Let me know what other comfyui topics you want me to cover! Also, these types of videos don't do to well on YT because watch time is everything. With tutorial channels people tend to skip ahead or just get what they need and bounce. Unfortunately that kills my retention metrics so any engagement, comment, like and share would really be appreciated! This is the best way to support the channel.
Gentleman, the goal is to use Flux as "The Ultimate Human Anatomy Refiner" for SD Series. That why I was asking for a workflow... Thank you for your attention.
You're very welcome and thank you for checking it out and taking the time to comment! By the end of the series you will "love spaghetti nodes!" hahaha! 😆😁
Thank you for these lessons! I would like to see some on inpainting in combination with Loras, also maybe one about how to insert an upscaler in any workflow.
using flux and also adetailer node to do faceswaps in flux would be nice to see how its done in comfyui . and then use xyz plot to do batch variations automatically without having to sit through dozens of images individually would make this the bomb ...would be cool...for a video .....
I've been wanting to cover that for some time. I wouldn't include it in this series though, instead I'd make a separate video for that on it's own. Appreciate the feedback!
Hi! First of all, thank you very much for your work. I often watch your tutorials, but I am very happy with the Comfyu series. I want to know how to paint my friends and relatives like they're by different painters in artistic painterly style .I've tried FaceSwap Lab and Roop or Reactor in Forge, but I wasn't always happy with the results and it took a lot of editing in photoshop. I've tried iPadapter FaceID workflow for Flux, which seems right, but there's still room for improvement. I would really like a step-by-step tutorial like this, and I think other watchers would like it too.
Very understandable! This step-by-step approach helps me understand the "noodles" and what connects to what much better. I do have a tip I have been doing (I'm sure I'm not the first) but your viewers might benefit from it. With so many models/loras being particular on some settings, when downloading them (mainly from Civitai, more complete info), I open Notepad and copy/ paste the developer tips and name it the same as the model. So I have a ???.safetensors file paired up with a ???.safetensors.txt in the same folder so I can go back to it when needed. Likewise, I often open up one of the "example" pics on a model page, (rt click>open in new tab or window) and it shows all the parameters for that creation, copy all and paste that into that same text file to test the model. Keep up the great work!
Glad to hear it! Then I've done my job. Great suggestion for all that information. I used to keep a running spreadsheet of all that info myself but now I kind of just know what works with certain models and lora's. Plus some platforms like Forge and Invoke AI have methods to save that sort of info. Comfyui also has custom nodes that do the same thing.
Very informative video. I hope you'll do a section on fixing hands, eyes and teeth. This is an issue even in Flux unless, of course, you're just doing "portrait of..." which is very heavily-trained. One thing that I like with Forge over Comfy is their LORA selection. You can actually use a generated image as a thumbnail right inside Forge so you can see what you're going to get. You can also set default strength, activation words and negative prompts. Granted, these break and actually the entire LORA breaks and can't load...but it's great when it works! If this can be done in Comfy, this would be huge. I suppose you could make a gen using one then use that as in img2img input, have the LORA in a group and then select the LORAs from a group of groups. Seems more bullet proof than in Forge. On second thought, just including these sample images with your node switch from this video would be even easier. Maybe add a note node to include weights, keywords, etc, although Forge is much simpler (again, when it works.) One thing to note (sorry to ramble here) is that you can change your LORA names in Explorer so they make a lot more sense inside Comfy or Forge. I also like to include a link to the Civit page the LORA is from to find prompts, keywords etc later on.
I'm long overdue to do a video on these things, there are several ways but in general those types of details almost always require some post production, it's just easier to fix with other tools. That being said, there are many ways and I will definitely make a video on it outside of this series. As for previewing loras and models there are nodes that do them but I find often times they end up not being updated and will break eventually. Hopefully the Comfyui team does something where it's default. Maybe on the desktop version once it's out of beta. But definitely agree with all your suggestions, there is still so much they can add to comfy to make the QOL experience better.
Dumb question ! Do you know how to make the manager button appear on the top ? Cuz in mine is not there :/ . Thanks for the tutorial is pretty neat btw !
If you DO have Manager already, click the cog wheel (Settings) in the lower left corner and in the "Comfy" section>Menu>Use new menu>dropdown box and you can choose to have those controls on top or bottom of screen. And there is a "Disabled" there also which may be how yours is set. Manager is the crucial part of ComfyUI; get if you don't have it!
If you haven't already installed the manager make sure to watch this part of the video. th-cam.com/video/z8efDtdBZn8/w-d-xo.htmlsi=YpbMA8K9UCcDmfiX&t=376 If you already have it, make sure to update to the latest version and if you are, check out what @metairieman55 said. It just has to be enabled in the settings.
Still to complicated! I watched Part 1, and hung in there, but in part two you talk about "Lora's"! But you don't explain what a Lora is? I'll see if I can find a real beginners video and maybe come back later!
Let me know what other comfyui topics you want me to cover! Also, these types of videos don't do to well on YT because watch time is everything. With tutorial channels people tend to skip ahead or just get what they need and bounce. Unfortunately that kills my retention metrics so any engagement, comment, like and share would really be appreciated! This is the best way to support the channel.
Gentleman, the goal is to use Flux as "The Ultimate Human Anatomy Refiner" for SD Series. That why I was asking for a workflow... Thank you for your attention.
Even in stuff you've been doing for a while now, you learn something new. These lessons are always useful. And well done. Thanks!
Thanks for the kind words! Hopefully it can help others as well.
Indeed. He shows you the structure, the elegance of the underlying logic.
Glad you can appreciate that! Always important to know also why we do what we do 😀.
Weeks of learning resumed in 2 videos, too late for me but so good for beginners. Good job !
Appreciate you checking it out regardless 😀🙌
These past few ComfyUI videos havebeen great. Or as the kids would say you're on fire bro. 👊
Appreciate the support! More fire coming soon. 🔥😁
Thank you, these videos are so useful!
You are very welcome! Thank you for taking the time to watch and comment!
Just wanted to thank you for clear and straightforward tutorial. Personally I hate the spaghetti-node concept but you just got me started 😊
You're very welcome and thank you for checking it out and taking the time to comment! By the end of the series you will "love spaghetti nodes!" hahaha! 😆😁
I've been looking for a video like this for a long time, thank you very much.
You're very welcome! Glad you found it helpful. More to come!
Thank you for these lessons! I would like to see some on inpainting in combination with Loras, also maybe one about how to insert an upscaler in any workflow.
You're welcome! Thanks for the suggestions. 👍
Thanks!
Welcome!
using flux and also adetailer node to do faceswaps in flux would be nice to see how its done in comfyui . and then use xyz plot to do batch variations automatically without having to sit through dozens of images individually would make this the bomb ...would be cool...for a video .....
I've been wanting to cover that for some time. I wouldn't include it in this series though, instead I'd make a separate video for that on it's own. Appreciate the feedback!
Hi! First of all, thank you very much for your work. I often watch your tutorials, but I am very happy with the Comfyu series. I want to know how to paint my friends and relatives like they're by different painters in artistic painterly style .I've tried FaceSwap Lab and Roop or Reactor in Forge, but I wasn't always happy with the results and it took a lot of editing in photoshop. I've tried iPadapter FaceID workflow for Flux, which seems right, but there's still room for improvement. I would really like a step-by-step tutorial like this, and I think other watchers would like it too.
You're welcome! Thanks for the feedback on topics to cover, I will add them to the list!
❤❤❤❤❤
🙌😊 Thank you for the support!
Very understandable! This step-by-step approach helps me understand the "noodles" and what connects to what much better. I do have a tip I have been doing (I'm sure I'm not the first) but your viewers might benefit from it. With so many models/loras being particular on some settings, when downloading them (mainly from Civitai, more complete info), I open Notepad and copy/ paste the developer tips and name it the same as the model. So I have a ???.safetensors file paired up with a ???.safetensors.txt in the same folder so I can go back to it when needed. Likewise, I often open up one of the "example" pics on a model page, (rt click>open in new tab or window) and it shows all the parameters for that creation, copy all and paste that into that same text file to test the model. Keep up the great work!
Glad to hear it! Then I've done my job. Great suggestion for all that information. I used to keep a running spreadsheet of all that info myself but now I kind of just know what works with certain models and lora's. Plus some platforms like Forge and Invoke AI have methods to save that sort of info. Comfyui also has custom nodes that do the same thing.
Very informative video. I hope you'll do a section on fixing hands, eyes and teeth. This is an issue even in Flux unless, of course, you're just doing "portrait of..." which is very heavily-trained.
One thing that I like with Forge over Comfy is their LORA selection. You can actually use a generated image as a thumbnail right inside Forge so you can see what you're going to get. You can also set default strength, activation words and negative prompts. Granted, these break and actually the entire LORA breaks and can't load...but it's great when it works! If this can be done in Comfy, this would be huge. I suppose you could make a gen using one then use that as in img2img input, have the LORA in a group and then select the LORAs from a group of groups. Seems more bullet proof than in Forge. On second thought, just including these sample images with your node switch from this video would be even easier. Maybe add a note node to include weights, keywords, etc, although Forge is much simpler (again, when it works.) One thing to note (sorry to ramble here) is that you can change your LORA names in Explorer so they make a lot more sense inside Comfy or Forge. I also like to include a link to the Civit page the LORA is from to find prompts, keywords etc later on.
I'm long overdue to do a video on these things, there are several ways but in general those types of details almost always require some post production, it's just easier to fix with other tools. That being said, there are many ways and I will definitely make a video on it outside of this series. As for previewing loras and models there are nodes that do them but I find often times they end up not being updated and will break eventually. Hopefully the Comfyui team does something where it's default. Maybe on the desktop version once it's out of beta. But definitely agree with all your suggestions, there is still so much they can add to comfy to make the QOL experience better.
Good content
Thanks so much, appreciate it!
Just a question! Is there a way to generate the person doing something? Eg a man taking his coat off, but it never seems to work properly.
I'm sure there is a way, you might just have to find the right words like "removing his jacket". Your other option is to use controlnet.
Dumb question ! Do you know how to make the manager button appear on the top ? Cuz in mine is not there :/ . Thanks for the tutorial is pretty neat btw !
It might be in the lower right if you haven't updated Comfy in a while (which, ironically, requires going into Manager.)
If you DO have Manager already, click the cog wheel (Settings) in the lower left corner and in the "Comfy" section>Menu>Use new menu>dropdown box and you can choose to have those controls on top or bottom of screen. And there is a "Disabled" there also which may be how yours is set. Manager is the crucial part of ComfyUI; get if you don't have it!
If you haven't already installed the manager make sure to watch this part of the video. th-cam.com/video/z8efDtdBZn8/w-d-xo.htmlsi=YpbMA8K9UCcDmfiX&t=376 If you already have it, make sure to update to the latest version and if you are, check out what @metairieman55 said. It just has to be enabled in the settings.
Still to complicated!
I watched Part 1, and hung in there, but in part two you talk about "Lora's"!
But you don't explain what a Lora is?
I'll see if I can find a real beginners video and maybe come back later!
3:36-5:13 explains it.