My Top 4 Favorite Ai Models - Civitai / A1111 / Stable Diffusion

Here are my Favorite AI Models. Why and how I use them. In this Video I show you Rev Animated, Realistic Vision, majicMIX realistic and Photon. These Models are Perfect for A1111, InvokeAI, ComfyUI and Google Colab

#### Links from the Video ####
Bonus Models:

#### Join and Support me ####

00:00 Intro
00:26 Rev Animated
04:46 Settings Guide
05:49 Realistic Vision
10:38 majicMIX realistic
14:39 Photon
16:51 Bonus Models

[OlivioSarikas]

#### Links from the Video ####
Bonus Models:

[Hexrun]

my favorite models are: AbsoluteReality, CyberRealistic, DreamShaper, Lyriel, A-Zovya Photoreal, Realistic Vision

[Simsonlover]

its SO AMAZING... the benefits of stable diffusion beeing open software and beeing able to train models and stuff... this makes it really awsome... i like dalle3 for its quality, but the restrictions and the slow speed makes it really annoying... but with this you could do literally anything you desire

and the quality got so good over the months... i remember the first version of stable diffusion
i was so impressed... even if the quality was not that good but seeing how it generated images and put characters into different styles... hit is what amazes me the most... its almost like magic... painting with reality and putting different characters and scenes into different genre and art styles while still seeing where its coming from...

i like dall e3 for prompt understanding and realization... for example i asked dall e3 (before it got resricted), fallout 4 power armor warehouse in cyberpunk 2077... and it did it so well... you saw the power armor you saw the cyberpunk elements and even signs... in stable diffusion i could not get it to work properly... but maybe with different prompting? i mean you could train your model very specically for every picture you wan tto make... but dall e3 just did it somehow


but still stable diffusion is so great putting out so much details and they will improve it even further

[tikishark]

Thank you for this information. I'm always on the look out for new models, and this was a very comprehensive overview. I also really appreciate the files to match. However, since SDXL has come out, most of my workflow has now shifted to ComfyUI and the model (v5 is also available). I've gotten some amazing results in SDXL, things I could only dream of with SD1.5

[AI.Creative.]

Very good models, thank you Olivio. I got good results :)

[WickedCitizen]

CyberRealistic, Juggernaut, pirsusEpicRealism, WyvernMix, I Can't Believe It's Not Photography, and ConsistentFactor are my favorites (other than ones already mentioned in the video).

[Afrmanpeace]

My favourite model for SD 1.0 XL is of course XL6 Hephaistos 2.1

[WifeWantsAWizard]

Since he brought up #SDXL1 models at the end there, I'd like to add "Juggernaut XL" if you're looking for realism. In my opinion, this particular model appears to be trained on thousands of images using the "cinematic" prompt keyword.

[victorproff]

Rev Animated, Reliberate and Realistic Vision

[mkaleborn]

I admit I've been sleeping on revAnimated for too long. It's been in my model library, but I haven't given it a proper try-out. Strangely, when I look at some of my most-often used personal Merges, many of my favorite pics (in illustrated, more realistic Anime style) have revAnimated in the mix. I guess that is not a coincidence!

2 of my favorites for illustrated, drawn, cgi (for SD1.5) are 'Simply Beautiful' and 'ProtogenX34Photorealism' (which you helped introduce me to!). Those 2 merged together produce very exciting and colourful drawn, painted or digitially rendered art. Combined with revAnimated I've produced some very good results with easy prompting.

And yes, for photorealistic images, the RealisticVisions series is hard to beat. I find it merges well with other realistic models, or illustrated/anime models for a nice variety of ethnicties and body types.

[AD-Dom]

You sir are a gem. Learned so much from you!! Thanks for the use of your prompts too, my alien fused people will once become normalized. My fave checkpoint is ArtUniverse v1.0-4.0.

[jcvijr]

For anime images, I’m using most Meina series, ToonYou and MistoonAnime. SDXL base model can generate good results with anime style too.

[mothified]

Very nice. I'd like to see more artifacts in your generations some time. Sometimes even prompts with more specific demands out of them. Ask for more, see what it makes

[TAKEASHORTBREAK]

Nice info, very usefull as always. My favorite models are RunDiffusionXL, lrmLiangyiusRealistic, Reliberate and icbinp.

[Pauluz_The_Web_Gnome]

Photon is the best for me so far! Very versatile! Why haven't you included Photon examples?

[codexer]

My latest models that I use most often are: IsoMix and RealCartoonRealistic

[pixelpop]

Actually the "1girl" is a descriptor, the number 1 designating the person most screen left on the image so that any prompts after apply to it before the next character ie: "2girl", this way you can specify eye/hair color etc for each character in the image.

[imatouristin]

Could you make a video of your current hardware setup? I think it would be useful for people who want to have an idea about how fast, how much, and how many gigabytes. Thanks

[StevicMacKay]

Ausgezeichnet! I've been watching your videos for a while and admittedly, have been avoiding Stable Diffusion as the UI reminds me of when I board a plane and catch a glimpse of the cockpit controls and realise they fly the giant metal bird with a 1970s modular synthesiser.

Now all the digital chickens have come home to crap on my car because I need very specific compositing and segmentation that only inpainting can provide (vs “the ol’ /imagine a wizard on a jetski trick” trick I’ve been Maxwell Smart’ing until now) in addition to this; unfortunately I’ve contracted a severe case of goldilocks complex exacerbated by AI art quality baselines creeping up like Tim & Eric’s Spagett bloke jumping out from a pot plant with weaponised parmesen. 

This comment in itself is procrastination from the jargon swamp and nebulous checkpoint naming conventions that I’m pretty sure drowned Artax…I have a deadline in 24 hours and I’m hoping enough coffee will create a Chronobowl allowing timeless iterations …I digress (again) my intent is to say, thanks for the time and effort you put into breaking down processes for noobs like me. 

P.S. I’m starting to hear, “Hello my friends!” when the TV is actually switched off. 👾

[tazztone]

is there a way to automatically load recommended prompts, VAE, embedding, etc each time i switch models? what would be the most convenient way to achieve this?