Stable Diffusion Inpainting Basics
HTML-код
- Опубликовано: 22 апр 2023
- Learn the basics of using stable diffusion inpainting with this easy-to-follow tutorial. In this video, I’ll cover everything you need to know about stable diffusion inpainting, including how it works, its benefits, and how to use it. Whether you’re a beginner or an experienced user, this tutorial will help you get started with stable diffusion inpainting in no time. Watch now and start creating stunning images today!
In this video I cover topics such as:
- Mask blur
- Inpaint area
- Masked content
- Masked padding
nice work. Have been inpainting for months, but you've really helped me in this review understanding whats happening.
Very nice tutorial! Perhaps the best I've seen so far, people don't explain those settings well.
You have a nice way of explaining things. You go far enough into the details to make them understandable without getting too bogged down in them. Nice job! Subbed.
Thank you
I appreciate the pacing of your tutorials. Very simple and straight forward.
I dove into SD beginning '23, downloading extensions and models left and right, quickly got frustrated when I didn't have the fundamentals covered. You've done well with this demo, thank you.
Video Feedback: With respect, I want to offer some thoughts on script/editing. If you find them useful, let me know...
1. The inpaint scenario is clear when you ***show*** an example of [red tie hitman]. Noting the affect of each slider/variable is very helpful. However, when your explanation goes beyond the actual demonstration, say "generating window on windowless house" analogy, some may follow your meaning but the video doesn't ***show*** what your conveying. YT is a visual medium. Add a couple more process images about that analogy to drive home your point.
2. Masking is an age-old process from the film photo development process. Pointing out affect of blur-radius is key, very good. Any reason for cropping the SD ui window? I have trouble following the white-cursor on the white-ui-browser. A filled ui (+zoom, browser) + more cursor contrast (or +size) might help to direct attention where you intend it to be.
3. Making reality of additional ideas is probably what I enjoy most about AI image gen/edit. Aligning your intent with the randomness of AI noise is no-doubt the challenge; a skill to work at for anyone. The wardrobe remix is a relatable concept. Considering the timelapse (process abbreviation), the "power" of inpainting isn't so much the end result, more so in the "control" of masking/guiding diffusion to one's intent. ***Show*** the image stages you had along the way, include your thought process. It could even be just a few images side-by-side, sans-timelapse. Original vs Final doesn't do justice for your efforts. Everyone learns if you share what had to happen in the middle.
All-round: An intuitive order of presentation for this cumbersome interface. Looking forward to more in the future. Cheers.
Your feedback does make sense and I will try and see if I can incorporate them in future videos.
I can understand you frustration with stable diffusion, the technology moves so quick and there are so many options. I personally find the best thing for this is to spend a period of time fully understanding one part such as inpainting, before moving onto another.
Thank you for your comment and constructive feedback, and I'm glad you enjoyed the video :) .
very helpful
Thank you for this fantastic tutorial. You have an excellent teaching style.
Thanks a lot, this is the most ever clever illustration for inpainting in SD. Keep the good work!
You're an awesome teacher, subbed
Very good, I agree with earlier comment about visual examples are better than explanations. Really good, hope you do more.
good explanation, you gain a subscriber
Nike air jordans haha. Nice touch and good tutorial!
for some reason you giving 47 a pair of Jordan absolutely cracks me up
Awesome, thank you man.
Seeing some of your video.
No bullshit, some real exemple and great elocution, you win new follower =)
Great job !
Very clear explanation, nice work.
Glad you enjoyed the video
Thank you for the video
Love the tutorials sir, ty. easy to understand. but the analogy windows is break my brain more than it needs. xd
aah agent 47 wearing jordans.. such a classic!
Hi! you explained inpainting really well and it helped me to be sure what i was doing was correct! thank you!!
I have a request / idea for your next video. basically, i have an image of a girl, but i would like to change the background what she is in. When i try to use inpainting, if i set the denoising strength too high, SD will sometimes add in extra hairs, extra blurry outlines, maybe extend the girl's stomach outwards, make her shoes or legs have some jagged edges or just have an 'aura' around her body. i have tried using the upload inpaint function to get a more accurate masking of the body of the girl, but SD still ends up adding alot of irrelevant stuff on her body... if i inpaint less of her, then SD might change her body structure (eg, make her waist thicker, boobs smaller, legs thicker, etc.).
i was wondering if you could do a tutorial on adding backgrounds without changing anything about the person in the image or having a weird 'aura' around the person? oh and definitely she is supposed to blend in with the background seamlessly.
thank you so much
Masked content AFAIK (maybe im wrong on how it works)
Fill = following the prompt ignoring the base image. (making new stuff) good for outpainting.
Original = following the prompt + the base image (detailing)
Latent noise = it will fill the image with the surrounding pixels, kinda like content aware fill in PS, it can follow the prompt use it at high denoising strength. don't use prompt if deleting stuff. (delete stuff, fill it with the surrounding)
latent nothing = just blur stuff idk what it is. xd
What acc'ent is that?
out of frame !