What happened?
The v2.16.0 update introduced automatic wiring for new model architectures (Z-Image, Anima, etc.), but the backend is failing to pass the required text encoders to ComfyUI.
Because the UI assumes the "workflow-appropriate text encoders" are being handled automatically, it completely hides the manual Text Encoder dropdowns from the config menu. This creates a hard block: the auto-router sends a None value to the ComfyUI backend, and the user is locked out of manually selecting an encoder to fix it.
This issue is not isolated to a single model; it occurs across multiple new architectures introduced in this update. I expected the automated pipeline to route the encoders correctly, or to at least retain the ability to manually assign my text encoders from the config menu. Instead, generation immediately crashes.
Steps to reproduce
-
Open Stability Matrix and navigate to the Inference tab.
-
Load a model that utilizes the new architectures (e.g., a Z-Image Turbo model or an Anima model).
-
Set the Workflow dropdown to match (e.g., Z-Image Turbo or Anima / SD).
-
Click the ⚙️ icon next to the Model; note that the Text Encoder selectors are completely missing.
-
Attempt to generate an image.
-
The generation fails immediately. ComfyUI throws a RuntimeError regarding missing CLIP input because Stability Matrix passed None.
Relevant logs
Comfy Error
Node execution encountered an error
{
"prompt_id": "e119c145-9650-41d3-9f3e-eb91cc1c71d0",
"node_id": "NegativeCLIP_Base",
"node_type": "CLIPTextEncode",
"executed": [],
"exception_message": "ERROR: clip input is invalid: None\n\nIf the clip is from a checkpoint loader node your checkpoint does not contain a valid clip or text encoder model.\n",
"exception_type": "RuntimeError",
"traceback": [
...Version
v2.16.0
What Operating System are you using?
Windows
What happened?
The v2.16.0 update introduced automatic wiring for new model architectures (Z-Image, Anima, etc.), but the backend is failing to pass the required text encoders to ComfyUI.
Because the UI assumes the "workflow-appropriate text encoders" are being handled automatically, it completely hides the manual Text Encoder dropdowns from the config menu. This creates a hard block: the auto-router sends a None value to the ComfyUI backend, and the user is locked out of manually selecting an encoder to fix it.
This issue is not isolated to a single model; it occurs across multiple new architectures introduced in this update. I expected the automated pipeline to route the encoders correctly, or to at least retain the ability to manually assign my text encoders from the config menu. Instead, generation immediately crashes.
Steps to reproduce
Open Stability Matrix and navigate to the Inference tab.
Load a model that utilizes the new architectures (e.g., a Z-Image Turbo model or an Anima model).
Set the Workflow dropdown to match (e.g., Z-Image Turbo or Anima / SD).
Click the ⚙️ icon next to the Model; note that the Text Encoder selectors are completely missing.
Attempt to generate an image.
The generation fails immediately. ComfyUI throws a RuntimeError regarding missing CLIP input because Stability Matrix passed None.
Relevant logs
Comfy Error Node execution encountered an error { "prompt_id": "e119c145-9650-41d3-9f3e-eb91cc1c71d0", "node_id": "NegativeCLIP_Base", "node_type": "CLIPTextEncode", "executed": [], "exception_message": "ERROR: clip input is invalid: None\n\nIf the clip is from a checkpoint loader node your checkpoint does not contain a valid clip or text encoder model.\n", "exception_type": "RuntimeError", "traceback": [ ...Version
v2.16.0
What Operating System are you using?
Windows