Want Stunning AI Images? Try these PRO Tricks! Stable Diffusion, InvokeAI, Playground, A1111
HTML-код
- Опубликовано: 18 янв 2023
- Unlock the secrets to creating stunning AI-generated images like a pro. With the help of one of the most skilled AI artists on Playground. Writing effective prompts in Stable Diffusion is key to achieving truly amazing results. But these these tried and true techniques will get you even better results will less effort. These methods work across various platforms such as Stable Diffusion, InvokeAI, Automatic 1111, and Playground. Master the art of AI image generation by playing with different settings and render modes, with a special focus on the powerful Image to Image Render function to refine and elevate your images. With this expert guide, you'll discover how to take your AI-generated images to the next level while maintaining a strong composition, giving your images a professional touch.
#### Links from Video ####
playgroundai.com/
Bella: playgroundai.com/profile/cl93...
VVinchi: playgroundai.com/profile/cl93...
Use this negative Prompt: ugly, tiling, poorly drawn hands, poorly drawn, poorly drawn face, out of frame, extra limbs, disfigured, deformed, body out of frame, blurry, bad anatomy, blurred, watermark, grainy, signature, cut off, draft, duplicate, coppy, multi, two faces, disfigured, kitsch, oversaturated, grain, low-res, mutation, mutated, extra limb, missing limb, floating limbs, disconnected limbs, malformed hands, blur, out of focus, long neck, long body, disgusting, childish, mutilated, mangled, old, heterochromia, dots, bad quality, weapons, draft
Support my Channel:
/ @oliviosarikas
Subscribe to my Newsletter for FREE: My Newsletter: oliviotutorials.podia.com/new...
How to get started with Midjourney: • Midjourney AI - FIRST ...
Midjourney Settings explained: • Midjourney Settings Ex...
Best Midjourney Resources: • 😍 Midjourney BEST Reso...
Make better Midjourney Prompts: • Make BETTER Prompts - ...
My Facebook PHOTOGRAPHY group: / oliviotutorials.superfan
My Affinity Photo Creative Packs: gumroad.com/sarikasat
My Patreon Page: / sarikas
All my Social Media Accounts: linktr.ee/oliviotutorials - Хобби
playgroundai.com/
Use this negative Prompt: ugly, tiling, poorly drawn hands, poorly drawn, poorly drawn face, out of frame, extra limbs, disfigured, deformed, body out of frame, blurry, bad anatomy, blurred, watermark, grainy, signature, cut off, draft, duplicate, coppy, multi, two faces, disfigured, kitsch, oversaturated, grain, low-res, mutation, mutated, extra limb, missing limb, floating limbs, disconnected limbs, malformed hands, blur, out of focus, long neck, long body, disgusting, childish, mutilated, mangled, old, heterochromia, dots, bad quality, weapons, draft
How do you do negative prompts in midjourney
@@rots.866 write --no and then everything after that is the negative prompt
@@OlivioSarikas it doesnt seem to work in midjourney
@@rots.866 ask in the midjourney discord. prompting works differently on midjourney than on stable diffusion. so you will not get the same results with the same prompt
@@OlivioSarikas Hey man how about a video on some of the more mature extensions and scripts for automatic webui? I just started using sd-dynamic-prompts and automatics promptgen and they are amazing, and there are a number of others that are really useful or interesting, like Image Browser, Shift Attention, Animator, Embedding Inspector, Checkpoint Merger.... in fact since there are so many developments I think it could be a reoccurring theme. What do you think?
Wow Super Logic and Very Helpful..👌😊🙏
Thanks for posting, Olivio! Fantastic information as always. We were playing around with Playground this morning and stumbled across the same basic concept. Generate an image, then run it through Image2Image and vary the prompt/sampler to get variations. You added confirmation and valuable details to what we were seeing. Thanks again!
This is the third video of yours I have watched and for some reason I learn a lot more from you than most other comparable vids. Thanks and keep up the fantastic work! You rock!
Congrats on 100k subs Olivio! Thank You for shoutout!❤ Also Bella amazing tips!
❤
Thank you so much!! 😍 And thank you for your help!
Thank you Olivio. Your videos are always helpful and very informative with very clear instructions and directions. Plus I like the added info and prompts on the video description. Very much appreciated. And hey, gratz on 100k+ subs. 💯cheers🥂
Excellent tips and tricks as always !
Thank you so very much for your wonderful explanations. It's so helpful! BLESS! 🙏
Very, very helpful. Many thanks!
Amazing tips , fantastic video
Thank you! Very clear and methodical, helps me as a beginner.
Great to hear!
Olivio, I have been using just about these same work steps in Playground as well. BUT one other major factor that can provide some pretty strong influence on your render outcome is to "Show Advanced Options" (lower RH side of the screen) to change the change the diffusion sampling method. This only works with the 1.5 version, not the 2.1 version Stable Diffusion option at the top of the RH side of the screen. I end up running the same prompt through all of those options to see what I like best. In addition, on the welcome page, I use the search function to come up with a topic or style I want to emulate or experiment with. Last nite I wanted to play around creating a psychedelic poster. So by searching for that topic I really gained a head start on the effects I wanted, instead of endlessly scrolling through the MANY random creations that are presented on the welcome page. But over all, great video as always! Thanks for sharing your knowledge.
awesome, thank you!
Bella's work is amazing
Great info thanks!😊
Great video!
I really struggle with inpainting/finetuning portraits/photos people in Auto1111, can you possibly go through what each setting does, how to change prompt when doing this and what are the best values? (things like mask padding, mask blur and so on).
Useful video, especially the part about Pexels. I mainly create art in Midjourney and Stable Diffusion for video thumbnails artwork for use in videos, so very informative for my work.
Another Tipp i can share is: use a low resolution for pre rendering ideas, and than cranking that up, because the AI just don't produce the same nice results in high resolution. I'm on a pro plan and for example use 320x640 for inspiration and then refine this image on 704x1408 or 512x1024. Sometime you also get "better" ideas on 256x256 than 512x512.
pro tips 👍 Thank you
Great video! TFS
Excellent
Thanks for the information! Love your vids…been playing with Playground AI and was wondering how users hide their prompts when posting so they can’t be copied? Is it possible to edit the prompts after the image is generated?
We need a new review on 1111 dreambooth! 🙏
Can you please make a complete beginners guide to this whole space of Stable Diffusion? From someone with zero coding knowledge but some basic graphic design.
Hi Olivio !Cool stuff . I just don't have stable diffusion 2 in Invoke AI . But that is already on my machine . Where to put models in InvokeAi to be able to use them ?
Nice thank you
I like useful content of all your videos and your clear pronouncing!
Super 🙂
im still new to this ai, is there a way to do speed paints and ai doing the final renders?
amazing!
Thank you!
I would like to know, how to get a better quality from my preview pics of the four drafts shown at first. When the perfect image is already within and I don`t want to upscale, just a bigger size of the preview picture, how can I do that??? Thanks for your content!
Great video. Do "seeds" come into play when you upload an image to create another? Thank u💥🎶🖤
VAE, Hypernetworks, Textual Inversion (positive and negative) and completely changing the model...try to find out what pleases your eye!!
subbribed, im looking forward to use this very useful tool
when i try remix some of them. I get an error about APi filters activation? is there any workaround for this?
Is it possible to use playground for commercial use?
how can you change image so it shows the pose you want
Hey yes this going to be really useful for fine tuning blimey that is what is needed >> great prompt crafting Olivio (!) 🌠your images are absolutely beautiful 👌💯👸😎🌟
Is playgroundAI a subscription platform?
Is there a free plan to try out the service?
I have never seen "Remix" create an identical image, even when using in on my own previously created work where I know I haven't changed the prompt. Am I missing something?
Thanks for the vid! Do you have to use Midjourney outside of this software? Thanks!
yes, Midjourney is not a open model. But you can try to use the openjourney or dreamlike model for stable diffusion that also have a very nice quality
@@OlivioSarikas TY appreciated 👍🏻
I really like playground ai's styles but I'm very annoyed at the constant false positives from the safety filters deleting my images. So I'm wondering; how do I calibrate my local auto1111 webui to create the same image style as playground ai? is such a thing even possible? any help is greatly appreciated
hey olivio how are you? have you notice on dreamboth that you cant proceed on training with any models to create? it keeps getting error, I tried reinstall nay times SD and the whole pack but nothing is working.
Thank you! ❤
You're welcome 😊
I tried this technique again and again and every single one of the pictures I picked obviously had the prompts hidden because I got very different images. 😕
Olivio, invoke currently trunc prompts over 75 tokens so anything after that isn’t used at all. 1111 has its way to fancy merge prompts. Prompt syntax is also different between invoke and 1111 so cut and paste does not work between the 2. You can see the trunc messages only in console for invoke. I use openai tokeniser to check my prompt lengths. Most public prompts are for 1111 so plugging them straight into invoke will not work as expected
There's a tiny edit you can do to the invoke source code that will enable tokenization output in the console that is very, very useful until Invoke adds their token stuff they're working on. Ask in the Discord and someone will link it for you, I'm sure.
@@blisterfingers8169 yes i know about that. just keeping it simple cuz i see new user questions in their discords and the patience of the server mods repeatedly pasting the same link to faq or pinned messages.
supposedly we'll have it in the prompt dialog in the next release.
@@WarrenKLiu I found it a lot more useful than the openai tokeniser though. It's quite misleading, counts weights etc
And yeah Invoke is just blowing everything out of the water.
Do you have a class or online course?
I’m a Midjourney user. Should I try playground or invoke next?
No
Hey guys, I am hoping someone knows if I upload an image to generate varieties, is my upload then in their database & used to expand their services?
So...out of interest, how much worse are these images if you remove the three artists named at the end of the prompt (Artgerm, Greg Rutkowski, Thomas Kinkaide)--I.E. in a world in which the machine would not recognize individual artist names in order for AI not to essentially "commit artistic identity theft" in a future patch. Would the image quality greatly decrease?
i used your negative prompt as my prompt.. scary😬
01:03 What's the point of the prompts if they're not the real ones?
Can anyone help me? I did with stable diffusion some images that seems pratically photos, they are very original (a lot of img2img and inpaint to make them perfectly what I wanted), and they contain persons, childs, that, I repeat, seems real. Since I have to publish a book, can I admit that are my own images made with stable diffusion, or I risk to be denunced? Hoping not... there is a lot of my work in them, so many hours, and drawing part with pen tablet. 😐
My automatic1111 stops working quite often with AtrributeError "Dropdown" object has no atrribute "multiselect" and this is the second time I had to reinstall it. Anyone know why it does that?
Delete your custom extensions and reinstall them. Auto-updates doesn’t always work properly. That error for me was caused by dreambooth.
@@WarrenKLiu oh thank you. Reinstalling it right now.
I have yet to experiment with AI so I'm not astute as you are. Is Invoke the only AI that will allow me to upload an image from my desktop (I paint with photoshop any other painting software) and start there as my base? Also, which AI will allow me to generate with an oil color or water color style once I'm happy with my initial results and ready to move on with styles or art mediums?
automatic1111 or invoke will do that, theres custom ckpts and embeddings for the painted style or just prompting
👋
Why must these ai websites use different terms or measures for every tiny aspect of StableDiffusion? Why does img2img strength go to 30? The value in the code is 0 to 1. Auto and Invoke and I beleive NovelAI all use 0-1. Do they think their users are idiots who don't undertstand decimals? Is it to obfusticate the service to make it harder for poeple to transfer away from it?
nice clickbait no a1111
Whyyyyyyyy are all these guys doing such videos having their own video talking in a tiny circle in the corner??????
Its distracting...and i much rather just hear you and see the images rather than seeing you in a tiny circle on the corner
Because i'm a person, not a machine. But nice that you only care about your needs and ask me to hide my face so everything can be 100% about you
INVOKE NOT WORKING BS WASTE OF TIME .😒
unfortunately, these Ais can only make human portraits. They are worst at designing robots, aliens, creatures etc... Unfortunately a human artist is still better at creativity. Let's hope these AIs improve.
That's because human images are constantly being trained into every new model but I've seen some models that been specifically trained on robots and have created some beautiful mechs and gundams.
lol. you didn't see robodiffusion then.
@@JanBadertscher I did, tried and it doesn't work. Gives very mediocre results, it definitely doesn't give any even close similar results to the images they have put.
@@G1Vortex I don't want mechs and gundam. I want something in line with, how I, Robot movie's robot look like, even more minimalist nothing see through. I tried many models, they all fail it at excellently.
So where's the "PRO" tricks? Pretty basic stuff in the video. Lots of places to borrow prompts were even before playground (lexica, etc.) and taking others prompts isn't pro or secret, it's widespread since the dawn of txt2img, using img2img also not pro trick as well as inpainting, everybody does it for a long time, so how are these "PRO" tricks? Didn't actually learned a thing from the video, except the name of new SD-based app which there are a ton already, not paid promotion btw, sure.
Truth is there are no pro tips to being an AI artist. And "talented" is hardly the word to describe them. If you can't be bothered to make it yourself, then you're not an artist but an imposter.
Uh I know a trick, is called learning how to do your own thing without blending other artist work
The neg prompt you used in Invoke had ((())) in it. That doesn't work in Invoke.