Stable Diffusion 3 Simple Workflow
5.0
0 reviews
Description
Credit & More Detail: https://comfyanonymous.github.io/ComfyUI_examples/sd3/
sd3_medium_incl_clips.safetensors and sd3_medium_incl_clips_t5xxlfp8.safetensors will work with this workflow, just make sure they are in your ComfyUI/models/checkpoints folder.
Node Diagram
Discussion
Using sd3_medium_incl_clips_t5xxlfp8.safetensors
Error occurred when executing CheckpointLoaderSimple: Dtype not understood: F8_E4M3 File "F:\apps\ComfyUI_windows_portable\ComfyUI\execution.py", line 151, in recursive_execute output_data, output_ui = get_output_data(obj, input_data_all) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "F:\apps\ComfyUI_windows_portable\ComfyUI\execution.py", line 81, in get_output_data return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "F:\apps\ComfyUI_windows_portable\ComfyUI\execution.py", line 74, in map_node_over_list results.append(getattr(obj, func)(**slice_dict(input_data_all, i))) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "F:\apps\ComfyUI_windows_portable\ComfyUI\nodes.py", line 516, in load_checkpoint out = comfy.sd.load_checkpoint_guess_config(ckpt_path, output_vae=True, output_clip=True, embedding_directory=folder_paths.get_folder_paths("embeddings")) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "F:\apps\ComfyUI_windows_portable\ComfyUI\comfy\sd.py", line 449, in load_checkpoint_guess_config sd = comfy.utils.load_torch_file(ckpt_path) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "F:\apps\ComfyUI_windows_portable\ComfyUI\comfy\utils.py", line 14, in load_torch_file sd = safetensors.torch.load_file(ckpt, device=device.type) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "F:\apps\ComfyUI_windows_portable\python_embeded\Lib\site-packages\safetensors\torch.py", line 310, in load_file result[k] = f.get_tensor(k) ^^^^^^^^^^^^^^^
Update: Needed to update ComfyUI!!!!!
(Edited)Yup, Comfy UI needs to be updated. They released the updated version yesterday
This is missing the text encoders that are required to get a good image.
https://huggingface.co/stabilityai/stable-diffusion-3-medium/tree/main/text_encoders
You can see them in the example workflows.
https://huggingface.co/stabilityai/stable-diffusion-3-medium/tree/main/comfy_example_workflows
It's not missing, it's official Comfy UI workflow. In this worklfow the text encoders are embedded inside the checkpoint. Either you can use it like that, or download the text encoders and input them
(Edited)All the official SD3 ComfyUI workflows use TripleCLIPLoader. Your example images, all used the 3 text encoders. This is how you get accurate text, one of the models main features.
It's the first thing listed in the 3 example workflows.
https://huggingface.co/stabilityai/stable-diffusion-3-medium/tree/main/comfy_example_workflows
It's the same way that SDXL requires special text encoders, to get the best images. I think only < SD 1.5 works well without special clip conditioning.
You would need a node for someone to be able to downloaded and input text encoders.
It's the "Official Comfy UI" workflow by @comfyanonymous, not official SAI workflow. I for Comfy UI already shared in the description: https://comfyanonymous.github.io/ComfyUI_examples/sd3/
When I click the QueuePrompt button, my terminal displays "press any key to continue" and the program terminates
furry
Node Details
Primitive Nodes (2)
EmptySD3LatentImage (1)
Note (1)
Custom Nodes (6)
ComfyUI
- VAEDecode (1)
- SaveImage (1)
- KSampler (1)
- CLIPTextEncode (2)
- CheckpointLoaderSimple (1)
Model Details
Checkpoints (1)
sd3_medium_incl_clips_t5xxlfp8.safetensors
LoRAs (0)