Add characters to ANY environment with Stable Diffusion

[@seraphin01]

you're literally the most under rated youtuber around for image generation. I love how you just don't slap prompts and call it a day. keep it up!

[@Not4Talent_AI]

thank you so much!! means a lot

[@SpeedTravelTH]

that's a nice informative video. i have never seen any other KZbinrs talk about many things in your video before. like almost no other youtuber talk about how to use SD for compositing image, not in so much deep details like your video. great skill. +1 sub

[@Not4Talent_AI]

Thank you so much!!!

[@blacksage81]

I guaranteed my watch time, but damn you really just cracked open "AI Art" in nine minutes and fifty-nine seconds. What a Legend.

[@Not4Talent_AI]

hahahahahah thanks!

[@DarkStoorM_]

The videos are getting better and better 💪 Also, you probably forgot about it already, but I really recommend playing with Camera Raw Filter as the final step. BTW, mask blur of 1px can *sometimes* be pretty annoying, requiring you to blend the results in photoshop, so the default 4px is mostly safe unless you have to inpaint with a very thin mask or where 1px doesn't create a visible seam, so for things like wires or some kinds decorations, but it's very rarely used anyway. By the way, I remember comparing CNet's Inpaint model to an inpainting version of a checkpoint *without CNet*, actually getting better results without CNet (and like no tweaking required). It is a good solution though if a model doesn't have an inpainting version, but if it does, I'd recommend using it instead, it's way better for blending stuff into the background unless the uploader does a bad job with merging the Inpainting version. Using both (CNet Inpaint + Inpainting version of the model) does not work well, in case you ask.

[@Not4Talent_AI]

thanks!! I use camera raw as a final step, but in this case I didnt really know what else to touch so I left it like that hahahha And for inpaint sketch I usually use 1 pixel because I draw stuff and dont want AI to mess it up. which makes it harder to blend, but easier to mantain the shape (maybe tile could help too, tbh havent tried that now that I think of it) And true, totally forgot to mention that. I made the tutorial not really thinking about it. I guess it works out just in case people dont have the inpaint model, but now that you say this It's a good idea to let people know. tyty for everything!! super useful insights as always

[@pastuh]

I was expecting to see a video about Transparent Image Layer Diffusion using Latent Transparency, which provides more options for controlling the presentation of results. And the jumping back and forth to Photoshop just shows that it's better to use the plugin directly in the Photoshop program.

[@Not4Talent_AI]

ngl haven't tried the plugin. now that you mention it it¡s probably a good idea to look into it. Also, I have no idea about what latent transparency even is. I'll check it out because it sounds very interesting. quickly looking online, it does look pretty cool. Maybe I'll do a video on it too. Thanks for the info!! (www.reddit.com/r/StableDiffusion/comments/1b1w6tt/transparent_image_layer_diffusion_using_latent/)

[@Rasukix]

the goatttttttt

[@Not4Talent_AI]

hahahahha hey there!! tyty!

[@miguelcrespo5545]

What an amazing tutorial, thank you so much for all your work.

[@Not4Talent_AI]

thank you for the kind words! Glad you liked it !

[@JavierPortillo1]

Sketchup has an amazing tool to match perspective from a photo to 3d using vanishing lines

[@Not4Talent_AI]

really??? had absolutely no idea, do you have a link for it? Maybe I could put it on the description. Ty for the info!

[@JavierPortillo1]

@@Not4Talent_AI yeah, you can also match the grid to a known height like door in the picture or another reference and have models be proportionate, so you can do things like importing other models like furniture or people from the library. It's pretty nice! kzbin.info/www/bejne/lYeucpeBorh7e6ssi=XcRxRfm4rIzkuVyH

[@Not4Talent_AI]

cool!! tyty!

[@lefourbe5596]

Mmmh ... I don't want a pirated PS... Gonna blend layers in GIMP. Color matching will be weird btw. Krita / clip studio would be beter but there is still the learning curve.

[@Not4Talent_AI]

photopea maybe? its basically photoshop at this point hahahaha

[@USBEN.]

I didn't understand p2p and tile model use case until now. Will help a lot for my next piece.

[@Not4Talent_AI]

Hope so!

[@chaeyang5533]

this is a piece of art right here. excellent video!!

[@Not4Talent_AI]

thank you very much!!

[@3Dsnapper]

Dang, gotta try this !

[@Not4Talent_AI]

Hope it helps!

[@Steamrick]

Thank you for the information density. No need to skip through 30+ minutes of video, it's all there in compact form and I can pause if I need to take a closer look at the settings.

[@Not4Talent_AI]

glad it works well for you!! tyty! (I'll see if it is too fast for people and balance it up a bit in case it is hahaha) keeping the density, but maybe stopping a bit more for people to soak the info in before moving onto the next topic

[@mysticalfoxie]

Pretty nice video again 😍 Could get a lot out of it. I really liked your time-lapse part. Maybe that could be a cool format? If you show us your process ✨️ This was absolutely the best part imo ❤ The last part of your process with so much photoshop going on, really impressed me tho :) Very nice work!! 👌 👍

[@Not4Talent_AI]

thank you so much!!! I will probably end up finding a way to upload process videos. Either as an extension of smaller tutorials, or as a category of vids themselves

[@mysticalfoxie]

@Not4Talent_AI They are definitely easier to make, so maybe a gap filler to improve your algorithm optimization? So maybe just some "Hey guys today I created this picture" videos :)

[@Not4Talent_AI]

could work, idk if it will match with the style of the other videos tho, which kind of worries me in case I did that. What I was thinking is to post the long format more chill version on ko-fi. (for free ofc). But I need to investigate before making a decision hahahaha@@mysticalfoxie

[@mysticalfoxie]

@Not4Talent_AI Yea that's what I'm talking about !! Lofi time-lapse (4x speed max to allow slowing it down with .25x to original). Just give it a fitting title instead of tutorial, it's own Playlist and you're ready to go :) I'd gladly watch that!! :p

[@robmos9368]

Thanks a lot for sharing your knowledge. Awesome work! :)

[@Not4Talent_AI]

happy to help! Thank you for leaving the kind comment!

[@veryA72]

YOU CONTENT IS AWESOME.

[@Not4Talent_AI]

thank you so much!! Really happy to see people enjoying it

[@billyn7757]

Please if possible do a tutorial about how to make gifs please next for Stable Diffusion 1.5. Keep up the good work on these tutorials. I've learned so much in this channel.

[@Not4Talent_AI]

ty so much!!! i already had a video planned next, but I could look into the gifs thing. What type of gifs would you be interested in??

[@billyn7757]

@@Not4Talent_AI Like animated Gif images like Meme gifs that you see on Tensor or like Tik Tok shorts. Thank you again.

[@Not4Talent_AI]

Could be hard but I'll see what I can find. Btw make sure that stuff like this isnt already what you are looking for: - kzbin.info/www/bejne/q5aUont4q9GoqKM&ab_channel=SebastianKamph -kzbin.info/www/bejne/Y325on6oq6t7rLs&ab_channel=NextDiffusion -kzbin.info/www/bejne/qJXRgJuDbbFrpKc&ab_channel=OlivioSarikas @@billyn7757

[@gohan2091]

Amazing. Would love to see you create a new image like this step by step, slowed down, you talking through everything you're doing and we seeing. Doesn't matter if it's an hour long. I think that would be very useful.

[@Not4Talent_AI]

ty!!! I will do that once I'm done with some other videos I have in mind. (only 1 remaining atm). I'll first make a video on a topic darkstoorm sent. but after that one I'll go over how to create dynamic poses, and then some full process stuff probably

[@gohan2091]

@@Not4Talent_AI great. You could really get this channel exploding in viewers if you provide very slow downed detailed tutorials step by step. A lot of people who aren't as skilled as you would really find it valuable. I'll stay tuned!

[@Not4Talent_AI]

I'll see what I can do hahahaa thank you for the support!! @@gohan2091

[@mycotina6438]

You go way beyond than anyone else as usual! Though, I wonder if you would make photoshop tutorial for basic concepts you often use such as modifying the lighting & masking? Oh, even just some keywords to kickstart my research should be enough. Thanks!

[@Not4Talent_AI]

tyty!!! I guess I could eventually do something like that, I'll keep it in mind! I remember responding to a comment with some keywords to look for. Let me check

[@Not4Talent_AI]

here it was: """ I still need to learn a lot too, but if I had to name some stuff to lean it would be this: (first, this is one of my fav. photoshop learning channels www.youtube.com/@PiXimperfect) -How to use photoshop layers -Photoshop basic tools -Selections -Masking -Free transform -Adjustment layers -Blending modes -Filters and image adjustments -Brushes and how to paint Tbh most of the important stuff would be easy to learn over the course of a week by watching the pinned series on the Pixiperfect channel (Photoshop for Complete Beginners | Lesson 1) After that is just practice. Learning how to draw over all will also be a massive help with retouching AI images. (assuming you dont already ofc). Mainly how to paint, and blend. The most important parts for retouching are: adjustment layers and filters--> for the over all look of the image and blending new stuff together Layers and blending modes --> to work non-destructively and be able to add stuff Learning about clone-stamp tool + content aware fill + generative fill --> you can do this in SD too, but sometimes it's better directly in photoshop. It's to erase stuff Selection, brushes and clipping masks--> to paint new stuff or correct things without affecting the rest of the image by accident. General knowledge of the tools and what is possible--> to know how to approach a problem. Even if you dont really know how to, knowing all the tools might give you a starting idea. Like "oh, I remember a tool for this, lets look up on youtube how I can use it for this scenario" """

[@mycotina6438]

@@Not4Talent_AI Thank you so much!!

[@armondtanz]

You need to do a basics course. Ive been on photoshop nearly 30 yrs... I found this both amazing and a nightmare. So so [SOOOO] much going on. Difficult to understand why you using sliders. I use midjourney because of ease of use but im frustrated at lack of control.. Nice vid and workflow.

[@Not4Talent_AI]

oh, for sure hahaaha if you come from midjourney it's probably a lot to take in at once. Most of what I explain in the video I've touched on in previous ones. The sliders I'm touching I think should be explained in: CN models, weight and ending step: kzbin.info/www/bejne/b568lnuKia9qf9U Denoising strength and inpainting: kzbin.info/www/bejne/nnXJe5Sgjt2CmZY (probably very fast videos anyway) ty for the feedback!

[@armondtanz]

@@Not4Talent_AI hope one day you can do absolute beginners guide. Like proper noob level. I've heard all the buzzwords like loras & contolnet, but never understood the do's n donts.

[@Not4Talent_AI]

might do one eventually, when stuff looks steadier and not ever-changing hahahhaha but it's a good idea. tyty! @@armondtanz

[@armondtanz]

@@Not4Talent_AI yea. You gotta be careful. I heard MJ are working on a controlnet type system. Also if a company like adobe bring out a suite with all these features then that will shine a huge light on them. I'm shocked no one has re done the UI and introduced icons. That's half the battle with this program.. One of the initial major breakthru's with photoshop. Those little icon tools down the left hand side were revolutionary.

[@Not4Talent_AI]

yeah hahahah in the AI space I think UI is probably the last thing they do, and you are lucky if they make one ahhahaaahha Had to make one from scratch for a topic I touched in a video xd I hope firefly and mj start making CN. Cuz atm I think SD 1.5 (one of the first that came out 1 year ago or more) is still the best one to use if you want control. Only and just because of controlnet. that's why Im not making a full course yet, as it would take a loooot of time, and it might be obsolete in 3 months@@armondtanz

[@Chocolatnave123]

Thanks for the vid, I'll try some of these tips out! Any chance you can do one on generating multiple characters in the same scene?

[@Not4Talent_AI]

Thanks! Im guessing it would be the same process but with composable lora+latent couple or with regional prompter. I have a video that focuses on multiple chars interacting, it would probably be a matter of mixing both vids 😂

[@cipus2009]

Thanks a lot for this tutorial.. 👍👍👍

[@Not4Talent_AI]

thank you for watching it! hope it helped

[@bluecloud111]

Great tutorial! Why not moving to comfyui to have even more control?

[@Not4Talent_AI]

I like the manual part of A1111 for now. And I'm more used to it too. I need to test a lot of stuff every day and dont really have the time to learn comfy UI atm. Planning on learning it eventually for sure though, opens up a whole new world to explore in terms of workflows and automation. Which could be very fun to play with. Once I complete the testing I need and find a good workflow, maybe I'll start focusing on automation. Thanks btw!!!

[@itzpaco5539]

Why not using sd xl ?

[@Not4Talent_AI]

sd 1.5 has tile controlnet and way more (and better) cn models. Anything that isn't controlnet tile or brightness, can be done with SDXL

[@lefourbe5596]

TO MY DEMISE :/. TILE is great at coherent enhancement of details but was never updated for SDXL. IP2P don't works too IIRC. X adapter is not yet ready to save us. Not even in comfyUi just yet.

[@Not4Talent_AI]

the day all of that goes to sdxl I'll be so happy. Or sd 3.1 at this point XD@@lefourbe5596

[@DarkStoorM_]

@@Not4Talent_AI Well, SDXL also has tile model, but apparently it "works best on ComfyUI" only or should be used for upscaling purposes only (ultimate sd upscale), so, pretty useless for things you do in this video ツ People say it's amazing, but I still haven't seen a good proof of that vs SD1.5

[@JakaJaka-kx2qj]

i am heree, hello friend. I verry enjoy with your video

[@Not4Talent_AI]

Hahahaha hii! Happy to hear that, tyty!

[@LexChan]

i like your tutorial for this.. but i think u need to slow it down and get more details and sample for this topic. if possible.. please include specific object (ready images. not promt out) to be blend into the image. but by far.. this is the closest AI photo manipulation i think most ppl want to know. dont know why.. most youtubers didnt get this right on what user really want... pls do more on this type of video. especially the perspective and lighting part... hope to get more information on it.. thanks

[@Not4Talent_AI]

I get a few samples before recording, but I show just some of those bc in past videos showing multiple ones got confusing. If I find a way to avoid that I could try adding more. I could make another video with what you said tho, sounds like a good idea. Idk when, but Ill add it to my list for sure. Thanks for the feedback and the ideas!!

[@seifergunblade9857]

tq

[@Hotsnown]

the davinci of image generation

[@Not4Talent_AI]

hahahahhah ty sm!!

[@Showbiz_Stuff]

Thanks for the video! Any idea why I have no instructp2p or where I can dowload the Controlnet model?

[@Not4Talent_AI]

weird, it should be in the regular controlnet models list. huggingface.co/lllyasviel/ControlNet-v1-1/tree/main "control_v11e_sd15_ip2p.pth" + yaml

[@Showbiz_Stuff]

@@Not4Talent_AI thank you!

[@onlyyoucanstopevil9024]

GG

[@Not4Talent_AI]

tyty!! I guess hahahah

[@bilalmurtaza8489]

any way i can use control net inpainting without downloading i have a amd gpu downloading stable diffusion locally donot work properly

[@Not4Talent_AI]

You can try to look for google colabs or pages like runpod Ai. Id like to know if there are other options but havent really tried atm

[@sherpya]

what about layerdiffusion?

[@Not4Talent_AI]

just learned about this through the comments XD I still dont think it would have helped much for the video. but probably worth mentioning anyway

[@LostBoyRichJ]

Can you also add someone or a person from an existing photo? Let's say you have created an environment with some people in a scene, but you also want to add someone else (who is not AI generated) is that possible? Cant find any tutorials about that

[@Not4Talent_AI]

Ive tried but atm I think AI changes the person too much. You couls try to add someone just like I did in the video. And then try to replace the face with a face swapper? Other than that my best advice is to composite directly with photoshop

[@LostBoyRichJ]

​@@Not4Talent_AI Thanks for your reply, i think i will stick to photoshop for tasks like this, but hopefully in the near future stable diffusion is able to do this

[@Not4Talent_AI]

@@LostBoyRichJ lets hope so hhahahaa npnp!

[@vikasverma2580]

Man this is crazy good.. But please make it an hr video where you explain everything slowly. It's just too fast to comprehend these many things at once.

[@Not4Talent_AI]

Might do one then. I want to start posting a long, slow vid to compliment each tutorial. Was planning on doing that starting from the next video. But maybe after that Ill do one for this

[@vikasverma2580]

@@Not4Talent_AI Some suggestions for your channel. Make each an hr long with focus on a single topic , for ex, basics of photoshop(that will be used in editing ai images), basic to Adv controlnet, basic to Adv inpainting, etc. Hope you get what I mean. Long hr videos will also increase your watch time.

[@vikasverma2580]

@@Not4Talent_AI also you can upload some short videos of 10min for those who already have Adv knowledge but do not know how to play with that.

[@Not4Talent_AI]

@@vikasverma2580 Hmm Idk about a tutorial on photoshop bc it's not really my target audience. But I could do one that's kind of like that if I make a tutorial on how to post process. (I still need practice before I can do that tho)

[@Not4Talent_AI]

@@vikasverma2580 The main focus would be this one, I mainly do tutorials for people who already know what they are getting into. There fore they are faster and just explain the idea more than a detailed how to. The hour long tutorials would be a reinforcement of those shorter ones. Just for people to see it in a slower format and have more time to take it in. But I dont plan on having a super formal script for those. Just explaining a bit of my though process and that's it. Dont have time to do more tbh. Ty for the kind comment and the suggestions btw!!

[@blasteekmag]

hi is there away to import a model or object to the background?

[@Not4Talent_AI]

Thats harder to do I think. You could follow the same process of composing the two toghether and then re-genersting. But youd need to fine tune the details a bit to get it matching properly. Maybe 2 seps. The first to match the lighting with a higher denoise. And then re-gain the character using inpainting and IP adapter

[@NicoSeymore]

Everytime i try to inpaint or inpaint sketch in empty areas i just get my mask as a black blob in output. Any ideas maybe?

[@Not4Talent_AI]

hmmmm wtf, are you getting any Nan errors? or maybe using some parameters that might interfiere?

[@NicoSeymore]

@@Not4Talent_AI it worked again after restarting the system lol dont know. thank you! Your tutorials are the best out here!

[@Not4Talent_AI]

@@NicoSeymore happy to hear ! And thank you so much ahhaha

[@burakgoksel]

It seems to open under control net, when I install it it does not open in a separate tab. What am I doing wrong, I can't put the puppet on the picture like here under controlnet.

[@Not4Talent_AI]

Hi, Im sorry but I dont really understand what you mean. It looks like if the message is missing one part. What seems to open under controlnet?

[@burakgoksel]

@@Not4Talent_AI i mean, openpose menu(upside txt2img,img2img... [openpose editor] eg) - doesnt appear, because of last version of controlnet. I will send a pic in discord. I think, i can ask in help-question channel right?

[@burakgoksel]

With the latest version of the Controlnet 1.445 plugin installed, when you install OpenposeEditor ext., it does not appear in the menu.(txt2img,img2img,extras...) I reset everything and installed opEditor first, problem solved.

[@Not4Talent_AI]

@@burakgoksel sure!

[@Not4Talent_AI]

@@burakgoksel ohhhhhh ok, thanks for sharing! thats annoying tho xD

[@tr1pod623]

Thank god I know how to draw makes everything way easier

[@Not4Talent_AI]

hahaahhaha 100%

[@avaric3z]

Too fast, and not show all processes/details. It is not how a tutorial video should be.

[@therookiesplaybook]

This is not EASILY

[@Not4Talent_AI]

Depends. To add the character is extremely easy. To add it good it gets harder. But its a pretty simple process aside from the drawing part imo