Required Files (v3)
To use the workflows, you need to have the following models installed in ComfyUI.
Place them in the corresponding directories as specified below.
<your_comfyui_dir> / models / checkpoints :
pixart_sigma-FP16.safetensors (1.2GB)
photon_refiner-FP16.safetensors (2.1GB)
<your_comfyui_dir> / models / clip :
t5_xxl_encoder-FP8.safetensors (4.9GB)
<your_comfyui_dir> / models / vae :
pixart_vae.safetensors (0.1GB)
Required Nodes (v3)
IMPORTANT: Ensure that your ComfyUI is updated to the latest version.
Additionally, the workflows require the following custom nodes to be installed:
ComfyUI_ExtraModels: provides support for PixArt-Sigma.
ComfyUI-Crystools: used for some simple string operations.
-- under construction --
Version History
v3
- Models in FP16 & FP8 for min VRAM usage (~6GB).
- Unified PixArt+refiner prompts (no more two separate prompts).
- 1536 x 1024 native resolution.
- Option to select portrait/landscape image orientation.
- Option to choose between fast/quality samplers.
- Option for the refiner to ignore the prompt.
- Variations achieved by changing the refiner seed.
- Six preconfigured workflows.
- CivitAI prompt extractor compatibility.
v2
- Simplified steps configuration.
- Added nodes to easily modify CFG and SEED parameters.
- Aesthetic node color/position changes for better usability.
- Updated install instructions with corrections and clarity.
- Improved final image organization into date-based folders.
v1
- The initial release providing core functionality.Description
Initial test version to experiment with the PixArt Sigma model.
FAQ
Comments (27)
Please photon XL, meu rei!
I was planning to work on that during my february vacation, but then the SD3 announcement came out, and I started doubting if it was even worth it. Now, with this new PixArt model, I feel even more torn. There are so many things I'd love to do, but I have so little time. Thanks for the encouragement though! I'll definitely try to make something happen when things settle down and new things stop popping up every day! haha
Haha, no worries, take your time! But I guess a new model from you would be very welcome, given the quality provided by photon until the present day.
@Primaveri It feels like the AI landscape has changed so much in just these 3 months! Flux has definitely been making waves, and while it's powerful, it's a bit of a beast to run. I've been playing around with PixArt-Sigma though. I'm actually thinking about training a Photon-Sigma model o something like that, combining PixArt's speed and creativity with Photon's style. I just need some free time, as always haha.
@Photographer That would be really nice! Pixart is really good, I've been using pixelalchemy and it's very powerful for details and creative realism. Flux is really powerful as well. I'm using flux as a second pass upscale/refiner and it really improves the overall quality of details from some XL gens. Maybe Flux Photon would be really great, given that Photon model shines where Flux fails, like HD clarity and skin details.
I followed the installation and get two error:
Error occurred when executing PixArtCheckpointLoader: No module named 'timm' File "C:\ComfyUI\ComfyUI\execution.py", line 151, in recursive_execute output_data, output_ui = get_output_data(obj, input_data_all) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\ComfyUI\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 "C:\ComfyUI\ComfyUI\execution.py", line 74, in map_node_over_list results.append(getattr(obj, func)(**slice_dict(input_data_all, i))) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\ComfyUI\ComfyUI\custom_nodes\ComfyUI_ExtraModels\PixArt\nodes.py", line 29, in load_checkpoint model = load_pixart( ^^^^^^^^^^^^ File "C:\ComfyUI\ComfyUI\custom_nodes\ComfyUI_ExtraModels\PixArt\loader.py", line 86, in load_pixart from .models.PixArtMS import PixArtMS File "C:\ComfyUI\ComfyUI\custom_nodes\ComfyUI_ExtraModels\PixArt\models\PixArtMS.py", line 14, in from timm.models.layers import DropPathand this one:
Error occurred when executing PixArtCheckpointLoader: No module named 'timm' File "C:\ComfyUI\ComfyUI\execution.py", line 151, in recursive_execute output_data, output_ui = get_output_data(obj, input_data_all) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\ComfyUI\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 "C:\ComfyUI\ComfyUI\execution.py", line 74, in map_node_over_list results.append(getattr(obj, func)(**slice_dict(input_data_all, i))) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\ComfyUI\ComfyUI\custom_nodes\ComfyUI_ExtraModels\PixArt\nodes.py", line 29, in load_checkpoint model = load_pixart( ^^^^^^^^^^^^ File "C:\ComfyUI\ComfyUI\custom_nodes\ComfyUI_ExtraModels\PixArt\loader.py", line 86, in load_pixart from .models.PixArtMS import PixArtMS File "C:\ComfyUI\ComfyUI\custom_nodes\ComfyUI_ExtraModels\PixArt\models\PixArtMS.py", line 14, in from timm.models.layers import DropPathI had the same issue. Use the "Standalone" Install option and it fixed the issue for me.
https://github.com/city96/ComfyUI_ExtraModels?tab=readme-ov-file#standalone-comfyui
Standalone ComfyUI
I haven't tested this completely, so if you know what you're doing, use the regular venv/git clone install option when installing ComfyUI.
Go to the where you unpacked ComfyUI_windows_portable to (where your run_nvidia_gpu.bat file is) and open a command line window. Press CTRL+SHIFT+Right click in an empty space and click "Open PowerShell window here".
Clone the repository to your custom nodes folder, assuming haven't installed in through the manager.
git clone https://github.com/city96/ComfyUI_ExtraModels .\ComfyUI\custom_nodes\ComfyUI_ExtraModels
To install the requirements on windows, run these commands in the same window:
.\python_embeded\python.exe -s -m pip install -r .\ComfyUI\custom_nodes\ComfyUI_ExtraModels\requirements.txt .\python_embeded\python.exe -s -m pip install bitsandbytes --prefer-binary --extra-index-url=https://jllllll.github.io/bitsandbytes-windows-webuiTo update, open the command line window like before and run the following commands:
cd .\ComfyUI\custom_nodes\ComfyUI_ExtraModels\ git pullSDXL as final i2i work much better. At least for photos.
Yes, it can be, it depends on the model. Anyway, the original idea behind this workflow is to create a fast process that minimizes resource usage and also maintains a visual aesthetic more similar to SD15.
An excellent and efficient workflow with clear instructions. Well done!
Thanks for the kind words! I put a lot of effort into making it user-friendly, and it's great to hear that it paid off.
Can anyone help me, I'm receiving an error when trying to use the PixArt .pth file. I read on the github that it needed to be converted to a safetensor somehow, but I'm unsure how to do so. Any help would be greatly appreciated :)
You need to install a bunch of other stuff for the .pth file to work. Please follow the instruction given here closely.
@NowhereManGo I've installed all the files and placed them in the correct directories. The only thing I'm unsure about is the 'diffusion_pytorch_model.safetensors' file, as there is no direction on where to place it. Does it just go in the models folder with all the other checkpoints?
@SLAPY7 You need to read the instruction more carefully, it's there 😎.
Rename the file 'diffusion_pytorch_model.safetensors' to 'pixart_sigma_vae.safetensors'
[...]
comfyui/models/vae
pixart_sigma_vae.safetensors
@NowhereManGo Okay great, done that. And the 'PixArt-Sigma-XL-2-1024-MS.pth' file can stay as a .pth and placed in the models/checkpoints folder?
@SLAPY7 Yes, you can keep the "PixArt-Sigma-XL-2-1024-MS.pth" in the models/checkpoints folder.
I'm actually working on version 3, which will not include any .pth files. Instead, it will use .safetensors for added security. Version 3 will also be much faster, although I'm not sure if it will be easier to install. The only way to ensure both optimization and ease of installation would be to create my own nodes. This is something I'm considering for v4, but I'm not sure if I'll have enough free time to dedicate to it.
Good wok bro
It's been a while, but thanks!
amazing work, thanks.
i tried also to make this work with PixArt-Sigma-XL-2-2K-MS, but it always give only noise pixels
my output also noise pixels, did you already know how to solved it ?
@Blackstyle for PixArt-Sigma-XL-2-2K-MS, no it still the same, works only with the specific model in the workflow
@gamal Oh you mean, the workflow working if use this PixArt-Sigma-XL-2-2K-MS model ?
@Blackstyle no it don't> it work only with its default model
There are some very impressive images here 👍.
Thank you for sharing the workflow and the instruction on how to set up PixArt Sigma.
I hadn't seen your message! PixArt-Sigma is amazing, even though it doesn't have the same level of coherence as FLUX, it's impressive what it can achieve.
@Photographer Yes, the PixArt team accomplished a lot with a small model trained with a modest dataset and GPU. I am very impressed by them. I am looking forward to the next version of Pix-Art (I guess I am kind of an A.I. model junkie 😂)