How To Create Consistent Characters In Fooocus

[@MonzonMedia]

Now before some of you comment "why don't you train a lora" I want to approach this without training as not everyone has the means to train lora's or models. Also, this is the most honest video on character consistency you will find! 😃

[@WILD_EPIC]

i have a question sir , like you have so many models i have only one why ? do fooocus not give us this model and after download this model how to use it in fooocus ?? plz reply

[@Hey-Its-Retro]

Thank you for another superb tutorial - I really look forward to your next one! Your videos are among the best I've seen and I really appreciate the way you explain things without any annoying distractions like flashing up 'comedy' GIFs every 7.5 seconds! Also, your tutorials are so easy to see! It's great that you screen-cap at a sensible resolution and also zoom in to what you're doing. I've seen other tutorials on other channels where it's impossible to see what's going on due to their massive monitor resolutions! On the iPad, the text is about 0.5mm tall! Sheesh! Anyway, thank you for all you do! You ROCK and that's OFFICIAL! 🤘

[@MonzonMedia]

Wow thanks my man, that really means a lot to me. I've always had the mindset to trade "flash" for value and I think for the most part most of my viewers appreciate that. This is the 3rd video I recorded on this topic because I felt the other recordings were not clear enough so I decided to just make a series out of it and take it one step at a time. 😁

[@Hey-Its-Retro]

@@MonzonMediaWell, I really appreciate your hard work and dedication as I'm sure everyone else here does! Thanks again and best wishes from Scotland! 🏴󠁧󠁢󠁳󠁣󠁴󠁿

[@taobaofy8500]

Can you show some advanced method for character consistency. You are right, nothing gives 100% same character. It would be great if you can teach us. Wish I watched your video earlier and didn't waste 12 hours, in trying different methods.

[@virgilitech]

Hello @MonzonMedia, first of all, thank you very much for all the content, I have been following your channel for a while to learn about the AI revolution, I thought you was one of the best channels for real. Putting that aside, I would like to ask if you know if is possible to use controlnet features through the API like the basic one that stable diffusion bring

[@MonzonMedia]

Hey bud, I appreciate the support and kind words. Do you mean controlnet for Fooocus or controlnet in general through API?

[@virgilitech]

@@MonzonMedia for fooocus or any other webui like forge, automatic1111.

[@dramalexi]

This helps a lot. Thank you so much.

[@MonzonMedia]

You're welcome!

[@Ryographix]

Ang galing po sir. salamat sa ideas.

[@MonzonMedia]

Welcome pare! Stay tuned for part 2!

[@iangillan1296]

Thanks for this tutorial! Didn't know you had your own channel. Can you tell please where can we download Mixpunk and Dream Haven (playground) models for 1.5 SD? It is very bad news that playground will no longer support 1.5 generations. I really love those models

[@MonzonMedia]

Hey bud! Yeah been at it a few years now, this channel is what got me the job for PG 😁 As for those models, they will be released to open source once after the 29th. Will let you know.

[@iangillan1296]

@@MonzonMedia thank you for response. It is the end of March, but there is no any news about releasing those models, or am I missing something?

[@CsokaErno]

Thank you, it is very useful, Fooocus is still good. The real question for me is that how is it possible to generate two or three consistent characters on the same picture and if it is possible to make consistent background... Or is it too complicated?

[@poussinge]

merci monsieur

[@harunsivasl4983]

Thanks for the good video good man with very calming sound!

[@MonzonMedia]

You're very welcome and I appreciate the positive feedback and support. Part 2 to come soon!

[@biorr6661]

Hey, thanks again for another great video. Iam running Fooocus in Google colab, if i want to use the model in colab, do you know which link i need to import the model?

[@bogdanarehere5855]

Thanks for share this knowledge. When did you make app for Mac?

[@MonzonMedia]

There is support for M1/M2 mac but it's considered beta.

[@BiancaMatsuo]

The next video of this series is already available?

[@MonzonMedia]

Still need to edit the video my friend, will work on that soon.

[@svetlanaLisova666]

is it possible to cut out an object or a person, leaving it without a background? The format (png) is a background object like in Photoshop

[@MonzonMedia]

Not in focus, you'd have to do that in another editor. But what are you trying to do? What are your intentions?

[@svetlanaLisova666]

@@MonzonMedia there are functions in Photoshop. but I do not know if it is in focus

[@davidwang6541]

What did it mean that you put a name : joel quero josh shiman in prompt, can it be an ID to tell system for Consistency ? Besides, the original character prototype shall be pure in its background and light effect and without full body cut ? Do I need to capture his back view also or different camera angles vertically ?

[@b.s.design]

Super! Thanks!

[@MonzonMedia]

Welcome!

[@franzschutzinger3350]

Very good Video. Easy to watch a Workflow through some function, with explonation. Im new to Foookus and KI Image, thanks this helps a lot.

[@MonzonMedia]

Glad to hear it! Make sure to check out the other Fooocus videos 👍🏼

[@HiProfileAI]

For the logos that keep popping up, why not just include logo , text and wrtiting in the negative propmpt? I've gotten mixed results with doing that but it helps overall in general

[@MonzonMedia]

Yes in the description I added that as my negative prompt, however they won't completely get rid of it but help minimize it. There is a false understanding of negative prompts, they don't necessarily work that way. Too much to get into but it still helps to put something in there. 👍

[@HiProfileAI]

@@MonzonMedia True, right about that neg prompt. As mentioned sometimes it helps. I personally find it easier to use the erase or replace in Photoshop and then bring back that image into Fooocus. But yeah trying to do as much as possible in the app I understand.

[@MonzonMedia]

@@HiProfileAI Yeah that's the idea. Very easy with a clone and stamp or healing brush. I will cover that on another video. I left things out on purpose because for tutorial video, watch time is brutal hahaha! Long form videos don't do well on my channel with people's short attention spans these days. :)

[@Pumpkin0fmayh3m]

hey I was wondering why you have "wearing hoodie" as a negative prompt?

[@MonzonMedia]

hahaha oops, probably I was generating something else at the time and just forgot to take it off....please ignore 😄

[@VNSALBT]

GG ! Thank you very munch, perfect video !

[@MonzonMedia]

You’re welcome! Part 2 to come soon!

[@byAndreasEkberg]

only right side audio?

[@Trendish_channel]

Hi, Thanks for the video Please make a tutorial how to train your own model from photos

[@MonzonMedia]

You’re welcome! Unfortunately my gpu can’t handle training but I’m hoping to upgrade some time this year. Maybe then?

[@h3video1ravenshadow]

do you use an 8 or a 16 gb nividia gpu for generating? I have a 4gb nividia gpu and its way to slow

[@MonzonMedia]

8GB, yeah 4GB can only do so much

[@AldiAnimation]

how do i use fooocus? i tried to find the url but cant find it

[@MonzonMedia]

Here you go! It’s a bit dated as it’s changed a bit but the basics are still the same and the process to install is pretty much the same. Make sure you meet the system requirements. kzbin.info/www/bejne/rWaocmZrba6DedUsi=YiAHpBFXN-n5K8SQ

[@Zero-nh4ke]

Can you show how to generate high quality transparent png image only with this software?

[@MonzonMedia]

It’s not possible in fooocus but you can upscale the image then bring it into another editor to do background remover

[@XGuarden1]

what i want to know its how make a lora or something that i can just call character name and then he know what to do. or just say in the house of MrX in the kitchen and he already can do it. will be wonderfull for manga making

[@MonzonMedia]

Unfortunately it's not possible on Fooocus, if you want to train Lora's locally you need the Koya interface github.com/bmaltais/kohya_ss and here's a good video on it kzbin.info/www/bejne/bWGrYWaZq5pqY9E&ab_channel=Aitrepreneur

[@XGuarden1]

@@MonzonMedia thanks for your reply its apreciated =:0)

[@wittchow]

I am wondering how you can get 2 characters in an image. Say I have a princess and and orange cat that is wearing a backpack and walks on 2 legs.

[@MonzonMedia]

Yes I'll cover that soon. Basically you would have to create them separately. Prompting multiple people lacks control with stable diffusion. You can do couples, and groups in general but again very difficult to control but I have a method that works quite well. Too much to go over via text here in the comments so video to come. 👍

[@wittchow]

@@MonzonMedia awesome. So yes being able to combine 2 distinct characters in a scene would be awesome. Looking forward to a video

[@jeeadvanced4719]

can you please make a vidoe on how to increase fooocus generation speed, i have rtx4050 140watt 6gb vram and it still takes 30sec to generate

[@MonzonMedia]

That's normal speeds for your card

[@dedraterru4676]

i dont have all them models available how you get those?

[@MonzonMedia]

You have to download them from a site like civitai.com and put them in your models folder. kzbin.info/www/bejne/rWaocmZrba6DedUsi=lClb_W87fGiB6uL2&t=145

[@dedraterru4676]

@@MonzonMedia Oh ok thanks. Are all the models super big in terms of space? Like the one in the description of this video was like 7GB if i remember correctly.

[@MonzonMedia]

@@dedraterru4676 for SDXL models they are mostly under 7gb, sd1.5 is under 3gb. I have a dedicated ssd just for my models 😊

[@dedraterru4676]

@@MonzonMedia oh ok and thanks for the fast reply

[@dedraterru4676]

@@MonzonMedia if im using the web version, can i still add models?

[@NotThatOlivia]

for what good is it, if there is NO FACEDETEAILER / AFTERDETAILER workflow option in Fooocus .... ? only for anime - pffff

[@MonzonMedia]

Up to this point I’ve had no need for that stuff. Inpaint works very well for adding detail.

[@falsettones]

AHAHAHA. I didn't even notice I'm commenting on my old account. I see my name! I see my name! XDDD

[@MonzonMedia]

hahaha! I use it a lot as it gives me a certain look I like 😁

[@falsettones]

@@MonzonMedia . Ahhhhh. I see your other username there earlier. XD

[@MonzonMedia]

@@falsettones Not sure what you mean...ahem! 😆

[@falsettones]

1st. 🤓

[@MonzonMedia]

😆 Yes you are! 🙌

[@falsettones]

@MonzonMedia . Talk to you later, Kuya. HAHAHA. XD

[@tikvision]

I wish everyone would stop using such idiotic and poetic words in the prompts. You are not talking to a person. The more words you use, the further the desired result goes away from you

[@MonzonMedia]

That may be true for SDXL since it has a limited token length but with the new models like Flux, SD3.5 there is a benefit for more verbose prompting however I hear what you are saying. Some words don't have any affect on the image and is considered slop. Especially if using a language model.