Add characters to ANY environment with Stable Diffusion

Insert characters into your scenes with AI,
Making them interact with your background so they feel alive and there.

------------- Links used in the VIDEO ----------

(to make it transparent go to settings - color - and lower the master opacity)

Extensions used in the video:

------------- Useful stuff for this ----------

Distortion and perspective fix:

(there are a lot more ways than just these tho)

Tips on matching the subject with the BG:

Change the light on your character in case you need to:



------------- Social Media ----------

Make sure to subscribe if you want to learn about AI and grow with the community as we surf the AI wave :3

#aiairt #characters #integrate #blend #addcharacterstoscene #digitalart #automatic1111#stablediffusion #ai #free #tutorial #betterart #goodimages #sd #digitalart #artificialintelligence #latentcouple #couple #composableLora #posing #controlnet #SD15 #inpainting #openpose #depthlibrary #AI #midjourney #interaction #relation #comic #storytelling #manga #anime

0:00 intro
0:12 Adding characters (the tacky way)
0:28 Posing the character
0:50 Important problems to solve
1:10 Scale
1:45 Perspective and lens
3:22 Adding characters (the good way)
4:22 Lighting and shadows
5:04 Adding (overlapped) characters
5:32 Creating interactions
7:18 Blending everything together (with AI)
8:45 Getting the OG background back
9:10 Final retouching tips
9:50 Big thanks!

[seraphin]

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!

[blacksage]

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

[SpeedTravelTH]

that's a nice informative video. i have never seen any other YouTubers 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

[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.

[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.

[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.

[JavierPortillo]

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

[billyn]

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.

[miguelcrespo]

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

[chaeyang]

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

[USBEN.]

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

[robmos]

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

[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!! 👌 👍

[mycotina]

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!

[gohan]

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.

[JakaJaka-kxqj]

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

[Chocolatnave]

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?

[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

[lefourbe]

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.

[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.