
+    elif 'Ovis-Image' in name:
+        model_type = 'ovis'
     # video models
     elif "CogVideo" in name:
         model_type = 'cogvideo'

modules/sd_detect.py

@@ -141,6 +141,8 @@ def guess_by_name(fn, current_guess):
         new_guess = 'Z-Image'
     elif 'longcat-image' in fn.lower():
         new_guess = 'LongCat'
+    elif 'ovis-image' in fn.lower():
+        new_guess = 'Ovis-Image'
     if debug_load:
         shared.log.trace(f'Autodetect: method=name file="{fn}" previous="{current_guess}" current="{new_guess}"')
     return new_guess or current_guess

modules/sd_models.py

@@ -479,6 +479,10 @@ def load_diffuser_force(detected_model_type, checkpoint_info, diffusers_load_con
             from pipelines.model_longcat import load_longcat
             sd_model = load_longcat(checkpoint_info, diffusers_load_config)
             allow_post_quant = False
+        elif model_type in ['Overfit']:
+            from pipelines.model_ovis import load_ovis
+            sd_model = load_ovis(checkpoint_info, diffusers_load_config)
+            allow_post_quant = False
     except Exception as e:
         shared.log.error(f'Load {op}: path="{checkpoint_info.path}" {e}')
         if debug_load:

pipelines/model_ovis.py

@@ -0,0 +1,36 @@
+import transformers
+import diffusers
+from modules import shared, devices, sd_models, model_quant, sd_hijack_te
+from pipelines import generic
+
+
+def load_ovis(checkpoint_info, diffusers_load_config=None):
+    if diffusers_load_config is None:
+        diffusers_load_config = {}
+    repo_id = sd_models.path_to_repo(checkpoint_info)
+    sd_models.hf_auth_check(checkpoint_info)
+
+    load_args, _quant_args = model_quant.get_dit_args(diffusers_load_config, allow_quant=False)
+    shared.log.debug(f'Load model: type=OvisImage repo="{repo_id}" config={diffusers_load_config} offload={shared.opts.diffusers_offload_mode} dtype={devices.dtype} args={diffusers_load_config}')
+
+    transformer = generic.load_transformer(repo_id, cls_name=diffusers.OvisImageTransformer2DModel, load_config=diffusers_load_config)
+    text_encoder = generic.load_text_encoder(repo_id, cls_name=transformers.Qwen3Model, load_config=diffusers_load_config)
+
+    pipe = diffusers.OvisImagePipeline.from_pretrained(
+        repo_id,
+        cache_dir=shared.opts.diffusers_dir,
+        transformer=transformer,
+        text_encoder=text_encoder,
+        **load_args,
+    )
+
+    pipe.task_args = {
+        'output_type': 'np',
+    }
+
+    del transformer
+    del text_encoder
+    sd_hijack_te.init_hijack(pipe)
+
+    devices.torch_gc(force=True, reason='load')
+    return pipe

pipelines/model_z_image.py

@@ -11,7 +11,7 @@ def load_z_image(checkpoint_info, diffusers_load_config=None):
     sd_models.hf_auth_check(checkpoint_info)
 
     load_args, _quant_args = model_quant.get_dit_args(diffusers_load_config, allow_quant=False)
-    shared.log.debug(f'Load model: type=Z-Image repo="{repo_id}" config={diffusers_load_config} offload={shared.opts.diffusers_offload_mode} dtype={devices.dtype} args={diffusers_load_config}')
+    shared.log.debug(f'Load model: type=ZImage repo="{repo_id}" config={diffusers_load_config} offload={shared.opts.diffusers_offload_mode} dtype={devices.dtype} args={diffusers_load_config}')
 
     transformer = generic.load_transformer(repo_id, cls_name=diffusers.ZImageTransformer2DModel, load_config=diffusers_load_config)
     text_encoder = generic.load_text_encoder(repo_id, cls_name=transformers.Qwen3ForCausalLM, load_config=diffusers_load_config)