Flux Kontext 360° Workflow - Seamless Panorama Generation
Complete workflow guide for creating seamless 360-degree panoramic images using Flux Kontext LoRA.
Workflow Overview
This specialized ComfyUI workflow transforms standard input images into immersive, seamless 360-degree panoramas using the trained Kontext LoRA model1. The key feature of this workflow is its ability to generate perfectly seamless wraparound images without visible seams or distortions at the connection points.
Key Features
Seamless Output: Generated 360° images connect perfectly at the edges, creating true wraparound panoramas
Context-Aware Generation: The LoRA understands spatial relationships to maintain visual coherence across the full 360° span
Quality Enhancement: Integrated tiled diffusion upscaling for superior final results
Workflow Steps
Input Processing: Load your source image and apply the Flux Kontext LoRA with appropriate strength settings. The model analyzes the input to understand the spatial context and visual elements that need to be extended.
360° Generation: The workflow generates the panoramic expansion while ensuring seamless edge matching. Special attention is paid to maintaining consistent lighting, perspective, and visual elements across the entire 360° span.
Seamless Integration: Advanced blending techniques ensure that the left and right edges of the generated panorama connect perfectly, creating a truly seamless wraparound experience.
Quality Enhancement: Apply tiled diffusion upscaling to significantly improve the final image quality and detail resolution.
Technical Specifications
Model: Flux with Kontext LoRA
Output Format: Seamless 360° panoramic images
Recommended Upscaling: Tiled diffusion for optimal results
Interface: ComfyUI workflow
The seamless nature of the output makes these panoramas perfect for VR environments, immersive displays, and any application requiring true 360-degree visual experiences without jarring transitions or visible seams.
Description
FAQ
Comments (30)
十分棒的工作流，作为设计师很喜欢这一类工作流，期待更新更多的功能
Thanks good workflow also for non 360 degree pictures. Danke
Do you have a link to the GitHub repo?
Is it technically possible to create vr180?
Yes with a LoRa trained with 180 Degree Images and Context Images
bug，，，SamplerCustomAdvanced
Boolean value of Tensor with more than one value is ambiguous
The lora is great and the workflow works, but there are a couple of things that, in my opinion, are overcomplicated or not useful in it.
Maybe I'm wrong, but what you're doing in the crop part (first and last) with 4 nodes, can easily be done with a single image concatenate.
Also there's a florance model that only describes the first image generated, but is not connected to anything else. Was it supposed to be used in the second conditioning to improve the effectivness?
I also don't have the fill_nodes so, I'm trying to figure out what the FL_SDUltimate_Slices does. Is it a 4x4 slicing (2x2 multiplied by 2)?
Thank you and keep up the good work.
Yeah, the Florence isn't really necessary — the idea was just to capture the cutouts and use them as prompts, but it also works without prompting. The Slixes node automatically calculates the tile size, though you could also build that with math nodes or calculate it manually and input the values.
wtf is HDR360.safetensors????
The 360 LoRa ???
https://civitai.com/models/682349/360-degree-flux-and-kontext
the workflow does not work for me. it only artificially widens the image by “slicing” or “expanding” it, without the actual spherical distortion and the horizon-to-pole gradient... what is the reason for this, what do I have to do?
its not working. all models downloaded and all nodes installed. workflow not giving any error but only giving upscaled version of what i give as an image.
Update Comfyui ?
me paso lo mismo ? todo actualizado, todo instalado, modelos y Loras intsalados, y solo me genera una upscale ??
denrakeiw The same to me. But I put this workflow on some online website and it worked. Running it locally gave no errors but it didn't work at all. I'm confused.
zengwusen330 Update Comfyui ?
denrakeiw Thanks for your reply. I updated it and tried countless methods. Just now, I finally changed a model and it worked. I changed to F.1-kontext-dev-fp8-e4m3f
denrakeiw Does this Lora support adding additional prompts, such as adding desired scenes to the invisible parts of the picture?
zengwusen330 Try it, but the prompt should be as short as possible — something like “Add xyc.”
This does really well with scenery and stuff! Can get some pretty mangled results with other subjects. Question- is it possible to have it output an image that is 4096 pixels wide? I couldn't really figure out how to adjust the resolution settings. Thanks for sharing!
Yes, that's possible, but you’ll need to reconfigure all the nodes that crop and merge the inpainting areas. You'll have to calculate the settings yourself—or have ChatGPT calculate them for you.
Is there a setting to control the position of the original image in the 360 image?
it's always in the center
Hello, the predefined resolution works relatively well, although the seams are not always completely fixed, you can always see a little bit where they were. However, when I try to use double the size -
orig - 1536 x 768
test x2 - 3072 x 1536
the output is not nice at all, the seams are not fixed at all. Is there any way to generate correct seams when enlarging?
Workflow works, it generates 360 pano that stiches very nice, but gives artifacts in few tiles, that ruine the result. Can`t solve this. Thank you.
Image
It gives this error: Some Nodes Are Missing: ETN_CropImage
If you load version 1.0.0 of the Tooling Nodes, then it will load the ETN_CropImage node.
I can't find it, the version 1.00 here https://github.com/Acly/comfyui-tooling-nodes . I think the most convenient thing would be for the creator of this workflow to adapt it to the latest version of Tooling nodes.
@traviesomorbo69826 in comfyui manager search for comfyui-tooling-nodes. Delete it and install again so you can select 1.0.0 version. Its all in ComfyUI Manager