ComfyUI Tutorial Series Ep 28: Create Consistent Characters with Flux + Train Loras Online

Learn how to create consistent characters using ComfyUI and Flux in this detailed tutorial. In episode 28, I guide you through generating character sheets, training Loras online, and using them locally for customized character generation. This step-by-step guide covers everything from setting up workflows to using advanced tools like OpenPose, ControlNet, and upscalers to refine your results.

Get the workflows and instructions from discord

Check Other Episodes

Lora used for Eyes
Download it and place it in the Loras Folder

Unlock exclusive perks by joining our channel:

#comfyui #fluxai #aitutorial

Chapters:
00:00 Introduction
00:27 Generate Images for Lora Training - Character Sheet
17:53 Lora Online Training
26:12 Generate Images with Your Lora

[SebAnt]

Thanks, what an amazing tutorial to kick off 2025 !!!

[Uday_अK]

Wow What A Start of the year! Amazing tutorial on creating consistent characters with Flux! 🎨 Your explanation of training LoRAs online was super clear. Can't wait to try these techniques in my ComfyUI workflow. Thanks, Pixaroma!

[aysenkocakabak]

Thank you for sharing this amount of knowledge with us.

[frankiberlin]

Thank you for this excellent video again with the great explanations and workflows!
The concept is completely different to what I have seen on the subject so far - much less complex and bloated - simply to the point - great!
The subject is currently occupying a lot of people.
It's fun to follow your ideas and continue to learn from them.
Greetings :)

[tornadowatcher-myej]

I normally do not like long tutorials because I am impatient and have other things to do. n Wow, i did not know the video took a few days. Not to mention learning the program and all before hand. thank you for your hard work. I am in the middle of developing characters from scratch for a short story. lots of time goes into it, but not as much with live action actors, sets, and productions. that outro is icing on the cake. love your content./

[AiFactory-ih]

Great tutorial thank you very clear & precise ive just started to learn about Lora training so will give this a try... keep the great tutorials coming. Love your Discord server also

[welx.musica]

Was hard to look for Flux LoRA training on TA. Glad I found this!

[BirdFlu]

Dude i was literally going to ask you about this today on your discord hahaha appreciate everything man, im at episode 23 of the series. You do a great job showing and explaining how things work. i feel like im back at school again but this time its fun! lol again appreciate your work bro :D

[tuga]

Nice, I have been waiting for this one

[mystifiedbird]

thank you so much for this tutorial... exactly what I needed. See you in the discord!

[paulonascimento]

Excellent work. Thank you for sharing such valuable content.🙏

[WebmasterDeals]

Very good tutorial. I was looking for such a tutorial

[JuliusZlioba]

These tutorial series is one of the best I came across learning AI image generation with ComfyUI. Most basics I already knew, but some more advanced techniques was very insightful. Wonderful work and I will be looking forward what you going to make next.

[Ozstudiosio]

as usually wonderful video with new year dear friend :)

[kristjanretter]

Really nice results! I'm definitely going to try it out. Do you have any tips on training a style in a similar fashion?

[systeresmeraldaobene]

I've said it before and I'll say it again. You're amazing. This is simple and perfect for many applications. Thank you for your all your work and effort making our work easier and faster. Your channel will continue to grow and I'm also subscribed to your second channel.

[franciscobutte]

Your channel is easily the most comprehensive on about Stable Difussion. Could you please make a video about how fix faces using a character Loras? Or how do you upsccale an image with a Lora model attached? It didn't show it in this video at the end when you refer to it.

[mgee_]

can you ttry a workflow with two image uploaders node that generate an image referencing the two images uploaded

[fredpourlesintimes]

And if you want to put two or more consistent character in same scenes, in a story book for example?
I also think that it'd be interesting to separate characters from the scene, like working with Photoshop layer, to juste regenerate a part of the image that doesn't satisfy us.

[doctorstrobe]

Hello! I wanted to thank you for this series of videos, I learn a lot watching them. I'm on a mac (M2 Mac Studio) and I have no issues with ComfyUI+SD but flux gives me headaches, do you know about a recent version of Flux that can work well on a mac M2 (currently, the one I use works but with something like 30 minutes to generate a simple 512*512 image with a very basic workflow)?