Hello everyone! I didn't expect this video to blow up and attract so many new viewers and beginners! I am currently working on the next version which will be even better, and this time I will also show the complete installation process from start to finish, so you will have an easier time installing everything!
Can't wait😍 A very cool thing would be to integrate instax union pro ControlNet into the latest workflow, so that you can generate the final images with a pose of your choice. That would be crazy
Wow this was so badass, it took me a while to set up because I am quite the noob in ComfyUI but the workflow was flawless. I ended up using another ControlNet model and voila. I even added the Load LORA object in the first step so I could generate the character sheet for a LORA. Super cool! Can't wait to follow your other tutorials
This is great. I've been doing 3x3 character grids, cropping them out manually in Photoshop and then doing post work on each. This will be a big time saver for creating character data sets, thank you :)
Thank you SO MUCH for this video. I literally spent 2 weeks trying to better understand how to create consistant characters, and training a lora. I finally got something to work, but wanted to try something better. I'll be giving this workflow a shot, and that Lora trainer. Your content is so good, at this rate you are going to end up with a new Patreon member. 😂
Thanks a lot for sharing your sweet work! Gonna try this out this wknd. Please continue uploading flux workflows for consistent characters. I'm really curious to test it out for my project.
Mastered the advanced functions of flux and comfyui... also uses paint to crop photos. 😅 Jokes aside, this is really impressive. I used to use an equivalent setup on webforge, but much more primitive, to generate reference photos for my 3d modeling. Basically I had a simple rendering with my 3 views (face, side and 3/4) as input and I used inpainting to generate a set of reference photos. But that's another level !
Just found you and couldn't agree more with Matt. Brilliant! Just subscribed to Patreon too. Thank you so much for breaknf all that down in a super digestible way.
It' really great to see your tutorials developing just with the opportunities of AI 😄 Keep up your great work and thank you very much for sharing so much knowledge for free!
I admire the effort u put into making a complete workflow from 0 to 100!. I'm Feeling excited to try it on my 4070super.... its gona be slow but I hope it gets the job done.
Have you already tried this workflow on that graphics card? I'm looking to upgrade my PC and I work with generative AI, but I don't know which card to buy.
i downloaded your tutorial and got this working!! that's remarkable because i struggle with this stuff a lot, but yours worked REALLY WELL!! TYSM for sharing! i'll probably do a video on my YT about ComfyUI some time and i'll have to give you a shoutout.
thanks for the video, it is amazing! one thing i noticed you not take the last lora but the 1 before. the last lora will be the lora without the digit. you can see it by the timestamp to.
just use pics of your own face from different angles, no? there's also ways (though I can't remember how) of using character sheets like he has, but with an existing pic as a reference, rather than generating a new character as he did.
@@encartauk I just didn't get a sense that there was a way to import files. I've watched others create using this format as well and nobody speaks about it.
@@ToddKauffman I've never used comfy, so I'm not sure at what stage you'd enter in your own pics. But there was a video i saw once that had about using character sheets in controlnet in Automatic1111, and then a reference pic of someone (even ai generated) to create the other angles of their face. It didn't work that great for me though in terms of likeness.
I have a question that no one knows! You are skilled and you know. With which ai or app or software, can you produce realistic similar photos (like flux or real graphic) from a sample photo? Of course the photo of humans A kind of clone means
You should make a workflow where it generates the characters with a contrasting background so you can green screen them wherever you want, it’ll eliminate the hassle of having the characters interacting with each other in the panels
I once read that to train Flux Loras you don't need to tag the images, I'm telling you in case you want to try it the next time you train a Lora, greetings!
yeah you don't need to tag it yourself but if you want to have better control over the picture that's gonna get created then you should tag it yourself. For example if you do tak the picture of the whole figure of person and tag it "picture showing full figure, looking straight at the camera" you gonna have more chances that it gonna create the picture you have in mind otherwise you have to kind of get to know how the flux scripts this command itself. So in the end you get to choose rather you want to spend some time tagging it by yourself and have better control over image creations or you just gonna learn how the flux scripts the positions of character
This is amazing! I see the option to use 2 LORAs, can one be for image style/cartoon style, and the other one for character? And combine these two? That would be insane
This is most exciting. Thank you so much for making this. What if I already have a character that I have designed in Mid Journey? Can I still import my character through your workflow so that I can later train an AI model?
Something worth mentioning is to recommended toggling advanced options and enabling buckets. This allows people to add images of varying aspect ratios. Otherwise, they will be limited to 512x512, 768x768, 1024x1024, etc. If they don't enable buckets and forget to use same aspect ratio images then their images will get cropped in the training process and ruin the model ie faces being half cropped out.
Super interesting. Thanks for sharing. I wonder - would it help to have character images in different lighting situations or is this completely irrelevant? In your example the woman looks like she is always lit with studio lights - a little bit i think or am i imagining things?
Thanks man this is great! will definitely consider subbing on pat soon if you keep these kind of guides coming :) Would love to see some comfy to code guides or app would be awesome too.
Awesome, will try this next week. I was wondering, is this also possible with a product like a shaver or synthesiser or any other product? No eyebrows, emotions...must be more easy... What do you think?
well done mate. loved the video. Just came across to your channel, and considering supporting you to try this workflow. My fear is I only have an Nvidia GTR 12VRAM will that be enough to play with Flux? I don't want to be waiting for hours for an image lol
i want to see a video where you will open civitai and select one picture check the details prompts checkpoint lora etc, and and build same workflow in comfyui and build the images !! , can you do that !! and try to do with may different photos and models
FYI, the non-numbered loras in the fluxgym output directory are the latest ones. You were using loras from epoch 12 (not necessarily a bad thing but you mentioned latest in video).
A lot has been done to get consistent faces, I am looking forward to when the outfits they wear can be more consistent. I had to settle on a very simple style to get some clothing consistency but it still varies wildly. Unfortunately I can't use flux locally as my laptop has no video card. Someday I hope to be able to take advantage of your excellent videos I watch them all and sooooo want to use your advice.
What if you made additional images based on extreme poses and angles, like combat poses, that way the AI is consistent even with clothes moves about and would want to create frame by frames? Then maybe use runway to interpolate the frames? Or maybe a custom comfy setup that can animate efficiently? Maybe control the position and ditance of the light angle? I'm curious on how things would go if we went full control into video designing! Maybe train how clothes work in certain artstyles, and transfer to a base model [underwear or not]. There's so much logic in this video it's drawing so many ideas to my head lol. Probably an option to link to an external/cloud AI GPU to make this quicker to render.
love your videos , is there a way to input an image of a character and input a pose image and apply a controlnet on using both input images (keeping clothes,face consistent )? or using controlnet with the lora is the only way to achieve this ?
This is the sort of thing I've been after for a long time. If I already have a character that I like, can I face-swap early in the workflow so that all of the images have that face? A year ago, when I created a LoRA, it took 7 hours to generate, would be interesting to so how long this takes.
Hello everyone! I didn't expect this video to blow up and attract so many new viewers and beginners! I am currently working on the next version which will be even better, and this time I will also show the complete installation process from start to finish, so you will have an easier time installing everything!
When will the new video come?
waiting for it
Can't wait😍
A very cool thing would be to integrate instax union pro ControlNet into the latest workflow, so that you can generate the final images with a pose of your choice.
That would be crazy
Awesome! Thanks!
Amazing video but I'm a complete bigger with confy UI I’m really looking forward to see the the installation process as well
Thank you. I realized that I am barely scratching the surface after watching your tutorials. God bless you!
Never seen such complex great working workflow Thanks a lot i did get great results
I am amazed about this PRO tutorial. I wish you could create a more slow step by step process for newby people like me! THANKS!
😄
That's right, this guy thinks he's creating content for the highest echelon! Go easy on us! hahahahaha
please make
Try this adjust your playback speed in TH-cam with your settings
Wow this was so badass, it took me a while to set up because I am quite the noob in ComfyUI but the workflow was flawless. I ended up using another ControlNet model and voila. I even added the Load LORA object in the first step so I could generate the character sheet for a LORA. Super cool! Can't wait to follow your other tutorials
can you please share your workflow?
This is great. I've been doing 3x3 character grids, cropping them out manually in Photoshop and then doing post work on each. This will be a big time saver for creating character data sets, thank you :)
Thank you SO MUCH for this video.
I literally spent 2 weeks trying to better understand how to create consistant characters, and training a lora. I finally got something to work, but wanted to try something better.
I'll be giving this workflow a shot, and that Lora trainer.
Your content is so good, at this rate you are going to end up with a new Patreon member. 😂
How's that going right now?
Thanks a lot for sharing your sweet work! Gonna try this out this wknd. Please continue uploading flux workflows for consistent characters. I'm really curious to test it out for my project.
compared to other ai voiced AI tutorials, the effort went into this video is isane bro. keep it up :D
Mastered the advanced functions of flux and comfyui... also uses paint to crop photos. 😅
Jokes aside, this is really impressive. I used to use an equivalent setup on webforge, but much more primitive, to generate reference photos for my 3d modeling. Basically I had a simple rendering with my 3 views (face, side and 3/4) as input and I used inpainting to generate a set of reference photos. But that's another level !
at 4:00 yea she is happy in different format :) great tutorial, I am trying to set it up after getting a decent nvidia card. Thanks for your videos!
The guy is a genius and truly ahead compared to others.
Just found you and couldn't agree more with Matt. Brilliant! Just subscribed to Patreon too. Thank you so much for breaknf all that down in a super digestible way.
The man keeps delivering. Million thanks.
Hello man. Can I ask you a question about this workflow please?
@@EternalAI-v9bof course but I think you meant to ask the creator and not me
I really appreciate you keeping us all up-to-date with the cutting edge- how soon before this can do 12 people consistently in the same shot?
I have said it before, and I will say it again... "BRILLIANT!" Thanks so much for sharing your incredible insights and workflows.
Hello matt, is there any way I can ask a question about the workflow please?
Your Channel is totally underrated. Fantastic learning curve... I tried your 2d set extension workflow couple of weeks ago... Awesome, thank you!
It' really great to see your tutorials developing just with the opportunities of AI 😄
Keep up your great work and thank you very much for sharing so much knowledge for free!
Had two of your vids open in different tabs and subbed on both vids.
Just subbed twice 😎
You are an insipiration. Following you from the beginning & look how far you have come. Congrats & thanks for your guidance. 🔥🔥
SO INFORMATIVE! Great tutorial, can't wait to try this! Thanks so much!
This is absolutely brilliant! Thank you for figuring this out and sharing with AI image generation enthusiasts!!!
This is insane. Thanks for sharing your workflow!
So much useful information! Thank you so much, keep up the good work!
thank you man for sharing and keeping them free. YOU ARE A SAINT
I admire the effort u put into making a complete workflow from 0 to 100!. I'm Feeling excited to try it on my 4070super.... its gona be slow but I hope it gets the job done.
Have you already tried this workflow on that graphics card? I'm looking to upgrade my PC and I work with generative AI, but I don't know which card to buy.
@@trabajoxmoney6857 I'm downloading the models and stuff atm. Will update once it's running.
Not as slow as you think
The 1st step is very excellent, It can helps me a lot in my creation. ty 4 sharing it!
This is fantastic and will help my workflow so much. Thank you!
verrrrrry cool ! can't wait to try this later ! thank you for all the work you did on this!
I really like the part of the entire video that is excellent! Love it!
i downloaded your tutorial and got this working!! that's remarkable because i struggle with this stuff a lot, but yours worked REALLY WELL!! TYSM for sharing! i'll probably do a video on my YT about ComfyUI some time and i'll have to give you a shoutout.
got a discord? i wanna learn we wanna build our own data set
thanks so much for these amazing workflows. So understandable and brilliant!
thanks for the video, it is amazing! one thing i noticed you not take the last lora but the 1 before. the last lora will be the lora without the digit. you can see it by the timestamp to.
You crushed this topic. Thanks for sharing. Gonna subscribe and donate to Patreon.
This is great - we need this but with the ability to add your own face or previously created character model
just use pics of your own face from different angles, no? there's also ways (though I can't remember how) of using character sheets like he has, but with an existing pic as a reference, rather than generating a new character as he did.
@@encartauk I just didn't get a sense that there was a way to import files. I've watched others create using this format as well and nobody speaks about it.
@@ToddKauffman I've never used comfy, so I'm not sure at what stage you'd enter in your own pics. But there was a video i saw once that had about using character sheets in controlnet in Automatic1111, and then a reference pic of someone (even ai generated) to create the other angles of their face. It didn't work that great for me though in terms of likeness.
Just use your own character sheet to make the LoRA. You don't need most of this process.
Amazing video! This is so complex, but you explain each every step
Man, you're great, you saved me and this industry!)
Thank you so much for this informative video. You fulfilled my request. Thank you.
I have a question that no one knows! You are skilled and you know.
With which ai or app or software, can you produce realistic similar photos (like flux or real graphic) from a sample photo? Of course the photo of humans
A kind of clone means
You should make a workflow where it generates the characters with a contrasting background so you can green screen them wherever you want, it’ll eliminate the hassle of having the characters interacting with each other in the panels
great ! just a small thing.. the last Lora generation is the one at the top with no numbers..Thank you
Yep, I left the same comment before reading this. (Don't worry, I deleted it. lol)
truly fantastic! amazing tutorial🔥
Dude, you are amazing!! Thanks a lot!
I love this video tutorial! 👍😀 Well done!
Great video. No hype, clear and useful. Many thanks.
Hello man. Can I ask you a question about this workflow please?
Very interesting! For a consistent character with different outfits, would you just create a sheet with just head shots and expressions?
I once read that to train Flux Loras you don't need to tag the images, I'm telling you in case you want to try it the next time you train a Lora, greetings!
yeah you don't need to tag it yourself but if you want to have better control over the picture that's gonna get created then you should tag it yourself. For example if you do tak the picture of the whole figure of person and tag it "picture showing full figure, looking straight at the camera" you gonna have more chances that it gonna create the picture you have in mind otherwise you have to kind of get to know how the flux scripts this command itself. So in the end you get to choose rather you want to spend some time tagging it by yourself and have better control over image creations or you just gonna learn how the flux scripts the positions of character
@@elninio92 Ah, no tenía ni puta idea de eso, gracias por el dato.
Wow that was really good
Thank youuuu!
Excellent video , smooth learning 👌
Awesome tutorial and thanks for the workflows and tutorials! I was wondering - how long does it take you a run trough the complete workflow?
This is amazing! I see the option to use 2 LORAs, can one be for image style/cartoon style, and the other one for character? And combine these two? That would be insane
Awesome Video Mick, Great work
Really great video, will give this a go.
@Mickmumpitz : This is an amazing tutorial. Love what you are doing. Thank you for sharing your knowledge 🙌🙌🙌
Finally a quality video I stumbled upon.
Great work! Thank you for the video
AMAZING! What if we need the characters to change outfits?
bro..... amazing
This is most exciting. Thank you so much for making this. What if I already have a character that I have designed in Mid Journey? Can I still import my character through your workflow so that I can later train an AI model?
Using kling you can do it as I know it
Thnak you so much for amazing videos , can you make video on how to make consistent character pose of custom character or self potrait .
Something worth mentioning is to recommended toggling advanced options and enabling buckets. This allows people to add images of varying aspect ratios. Otherwise, they will be limited to 512x512, 768x768, 1024x1024, etc. If they don't enable buckets and forget to use same aspect ratio images then their images will get cropped in the training process and ruin the model ie faces being half cropped out.
U earn follow. Please more share "how to create character" concept like this. This is amazing work.
Thank you! I'm already working on the next version!
Oh! This is reakky good! Thank you!!
thanks for this!
Flux is such a game changer,
You are incredible !
Thank you! Very nice tutorial.
greatest one ever!
damn nice video man, taught me alot. keep up the good work!!
Super interesting. Thanks for sharing. I wonder - would it help to have character images in different lighting situations or is this completely irrelevant? In your example the woman looks like she is always lit with studio lights - a little bit i think or am i imagining things?
Nicely done
Thanks
Thanks man this is great! will definitely consider subbing on pat soon if you keep these kind of guides coming :)
Would love to see some comfy to code guides or app would be awesome too.
Awesome, will try this next week. I was wondering, is this also possible with a product like a shaver or synthesiser or any other product? No eyebrows, emotions...must be more easy... What do you think?
well done mate. loved the video. Just came across to your channel, and considering supporting you to try this workflow. My fear is I only have an Nvidia GTR 12VRAM will that be enough to play with Flux? I don't want to be waiting for hours for an image lol
Before I watched this video, I thought I know one or 2 things about generative ai…well not anymore. Fire vid!
amazing, good job !
i want to see a video where you will open civitai and select one picture check the details prompts checkpoint lora etc, and and build same workflow in comfyui and build the images !! , can you do that !! and try to do with may different photos and models
You are the best mate... Blesses
Simply easier and superb
FYI, the non-numbered loras in the fluxgym output directory are the latest ones. You were using loras from epoch 12 (not necessarily a bad thing but you mentioned latest in video).
correct
A lot has been done to get consistent faces, I am looking forward to when the outfits they wear can be more consistent. I had to settle on a very simple style to get some clothing consistency but it still varies wildly. Unfortunately I can't use flux locally as my laptop has no video card. Someday I hope to be able to take advantage of your excellent videos I watch them all and sooooo want to use your advice.
What if you made additional images based on extreme poses and angles, like combat poses, that way the AI is consistent even with clothes moves about and would want to create frame by frames? Then maybe use runway to interpolate the frames? Or maybe a custom comfy setup that can animate efficiently? Maybe control the position and ditance of the light angle? I'm curious on how things would go if we went full control into video designing! Maybe train how clothes work in certain artstyles, and transfer to a base model [underwear or not]. There's so much logic in this video it's drawing so many ideas to my head lol. Probably an option to link to an external/cloud AI GPU to make this quicker to render.
Amazing work!
thanks bro
you make it harder and harder. thank you.
Thanks a lot buddy. Great video.
Great tutorial! Can you get two (or three) consistant characters in same image?
Wow! Great tutorial!
Thank you for the great video. But I want to ask before trying: Can a 10-year-old laptop do this?
waiting for it
Wow, this is mind-blowing! Could you share where I can find good resources or guides to get started? (I am beginner)
love your videos , is there a way to input an image of a character and input a pose image and apply a controlnet on using both input images (keeping clothes,face consistent )? or using controlnet with the lora is the only way to achieve this ?
Yes please
This is the sort of thing I've been after for a long time. If I already have a character that I like, can I face-swap early in the workflow so that all of the images have that face? A year ago, when I created a LoRA, it took 7 hours to generate, would be interesting to so how long this takes.
Un travail incroyable 🎉
THANK YOU
Great job! Thank you!
Love this ! Thanks. How would you suggest I run this workflow through API ? I don't have a powerfull enough machine but i really love this workflow
Great video! Well done.
Liked and subscribed. Great stuff
yea those look good