Hey, I figured one way to erase the unwanted pixels of the product when using canny filter to generate the background. Simple use of a Lama inpainting > use the mask of the original product > grow mask by 2-4 > should erase the pixels and regenrate them according to the background. Awesome video
you are a genius!!! This is the best WF for product photos!!! I added an upscaler, AuraSR is the best one to keep the text or details. Consider to remove the compression and JPG artifacts before scaling (any DeJPG model works). I'm trying to be able to place props in the scene, it would be great to be able to add objects to the scene. Imagine a hamburger, being able to add the toppings, a glass of beer, or another prop. If you have an idea of how to do this I would be very grateful 🤩🤩🤩🤩🤩🤩🤩 Sending you a hug from Argentina! 🤗
Thank you so much for the kind words and hug! 😊 I really appreciate it! When it comes to adding elements to a picture, inpainting is the way to go. From what I know, maybe considering using a specialized model like Juggernaut XL inpainting for that, or BrushNet for inpainting, or maybe even Photoshop's AI tools. Have you tried any of these methods?
Hi, thanks for sharing your workflow. I have only 1 problem, at the relighting the image stage, somehow the background that i generated from Easy Diffusion changes. Although it is only slight change, but is there a way that i can preserve the background image? It it possible to include Lora that i use to generate the image in Easy diffusion in here? Sorry for so many questions since i’m new to SD😄 (I’m using easy diffusion to generate background image since i cannot get the result i want in comfyui. Do tell me if there is a way to make it better though)
Thanks for bringing this up-it's a great question! Right now, when you do relighting, there will be some changes to the background; it's just part of the process with the current capabilities. However, if ic-light gets an upgrade in the future, it could handle this even better. I hope this clears things up, and I'm always here to help if you have more questions!
Hello blogger, there is an error in my use of workflow. I downloaded three models in lllyasviel/ic-light, but the lighting group still shows that the model loading fails. What model do you use? I can't see the model name, did I download it wrong?😣
Love this!!! Installed but got the following error "When loading the graph, the following node types were not found: "LayerColor: Brightness Contrast" Could you please assist? @my-ai-force. I tried downloading the raw code from Github and placed in in the custom nodes folder, tried this multiple times and still getting the error.
Thanks so much for enjoying the video! 😊 I’ve encountered this issue in earlier versions of ComfyUI as well. Sometimes, simply refreshing your browser or clicking the “Try Fix” button in the node location within the ComfyUI Manager can resolve the problem. In the more recent versions of ComfyUI, I haven’t come across any node unavailability or conflicts. Have you updated to the latest version of ComfyUI? Let me know if you need any further assistance!
Workflow V2 Updates: th-cam.com/video/64bOxqzQ0pg/w-d-xo.html
Hey, I figured one way to erase the unwanted pixels of the product when using canny filter to generate the background.
Simple use of a Lama inpainting > use the mask of the original product > grow mask by 2-4 > should erase the pixels and regenrate them according to the background.
Awesome video
Great points! Thank you, I will definitely have a try.
hi yvan, how does he do it in this video? th-cam.com/video/PkyKWVwnuM0/w-d-xo.htmlsi=BtmFRmodvy5kz0_b&t=122 is that what you mean?
I love you man!!
Omg you are amazing wei!!❤❤❤❤❤❤❤❤❤❤❤
Thanks a ton for your warm words.
you are a genius!!! This is the best WF for product photos!!! I added an upscaler, AuraSR is the best one to keep the text or details. Consider to remove the compression and JPG artifacts before scaling (any DeJPG model works). I'm trying to be able to place props in the scene, it would be great to be able to add objects to the scene. Imagine a hamburger, being able to add the toppings, a glass of beer, or another prop. If you have an idea of how to do this I would be very grateful 🤩🤩🤩🤩🤩🤩🤩 Sending you a hug from Argentina! 🤗
Thank you so much for the kind words and hug! 😊 I really appreciate it! When it comes to adding elements to a picture, inpainting is the way to go. From what I know, maybe considering using a specialized model like Juggernaut XL inpainting for that, or BrushNet for inpainting, or maybe even Photoshop's AI tools. Have you tried any of these methods?
what about glossy surfaces? will it reflects the product?
Hi, thanks for sharing your workflow. I have only 1 problem, at the relighting the image stage, somehow the background that i generated from Easy Diffusion changes. Although it is only slight change, but is there a way that i can preserve the background image? It it possible to include Lora that i use to generate the image in Easy diffusion in here? Sorry for so many questions since i’m new to SD😄
(I’m using easy diffusion to generate background image since i cannot get the result i want in comfyui. Do tell me if there is a way to make it better though)
Thanks for bringing this up-it's a great question! Right now, when you do relighting, there will be some changes to the background; it's just part of the process with the current capabilities. However, if ic-light gets an upgrade in the future, it could handle this even better. I hope this clears things up, and I'm always here to help if you have more questions!
Does it also apply to the human face or is it only for the product?
Another question . What Laura model was that you used?
It can also be applied to faces, but it won't work perfectly. You can try this lora: civitai.com/models/251417/midjourney-mimic
Hello blogger, there is an error in my use of workflow. I downloaded three models in lllyasviel/ic-light, but the lighting group still shows that the model loading fails. What model do you use? I can't see the model name, did I download it wrong?😣
iclight_sd15_fc.safetensors
Can this all be automated and then called by an API or webhook? Thanks !
I’m afraid this technique is outside my area of expertise at the moment.
Congrats. Do you think that Flux Dev should improve the result?
Love this!!! Installed but got the following error "When loading the graph, the following node types were not found: "LayerColor: Brightness Contrast" Could you please assist? @my-ai-force. I tried downloading the raw code from Github and placed in in the custom nodes folder, tried this multiple times and still getting the error.
Thanks so much for enjoying the video! 😊 I’ve encountered this issue in earlier versions of ComfyUI as well. Sometimes, simply refreshing your browser or clicking the “Try Fix” button in the node location within the ComfyUI Manager can resolve the problem. In the more recent versions of ComfyUI, I haven’t come across any node unavailability or conflicts. Have you updated to the latest version of ComfyUI? Let me know if you need any further assistance!
@@my-ai-force I have updated mine and the error persisted