GPT-4o Ghibli Stylization
5.0
6 reviewsDescription
https://github.com/logtd/ComfyUI-Fluxtapoz
https://civitai.com/models/1411365/studio-chatgpt-ghibli
Note: The recommended value for 'skip steps' in the Flow Edit Sampler node is 0-4.
Node Diagram
Discussion
Can you provide CFG-Zero-star node
Update ComfyUI to the latest version, and you'll have this node. Of course, you can bypass it—it seems to only work for text-to-image tasks.
Ok thanks
Thaaks for this one comment about text to image. Saves me hours of downloading and trying to..
I fixed this by downloading this comfyUi: https://github.com/comfyanonymous/ComfyUI
It comes with the CFGZeroStar node
I already installed that version and it still does´nt work
感谢大佬分享,这是我人生跑通的第一个第三方workflow,效果非常好,除了速度有点慢,一张图M2 16g的MacBookPro大概需要6600秒。
I don't have a Mac, but on an NVIDIA 40-series GPU with 12GB VRAM, it runs in under 60 seconds.
This is by far the best workflow. And I have tried them all.
Congrats!
anyone got this error? is it due to amd card
CLIPTextEncode
Cannot access storage of OpaqueTensorImpl
(Edited)GOOD WORK, CAN I KNOW WHY MY GENERATE IMAGE SEEMS TO BE MIRRORED?
THANKS
Thanks a lot @makisekurisu for this is a great workflow!
I just have some pictures that don't seem to be getting the effect applied correctly for some reason, it creates what seems to be "something in between" I'm running 1200/1800px images, you can see the input/output images here: https://ibb.co/20fSxjyN
On a side node: Sometimes it struggles a bit to correctly represent some skin tones, like for example on this picture: https://ibb.co/RT8QXxhV, the person on the right has a darker skin tone but on the result image I'm getting both with the same skin tone color.
Would appreciate very much any help on those issues! Thanks a lot once again!
Set the skip steps to 0;
Generate multiple times to see if you get a satisfying result.
Thanks a lot for the response!
Using 0 skip steps results in an image very close to the original image with no Ghibli effect applied, if I go above 4 skip steps like 6,7,8 I just end up a blurry image... For some reason my bartender photo never get's the effect, even when I try smaller resolutions like 600x900 or 400x600...
Unfortunately neither increasing or decreasing the skip steps solved any of the issues...
skip_steps 0
refine_steps 28
If it still doesn't work, please try adjusting parameters such as the strength of 'applystylemode' yourself. In fact, your image is not particularly unique—I would simply think that you haven't properly understood or utilized this workflow.
(Edited)Indeed your workflow is new to me with a different approach so I'm still learning it bit by bit.
Changing the skip/refine steps did not solve the problem but increasing the strength of the 'applystylemode' did! Thanks a lot for the help! I see it does struggle a bit with medium to dark skin woman in some cases, but that might more lora specific than workflow related.
Really appreaciate your help thanks!
Prompt execution failed
Prompt outputs failed validation:
VAELoader:
- Value not in list: vae_name: 'ae.safetensors' not in []
UnetLoaderGGUF:
- Value not in list: unet_name: 'flux1-dev-Q8_0.gguf' not in []
DualCLIPLoaderGGUF:
- Value not in list: clip_name1: 'clip_l.safetensors' not in []
- Value not in list: clip_name2: 't5-v1_1-xxl-encoder-Q8_0.gguf' not in []
LoraLoaderModelOnly:
- Value not in list: lora_name: 'studiochatgpt-ghibli-v1.safetensors' not in []
StyleModelLoader:
- Value not in list: style_model_name: 'flux1-redux-dev.safetensors' not in []
CLIPVisionLoader:
- Value not in list: clip_name: 'sigclip_vision_patch14_384.safetensors' not in []
Hi, can you tell me if it is possible to use your workflow with other Lors? Which lore should be considered?
嗨!
我是 BoredWhy 的 Marcio。
您的这个工作流程真是太棒了!我们已将其在 BoredWhy 上线,方便大家体验和使用。希望您不介意!
请点击这里查看:https://boredwhy.com/workflow/68345328161b78915e59d727
当然,所有的功劳都归您所有!
-----------------
Hey there! Marcio from BoredWhy here.
Super cool workflow! We've put it online for people to easily try out on BoredWhy – hope you don't mind!
Check it out here: https://boredwhy.com/workflow/68345328161b78915e59d727
And of course, full credit goes to you!
Node Details
Primitive Nodes (11)
Anything Everywhere3 (1)
CFGZeroStar (1)
ConfigureModifiedFlux (1)
DualCLIPLoaderGGUF (1)
FlowEditGuider (1)
FlowEditSampler (1)
FluxGuidance (2)
Image Comparer (rgthree) (1)
OutFluxModelSamplingPred (1)
UnetLoaderGGUF (1)
Custom Nodes (15)
ComfyUI
- SamplerCustomAdvanced (1)
- VAEDecode (1)
- BasicScheduler (1)
- VAEEncode (1)
- DisableNoise (1)
- VAELoader (1)
- LoraLoaderModelOnly (1)
- StyleModelLoader (1)
- ConditioningConcat (1)
- StyleModelApply (1)
- CLIPVisionLoader (1)
- CLIPVisionEncode (1)
- CLIPTextEncode (2)
- LoadAndResizeImage (1)
Model Details
Checkpoints (0)
LoRAs (1)
studiochatgpt-ghibli-v1.safetensors