The official subgraph workflow for dual-image editing in Flux2 has completely failed. It can only perform single-image editing. The images produced by the dual-image editing are completely unrelated to the reference image?

### Custom Node Testing

- [ ] I have tried disabling custom nodes and the issue persists (see [how to disable custom nodes](https://docs.comfy.org/troubleshooting/custom-node-issues#step-1%3A-test-with-all-custom-nodes-disabled) if you need help)

### Expected Behavior

Please handle it as soon as possible, including the issue of the random seeds in the subgraphs failing. I have to manually extract the random noise and use it outside.

### Actual Behavior

I don't know if it's due to the environment, but the official workflow for double-image editing in Flux2 has the same process as the official sub-image workflow and the workflow I set up myself. However, the sub-image workflow completely fails and can only perform single-image editing. The double-image edited images have no relation to the reference image?

### Steps to Reproduce

With the same random seed, the official Klein subgraph workflow completely fails.

### Debug Logs

```powershell
E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3>.\python_embeded\python.exe -s ComfyUI\main.py --windows-standalone-build --front-end-version Comfy-Org/ComfyUI_frontend@latest --use-pytorch-cross-attention
Adding extra search path checkpoints I:\sd-webui-aki\sd-webui-aki-v4.9\models\Stable-diffusion
Adding extra search path configs I:\sd-webui-aki\sd-webui-aki-v4.9\models\Stable-diffusion
Adding extra search path vae I:\sd-webui-aki\sd-webui-aki-v4.9\models\VAE
Adding extra search path loras I:\sd-webui-aki\sd-webui-aki-v4.9\models\Lora
Adding extra search path loras I:\sd-webui-aki\sd-webui-aki-v4.9\models\LyCORIS
Adding extra search path upscale_models I:\sd-webui-aki\sd-webui-aki-v4.9\models\ESRGAN
Adding extra search path upscale_models I:\sd-webui-aki\sd-webui-aki-v4.9\models\RealESRGAN
Adding extra search path upscale_models I:\sd-webui-aki\sd-webui-aki-v4.9\models\SwinIR
Adding extra search path embeddings I:\sd-webui-aki\sd-webui-aki-v4.9\embeddings
Adding extra search path hypernetworks I:\sd-webui-aki\sd-webui-aki-v4.9\models\hypernetworks
Adding extra search path controlnet I:\sd-webui-aki\sd-webui-aki-v4.9\models\ControlNet
Adding extra search path unet I:\sd-webui-aki\sd-webui-aki-v4.9\models\unet
Starting up lumi_batcher_service...
Registering lumi_batcher_service module...
Successfully registered module: lumi_batcher_service
lumi_batcher_service started successfully.
[START] Security scan
[DONE] Security scan
## ComfyUI-Manager: installing dependencies done.
** ComfyUI startup time: 2026-02-05 14:57:50.394
** Platform: Windows
** Python version: 3.12.7 (tags/v3.12.7:0b05ead, Oct 1 2024, 03:06:41) [MSC v.1941 64 bit (AMD64)]
** Python executable: E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\python_embeded\python.exe
** ComfyUI Path: E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI
** ComfyUI Base Folder Path: E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI
** User directory: E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\user
** ComfyUI-Manager config path: E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\user\__manager\config.ini
** Log path: E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\user\comfyui.log

Prestartup times for custom nodes:
0.0 seconds: E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\custom_nodes\rgthree-comfy
0.0 seconds: E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\custom_nodes\ComfyUI-Easy-Use
0.0 seconds: E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\custom_nodes\ComfyUI-DepthAnythingV3
0.0 seconds: E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\custom_nodes\comfyui-lumi-batcher
6.2 seconds: E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\custom_nodes\comfyui-manager

E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\python_embeded\Lib\site-packages\torch\cuda\__init__.py:65: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you.
import pynvml # type: ignore[import]
Checkpoint files will always be loaded safely.
Found comfy_kitchen backend triton: {'available': True, 'disabled': True, 'unavailable_reason': None, 'capabilities': ['apply_rope', 'apply_rope1', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8']}
Found comfy_kitchen backend eager: {'available': True, 'disabled': False, 'unavailable_reason': None, 'capabilities': ['apply_rope', 'apply_rope1', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8', 'scaled_mm_nvfp4']}
Found comfy_kitchen backend cuda: {'available': True, 'disabled': False, 'unavailable_reason': None, 'capabilities': ['apply_rope', 'apply_rope1', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8', 'scaled_mm_nvfp4']}
Total VRAM 16302 MB, total RAM 81836 MB
pytorch version: 2.10.0+cu130
xformers version: 0.0.34
Set vram state to: NORMAL_VRAM
Device: cuda:0 NVIDIA GeForce RTX 5070 Ti : cudaMallocAsync
Using async weight offloading with 2 streams
Enabled pinned memory 36826.0
working around nvidia conv3d memory bug.
Using pytorch attention
Python version: 3.12.7 (tags/v3.12.7:0b05ead, Oct 1 2024, 03:06:41) [MSC v.1941 64 bit (AMD64)]
ComfyUI version: 0.11.1
Initializing frontend: Comfy-Org/ComfyUI_frontend@latest, requesting version details from GitHub...
[Prompt Server] web root: E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\web_custom_versions\Comfy-Org_ComfyUI_frontend\1.39.7
Web extensions folder found at E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\web\extensions\ComfyLiterals
[custom_nodes.comfyui_controlnet_aux] | INFO -> Using ckpts path: E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\custom_nodes\comfyui_controlnet_aux\ckpts
[custom_nodes.comfyui_controlnet_aux] | INFO -> Using symlinks: False
[custom_nodes.comfyui_controlnet_aux] | INFO -> Using ort providers: ['CUDAExecutionProvider', 'DirectMLExecutionProvider', 'OpenVINOExecutionProvider', 'ROCMExecutionProvider', 'MIGraphXExecutionProvider', 'CPUExecutionProvider', 'CoreMLExecutionProvider']
DWPose: Onnxruntime with acceleration providers detected
E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\python_embeded\python.exe
ComfyUI-GGUF: Allowing full torch compile
======================================== ComfyUI-nunchaku Initialization ========================================
Nunchaku version: 1.3.0.dev20260202
ComfyUI-nunchaku version: 1.2.1
NumExpr defaulting to 16 threads.
2026-02-05 14:58:12.925739: I tensorflow/core/util/port.cc:153] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.
2026-02-05 14:58:14.926285: I tensorflow/core/util/port.cc:153] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.
E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\python_embeded\Lib\site-packages\keras\src\export\tf2onnx_lib.py:8: FutureWarning: In the future `np.object` will be defined as the corresponding NumPy scalar.
if not hasattr(np, "object"):
E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\python_embeded\Lib\site-packages\albumentations\__init__.py:13: UserWarning: A new version of Albumentations is available: 2.0.8 (you have 1.4.15). Upgrade using: pip install -U albumentations. To disable automatic update checks, set the environment variable NO_ALBUMENTATIONS_UPDATE to 1.
check_for_updates()
E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\python_embeded\Lib\site-packages\timm\models\layers\__init__.py:48: FutureWarning: Importing from timm.models.layers is deprecated, please import via timm.layers
warnings.warn(f"Importing from {__name__} is deprecated, please import via timm.layers", FutureWarning)
'nunchaku_versions.json' not found. Node will start in minimal mode. Use 'update node' to fetch versions.
=================================================================================================================
Efficiency Nodes: Attempting to add Control Net options to the 'HiRes-Fix Script' Node (comfyui_controlnet_aux add-on)...Success!
Loaded Efficiency nodes from E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\custom_nodes\efficiency-nodes-comfyui
Loaded comfyui_controlnet_aux nodes from E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\custom_nodes\comfyui_controlnet_aux
Loaded AdvancedControlNet nodes from E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\custom_nodes\ComfyUI-Advanced-ControlNet
Could not find AnimateDiff nodes
Loaded IPAdapter nodes from E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\custom_nodes\ComfyUI_IPAdapter_plus
Loaded VideoHelperSuite from E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\custom_nodes\ComfyUI-VideoHelperSuite
### Loading: ComfyUI-Impact-Pack (V8.28.2)
### Loading: ComfyUI-Impact-Pack (V8.28.2)
[Impact Pack] Wildcard total size (0.00 MB) is within cache limit (50.00 MB). Using full cache mode.
Loaded ImpactPack nodes from E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\custom_nodes\ComfyUI-Impact-Pack
[Impact Pack] Wildcards loading done.
[Impact Pack] Wildcard total size (0.00 MB) is within cache limit (50.00 MB). Using full cache mode.
[Impact Pack] Wildcards loading done.
HTTP Request: GET https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json "HTTP/1.1 200 OK"
[ComfyUI-DepthAnythingV3] Initializing custom node...
[ComfyUI-DepthAnythingV3] [OK] Node classes imported successfully
[ComfyUI-DepthAnythingV3] [OK] Loaded successfully!
[ComfyUI-Easy-Use] server: v1.3.6 Loaded
[ComfyUI-Easy-Use] web root: E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\custom_nodes\ComfyUI-Easy-Use\web_version/v2 Loaded
ComfyUI-GGUF: Allowing full torch compile
E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\python_embeded\Lib\site-packages\timm\models\registry.py:4: FutureWarning: Importing from timm.models.registry is deprecated, please import via timm.models
warnings.warn(f"Importing from {__name__} is deprecated, please import via timm.models", FutureWarning)
E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\python_embeded\Lib\site-packages\timm\models\helpers.py:7: FutureWarning: Importing from timm.models.helpers is deprecated, please import via timm.models
warnings.warn(f"Importing from {__name__} is deprecated, please import via timm.models", FutureWarning)
### Loading: ComfyUI-Impact-Pack (V8.28.2)
[Impact Pack] Wildcard total size (0.00 MB) is within cache limit (50.00 MB). Using full cache mode.
[Impact Pack] Wildcards loading done.
### Loading: ComfyUI-Impact-Subpack (V1.3.5)
[Impact Pack/Subpack] Using folder_paths to determine whitelist path: E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\user\default\ComfyUI-Impact-Subpack\model-whitelist.txt
[Impact Pack/Subpack] Ensured whitelist directory exists: E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\user\default\ComfyUI-Impact-Subpack
[Impact Pack/Subpack] Loaded 0 model(s) from whitelist: E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\user\default\ComfyUI-Impact-Subpack\model-whitelist.txt
[Impact Subpack] ultralytics_bbox: E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\models\ultralytics\bbox
[Impact Subpack] ultralytics_segm: E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\models\ultralytics\segm
### Loading: ComfyUI-Inspire-Pack (V1.23)
--------------------- Jake Upgrade Nodes ---------------------
🔶 Version 2.4.1
🔶 Config file loaded successfully
🔶 Using standard version of RandomPrompter
🔶 All main modules loaded
🔶 No deprecated nodes loaded
🔶 Total nodes: 105
--------------------------------------------------------------
[LaoLi Lineup] 插件已加载：显存排队系统准备就绪
[llama-cpp_vlm] Model cleanup hook applied!
comfyui_lumi_batcher_workspace has already been created
comfyui_lumi_batcher_workspace\resources has already been created
comfyui_lumi_batcher_workspace\db has already been created
comfyui_lumi_batcher_workspace\download has already been created
comfyui_lumi_batcher_workspace\upload has already been created
### Loading: ComfyUI-Manager (V3.39.2)
[ComfyUI-Manager] network_mode: public
[ComfyUI-Manager] ComfyUI per-queue preview override detected (PR #11261). Manager's preview method feature is disabled. Use ComfyUI's --preview-method CLI option or 'Settings > Execution > Live preview method'.
### ComfyUI Version: v0.11.1-19-g021ba207 | Released on '2026-02-01'
✅ MiniCPM loaded: 4 nodes from 2 modules
--- [ComfyUI-Nkxx] 正在检查核心依赖... ---
--- 正在加载自定义节点包：ComfyUI-Nkxx ---
> 正在配置全局函数...
> 已成功加载节点文件: banana_nodes.py
> 已成功加载节点文件: banana_nodes_async.py
> 已成功加载节点文件: grsaivlmllm_nodes.py
> 已成功加载节点文件: sora2_nodes.py
[Sora2 Node] yt_dlp 库已找到。
> 已成功加载节点文件: sora2_nodes_async.py
> 已成功加载节点文件: veo3_nodes.py
--- ComfyUI-Nkxx 加载完成，共找到 17 个节点 ---
======================================== ComfyUI-nunchaku Initialization ========================================
Nunchaku version: 1.3.0.dev20260202
ComfyUI-nunchaku version: 1.2.1
'nunchaku_versions.json' not found. Node will start in minimal mode. Use 'update node' to fetch versions.
=================================================================================================================
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/alter-list.json
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/model-list.json
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/github-stats.json
ComfyUI-GGUF: Allowing full torch compile
✨提示词小助手 V2.0.2 已启动
Nvidia APEX normalization not installed, using PyTorch LayerNorm
Nvidia APEX normalization not installed, using PyTorch LayerNorm
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/extension-node-map.json

[ReActor] - STATUS - Running v0.6.2-a3 in ComfyUI
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/custom-node-list.json
Torch version: 2.10.0+cu130
Warning: GroundingDINO not available. Text prompts will use fallback method.
Error loading AILab_SegmentV2.py: No module named 'groundingdino'
[ComfyUI-RMBG] v2.9.0 | 31 nodes Loaded
⚡ SeedVR2 optimizations check: SageAttention ✅ | Flash Attention ✅ | Triton ✅
🔧 Conv3d workaround active: PyTorch 2.10.0, cuDNN 91200 (fixing VAE 3x memory bug)
📊 Initial CUDA memory: 14.60GB free / 15.92GB total

[notice] A new release of pip is available: 25.3 -> 26.0
[notice] To update, run: python.exe -m pip install --upgrade pip
Traceback (most recent call last):
File "E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\nodes.py", line 2216, in load_custom_node
module_spec.loader.exec_module(module)
File "", line 995, in exec_module
File "", line 488, in _call_with_frames_removed
File "E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\custom_nodes\ComfyUI-Trellis2\__init__.py", line 1, in
from .nodes import NODE_CLASS_MAPPINGS, NODE_DISPLAY_NAME_MAPPINGS
File "E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\custom_nodes\ComfyUI-Trellis2\nodes.py", line 20, in
import cumesh as CuMesh
File "E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\python_embeded\Lib\site-packages\cumesh\__init__.py", line 1, in
from . import remeshing
File "E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\python_embeded\Lib\site-packages\cumesh\remeshing.py", line 4, in
from . import _C
ImportError: DLL load failed while importing _C: 找不到指定的程序。

Cannot import E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\custom_nodes\ComfyUI-Trellis2 module for custom nodes: DLL load failed while importing _C: 找不到指定的程序。
ImageCompositeMaskedWithSwitch: 检测父节点类型 - V3支持=True, 父节点是V3=True
ImageCompositeMaskedWithSwitch: 使用 V3 Schema 模式
(pysssss:WD14Tagger) [DEBUG] Available ORT providers: TensorrtExecutionProvider, CUDAExecutionProvider, CPUExecutionProvider
(pysssss:WD14Tagger) [DEBUG] Using ORT providers: CUDAExecutionProvider, CPUExecutionProvider
------------------------------------------
Comfyroll Studio v1.76 : 175 Nodes Loaded
------------------------------------------
** For changes, please see patch notes at https://github.com/Suzie1/ComfyUI_Comfyroll_CustomNodes/blob/main/Patch_Notes.md
** For help, please see the wiki at https://github.com/Suzie1/ComfyUI_Comfyroll_CustomNodes/wiki
------------------------------------------
[E:\ComfyUI_windows_gguf\ComfyUI_windows_0_x_3\ComfyUI\custom_nodes\comfyui_controlnet_aux] | INFO -> Using ckpts path: E:\ComfyUI_windows_gguf\ComfyUI_windows_0.3\ComfyUI\custom_nodes\comfyui_controlnet_aux\ckpts
[E:\ComfyUI_windows_gguf\ComfyUI_windows_0_x_3\ComfyUI\custom_nodes\comfyui_controlnet_aux] | INFO -> Using symlinks: False
[E:\ComfyUI_windows_gguf\ComfyUI_windows_0_x_3\ComfyUI\custom_nodes\comfyui_controlnet_aux] | INFO -> Using ort providers: ['CUDAExecutionProvider', 'DirectMLExecutionProvider', 'OpenVINOExecutionProvider', 'ROCMExecutionProvider', 'MIGraphXExecutionProvider', 'CPUExecutionProvider', 'CoreMLExecutionProvider']
** Comfly Loaded : Go Younjung Is My Favorite Girl!--T8star

### [START] ComfyUI AlekPet Nodes v1.0.91 ###
Node -> ArgosTranslateNode: ArgosTranslateCLIPTextEncodeNode, ArgosTranslateTextNode [Loading]
Node -> ChatGLMNode[Loading]
Node -> DeepTranslatorNode: DeepTranslatorCLIPTextEncodeNode, DeepTranslatorTextNode [Loading]
Node -> ExtrasNode: PreviewTextNode, HexToHueNode, ColorsCorrectNode [Loading]
Node -> GoogleTranslateNode: GoogleTranslateCLIPTextEncodeNode, GoogleTranslateTextNode [Loading]
Node -> IDENode: IDENode [Loading]
Node -> PainterNode: PainterNode [Loading]
Node -> PoseNode: PoseNode [Loading]
### [END] ComfyUI AlekPet Nodes ###
Total VRAM 16302 MB, total RAM 81836 MB
pytorch version: 2.10.0+cu130
xformers version: 0.0.34
Set vram state to: NORMAL_VRAM
Device: cuda:0 NVIDIA GeForce RTX 5070 Ti : cudaMallocAsync
Using async weight offloading with 2 streams
Enabled pinned memory 36826.0
FETCH ComfyRegistry Data: 5/123
LOADED 6 FONTS
No OpenGL_accelerate module loaded: No module named 'OpenGL_accelerate'
```

### Other

With the same random seed, the official Klein subgraph workflow completely fails.