Why Is EVERYONE Using This Model?! - Rev Animated for Stable Diffusion / A1111

Рет қаралды 88,124

Жыл бұрын

Why is everyone using Rev Animated for Stable Diffusion? Here are my best Tricks for this Model. Get stunning Results in A1111 in no Time. Full Prompt Provided below. Here is the Prompt Engineering and my Workflow to get the MOST quality. We are also using negative Embeddings for this.
#### Links from the Video ####
LIVE Stream: kzfaq.infoOJW8psTylv4
Join my Discord: / discord
Join my Facebook: / theairevolution
Rev Animated Model: civitai.com/models/7371/rev-a...
VAE: huggingface.co/stabilityai/sd...
Negative Mebeddings:
huggingface.co/Xynon/models/t...
huggingface.co/datasets/Nerfg...
huggingface.co/yesyeahvh/bad-...
huggingface.co/datasets/gsdf/...
civitai.com/models/4629/deep-...
huggingface.co/NiXXerHATTER59...
Prompt and Settings:
((best quality)), ((masterpiece)), ((realistic)), (detailed), portrait, close up, young female fantasy mage, freckles, grey eyes, wearing blue robes, highly detailed skin, looking at the viewer, cyberpunk art, stunning gradient colors, no watermark signature, detailed background, town, alley, dark alley, closed mouth, insanely detailed, , ((masterpiece)), absurdres, HDR
Negative prompt: wires, ear rings, dirty face (deformed iris, deformed pupils, bad eyes, semi-realistic:1.4) (bad-image-v2-39000, bad_prompt_version2, bad-hands-5, EasyNegative, NG_DeepNegative_V1_4T, bad-artist-anime:0.7),(worst quality, low quality:1.3), (depth of field, blurry:1.2), (greyscale, monochrome:1.1), nose, cropped, lowres, text, jpeg artifacts, signature, watermark, username, blurry, artist name, trademark, watermark, title, (tan, muscular, loli, petite, child, infant, toddlers, chibi, sd character:1.1), multiple view, Reference sheet, long neck
Steps: 20, Sampler: Euler a, CFG scale: 6, Seed: 3458837599, Size: 512x768, Model hash: d725be5d18, Model: revAnimated_v11
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 Affinity Photo Creative Packs: gumroad.com/sarikasat
My Patreon Page: / sarikas
All my Social Media Accounts: linktr.ee/oliviotutorials

Пікірлер: 219

@Silent1Majority Жыл бұрын

"WANT!" video covering the different VAE's and how they can best be used.

@OlivioSarikas Жыл бұрын

🔥😍

@RobertJene Жыл бұрын

@@OlivioSarikas I also want this

@syxtiwatt3857 Жыл бұрын

@@OlivioSarikas Pleeeeeaaaase!!!!

@yusufharris1410 Жыл бұрын

Yes, I want this as well ❤️‍🔥

@5958637 Жыл бұрын

Please!

@s6yx Жыл бұрын

thank you again for featuring my model. I appreciate your added input as well on prompting

@CheckoFresco Жыл бұрын

Such an amazing tutorial, I've experimented now with this, and the results are epic, Thanks Olivio!

@xumatyt Жыл бұрын

0:56 Yeah, those six fingers look stunning 😁

@OlivioSarikas Жыл бұрын

It's a extra finger for bitch-slapping the Anti-AI crowd ;)

@KadayiPolokov Жыл бұрын

Olivio you can put both the clipskip and VAE dropdown on the UI front end, which saves having to go into settings all the time to load them (useful if you have a few VAE) Goto Settings>User Interface Then in the box just under 'Quicksettings list' post the following code: sd_model_checkpoint, CLIP_stop_at_last_layers, sd_vae Save and reload the UI It's a real time saver

@herval Жыл бұрын

epic tip!

@Cyan101 Жыл бұрын

can also just bake the vae into the model to save hassle of turning it off when swapping to a model with its own vae/dont want to use that one

@KadayiPolokov Жыл бұрын

@@Cyan101 True enough, but it's nice to have the option to move between them as well.

@techcryptoin Жыл бұрын

Thank you for leaving all the links, made things very easy.

@barry_2191 Жыл бұрын

Thanks a lot Olivio!, this channel is amazing!

@Voidgamer21 Жыл бұрын

Very Special Thanks to you Jorge, You are a real asset to this community, and a major contributor, especially now this medium is moving toward Monetization its great to still be able to garner this quality of information. All the best to you, and YES! A tutorial on VAE's would be awesome, especially how to make your own.

@Olstman Жыл бұрын

Yes please! Explaining VAE Video! Tutorial great as always!

@OlivioSarikas Жыл бұрын

🔥😍

@betodeth Жыл бұрын

Saw you used that model in previous video and downloaded it! Amazing my favorite model right now! ❤ I will install VAE thanks to you. I want a video about it, I like the way you explain.

@OlivioSarikas Жыл бұрын

😍Really love this model. Such awesome results and details

@Blood3533 Жыл бұрын

Thank you sir. Great guide.

@ShinichiKudoQatnip Жыл бұрын

Love you for such a detailed explanation ❤

@JuanDArango Жыл бұрын

Bro incredible work you are doing , thanks in advanced

@Maisonier Жыл бұрын

Yes please, Could be awesome more VAE videos. Liked and subscribed.

@Flackon Жыл бұрын

A video on VAEs would be nice, for sure

@roseydeep4896 Жыл бұрын

Yes please. I'd love to learn more about how VAEs affect the generated pictures and what are some of the best VAEs for different types of models and images!!

@OlivioSarikas Жыл бұрын

🔥😍

@toughnumber Жыл бұрын

Your tutorials are brilliant, I follow along exactly and it always works out. Can't wait to try this one!

@Elwaves2925 Жыл бұрын

I stumbled on this model about a week ago, when a prompt used it. Loved it from the first go. Thanks for the top about using a VAE, I didn't really know what they were and as others have said, I'd love a video all about them when you get chance.

@OlivioSarikas Жыл бұрын

Thank you very much :)

@3diva01 Жыл бұрын

Excellent video! Thank you! Did you ever make a video about different VAEs and what they are for and how to use them?

@yuryruban 6 ай бұрын

Thank you wizard!

@lavahawk Жыл бұрын

been my favorite model for a while

@gu9838 Жыл бұрын

i use it all the time lately its great.

@syno3608 Жыл бұрын

A simple but effective NOTE :- Use Clip skip of 2 - the effects are of magnitudes .

@DavidMiller212 Жыл бұрын

This model is wicked good!

@NoMoreCoolNames Жыл бұрын

Thank you, I really appreciate when you showing where to put each file

@OlivioSarikas Жыл бұрын

My pleasure :)

@OlivioSarikas Жыл бұрын

#### Links from the Video #### LIVE Stream: kzfaq.infoOJW8psTylv4 Join my Discord: discord.gg/XKAk7GUzAW Join my Facebook: facebook.com/groups/theairevolution Rev Animated Model: civitai.com/models/7371/rev-animated VAE: huggingface.co/stabilityai/sd-vae-ft-mse-original/tree/main Negative Mebeddings: huggingface.co/Xynon/models/tree/main/experimentals/TI huggingface.co/datasets/Nerfgun3/bad_prompt/blob/main/bad_prompt_version2.pt huggingface.co/yesyeahvh/bad-hands-5 huggingface.co/datasets/gsdf/EasyNegative civitai.com/models/4629/deep-negative-v1x huggingface.co/NiXXerHATTER59/bad-artist Prompt and Settings: ((best quality)), ((masterpiece)), ((realistic)), (detailed), portrait, close up, young female fantasy mage, freckles, grey eyes, wearing blue robes, highly detailed skin, looking at the viewer, cyberpunk art, stunning gradient colors, no watermark signature, detailed background, town, alley, dark alley, closed mouth, insanely detailed, , ((masterpiece)), absurdres, HDR Negative prompt: wires, ear rings, dirty face (deformed iris, deformed pupils, bad eyes, semi-realistic:1.4) (bad-image-v2-39000, bad_prompt_version2, bad-hands-5, EasyNegative, NG_DeepNegative_V1_4T, bad-artist-anime:0.7),(worst quality, low quality:1.3), (depth of field, blurry:1.2), (greyscale, monochrome:1.1), nose, cropped, lowres, text, jpeg artifacts, signature, watermark, username, blurry, artist name, trademark, watermark, title, (tan, muscular, loli, petite, child, infant, toddlers, chibi, sd character:1.1), multiple view, Reference sheet, long neck Steps: 20, Sampler: Euler a, CFG scale: 6, Seed: 3458837599, Size: 512x768, Model hash: d725be5d18, Model: revAnimated_v11

@PWNEDPIXELS Жыл бұрын

very helpful, thank you so much

@OlivioSarikas Жыл бұрын

You're welcome!

@c0nsumption Жыл бұрын

Videos on VAEs for sure!!!

@OlivioSarikas Жыл бұрын

🔥😍😍🔥

@hellomoney333 Жыл бұрын

Thank you Olivio, for your knowledge on how to create illusions with magic spells

@aaronhhill Жыл бұрын

I've been using it for a while now. That and RPG and Deliberate are my starting models. Then I refine in Realistic or Faetastic depending on what type of style I'm looking for.

@OlivioSarikas Жыл бұрын

I'm a big fan of Deliberate too :)

@Shingo_AI_Art Жыл бұрын

I have to admit this is a wonderful model, i just made a random try with a scribble yesterday and the results are impressive, but you really notice it when you use Hires upscale with Latent upscaler

@OlivioSarikas Жыл бұрын

Yes, totally. Latent upscale sets this model on fire 🔥😍

@DennisHellberg Жыл бұрын

Quality content like always. I would like to see a video about vae's =) Cheers mate

@OlivioSarikas Жыл бұрын

🔥😍 Thank you

@IgorNV Жыл бұрын

I definitely want a video on VAEs!

@irangel1958 Жыл бұрын

Excellent informationthank you. I would vote yes for a video on VAE's.

@OlivioSarikas Жыл бұрын

thank you :)

@Mocorn Жыл бұрын

After seeing your video on upscaling I was reminded about the 4xsupersampler and started tinkering around with enabling it for the native hiresfix function and got some seriously nice results. My workflow was very similar to what you do in this video. Once I had found a good seed and a nice positive prompt I would then use the high-res function to upscale before sending the picture to img2img. The trick here being that the high-res function upscales through the latent space rather than from pixels. This means that new details can be created from latent scratch rather than trying to enhance existing pixels. The trick to get a good result for me was to use the high-res fix with the 4X super sampler enabled at 1.3 (instead of 2) and a low denoise value so that it didn't get to creative. This way I ended up with pretty high resolution images that was very similar to my first image. This became an excellent starting point for the img2img tab. Is this something that you've experimented with yourself?

@lollihonk Жыл бұрын

How much vram does it need?

@Mocorn Жыл бұрын

@@lollihonk I only have 11GB so I have to work around that a bit. I can render with this technique to get to about 1300px on the original. That's a decent point to upscale from

@dexter0010 Жыл бұрын

quick tip go into settings, user interface and add sd_vae where it says sd checkpoint or something like that, it'll add a dropdown menu next to the checkpoint one in the webui for the vae!

@RamkisanMohan 11 ай бұрын

Olivio! first of all your videos are great. You explain in a way where even a layman can understand. Is there a way we can add multiple characters to a single prompt and define params for each character and interactions

@BlakeEM Жыл бұрын

I tend to use epiNoiseoffset_v2 LoRa instead. In your example, I think the vanilla one looks better. Warmer colors, more detailed skin, better eyelashes, more natural lighting on the skin.

@BillMeeksLA Жыл бұрын

Absolutely want a video on VAEs.

Жыл бұрын

1:05 you prise the "nice" hand... with 6 fingers :P

@OlivioSarikas Жыл бұрын

Can't have to many fingers in times of prompting :)

@sorijin Жыл бұрын

A bunch of my deforum videos use this model many props to the creator!

@OlivioSarikas Жыл бұрын

cool. I didn't try it for that yet 🔥😍

@s6yx Жыл бұрын

thank you, ill check your videos on using my model

@hairy7653 Жыл бұрын

Yes a VAE vid would be great please

@OlivioSarikas Жыл бұрын

🔥😍

@Heisenburg17 Жыл бұрын

Looks like famous Sakimichan art! Almost perfect!

@OlivioSarikas Жыл бұрын

Uhhhh, saucy. Never heard of Sakimichan

@dschroeder7354 Жыл бұрын

You are awesome!

@OlivioSarikas Жыл бұрын

Thank you 🔥😍

@bellcurve0 Жыл бұрын

Olivio wondering if u could make a vid that breakdowns the differences is prompt creation between MJ and SD Like are the approaches similar of completely different. I see SD prompts with parentheses and way more. Reliant on negative prompting? Idk, on MJ and wanting to explore SD but SD seems really intimidating to learn.

@XViewer Жыл бұрын

Cool, Which site or tool are you using?

@zephilde Жыл бұрын

Hi! I really like your videos because you dive deep into config and settings... But, it would be really helpfull for understanding if you could put some quick incrust of an image sshowing the diff in settings you speak about! ( i know it may represent some work, but it will help newcomers! ) Cheers

@twitterglobalarmy Жыл бұрын

Models are cool and all that, but getting TOOLS to make cool things, THIS is the kind of stuff I like to come to your channel for. You appear to be quite knowledgeable on the state of AI. I recently learned about BlueWillow, but I need a tutorial with more information

@valter987 Жыл бұрын

This is the best model i ever used, the level of details are amazing, so it was an amazing tip from this channel

@Daydricoldsmith Жыл бұрын

Another video idea you could do would be showing everyone how to make their own safetensor through merging two others together.

@krio_gen Жыл бұрын

This is one of the best models, one with a huge disadvantage -the bodies look plastic if you open the skin.

@Vokoca Жыл бұрын

Great video! However, that negative prompt seems like such overkill, especially with all those weights. I've been experimenting with just a single negative embedding, or even just stuff like "bad prompt, worst quality" on its own and it seems to work similarly enough on a lot of models. I didn't try this model specifically yet, but it might be something that's worth looking into - it feels like there is a lot of mysticism around negative prmpts ending up with a lot of bloat and confusion.

@adamquinnart Жыл бұрын

It is a really great model currently my favorite 🙂

@OlivioSarikas Жыл бұрын

Mine too

@caiubymenezesdacosta5711 Жыл бұрын

Tanks for sharing. Please , if its possible, make a video explaining the types of VAE (Variational Autoencoder)!

@OlivioSarikas Жыл бұрын

🔥😍

@jeriontop Жыл бұрын

Videos on the best VAEs please 🙏

@designproductwithbing Жыл бұрын

hey Olivo, could you make video about the architechture design or product design using stable diffusion? It would be much much help of understanding the capability of AI in design industry.

@komakaze1 Жыл бұрын

Yes, how to describe objects that don't exist? I once started with a photo of an iron and turned it into a spaceship. It turned out nothing like I intended, perhaps if I Photoshopped the iron a bit more like a bad collage and iterated on that I could get better results. It requires a whole different library of prompt words to create objects instead of people or landscapes.

@AustinGlamourPhoto Жыл бұрын

Start playing with DDIM sampler. It's super fast and you will get very nice backgrounds.

@AndyHTu Жыл бұрын

Can you make videos on creating beautiful cities? I've been trying to build a lot of dystopian cities, sin city, skyscrapers etc, but I get the ugliest lines. I've been also trying to composite images into my photos, but I kept getting weird results. Specifically trying to composite a giant person over a neon lit city, like he's watching over the city. I will love you forever if you can help me with that! :)

@frazuppi4897 Жыл бұрын

it would be very cool if you can share the prompts as well in the video description

@EditArtDesign Жыл бұрын

Where should these negative models be located and how to use them is not very clear?? Thank you in advance!

@talessin Жыл бұрын

Hi! Why safetensors are better then ckpt? and yeah VAE differences video will be cool. one mor thing is where what folder we should put Negative Mebeddings

@MHumanoid Жыл бұрын

Are these negative embeddings useful for realistic images as well?

@TheSchwarzKater Жыл бұрын

wow, so leaving a comma with space is like some hack. Shows how early this tech actually is, kind of funny. Reminds me of old code when there was something like "leave a space here for the interpreter to make a pause and free RAM" lol leave space, now dance three times counterclockwise and clap your hands.

@bhaskarcholasamudram1128 Жыл бұрын

Olivio would this model also work with Mochi Diffusion on mac?

@erikschiegg68 Жыл бұрын

Bring a VAE explaining video, please.

@OlivioSarikas Жыл бұрын

🔥😍

@TheRoadTraveledByFew Жыл бұрын

I've been using InvokeAI and when I try to use the EasyNegative embedding it tells me that it's unusable in the command line because it was trained on a 2.0, 768x768 model (and yes I have it in the correct folder, all my other embeddings are working). Considering this is a model trained on 1.5, I would assume that people are not noticing that this embedding is not being applied.. or is this only an Invoke thing? Anyone else notice this? Thx to anyone with insight on this. Also yes a video on VAEs would be great! Love the channel.

@thewebheadgt 3 күн бұрын

This vs Dreamshaper 8? What do you recommend?

@clearandsweet Жыл бұрын

I haven't messed around with any vae because I didn't know what they were used for I would love a video describing them

@OlivioSarikas Жыл бұрын

Me neither to be honest. I learned a lot about them from ComfyUI

@shambhangal438 Жыл бұрын

Variational Autoencoders (vaes) give you better color, contrast and better low level details (eyes, hair, cloth, etc). You will already know that the main model (e.g. stable diffusion 1.5) is trained on the big subjects in a scene, think of a vae as being a smaller model trained on optimizing the lower level details. In terms of what they actually do, AI image generation computation works within 'latent space'; a vectorized version of the problem. You have to compress and decompress the data going in (or to be more specific, chose 'an optimal subset of properties that describe the problem' - this reduction in information is why it used to take high memory GPUs to render stable diffusion images, but now consumer GPUs can do it!) and coming back out of the latent space, and this is usually done via a fixed translation. A vae is a neural net that has been trained to optimize this translation, and this comes out as better low level information in the final output; better detail in small areas of the scene and better color/contrast. The standard block diagram image you see is of an hourglass on its side like this; >o< The > is the compress part (encoder), the < is the decompress part (decoder) and the o is latent space. The vae optimises the > and

@imanonymus5745 Жыл бұрын

yes we want vae video

@OlivioSarikas Жыл бұрын

🔥😍

@simonbe Жыл бұрын

Stunning six finger hand 😂

@BlackJade_OFM Жыл бұрын

So, How once i get the image i want, i get different shots, say Closeup, cowboy, full body without the AI changing some part of the image? I have saved and reused the seed. no settings have changed

@duskfallmusic Жыл бұрын

THAT MODEL IS AMAZEBALLS when I test my lycoris/lora, that and Kenshi. I also use Nyan mix a lot.

@duskfallmusic Жыл бұрын

Wait Sample prompts are refined? Wow, LMAO nobody go looking up duskfallcrew models adn expecting refined mature amazing prompts XD i'm usually cranky and just shove whatever in XD

@OlivioSarikas Жыл бұрын

Awesome!

@MacS7n Жыл бұрын

Please make a video about VAE. Many people don't understand it including myself.

@OlivioSarikas Жыл бұрын

🔥😍

@Mirko_ai Жыл бұрын

Where to put a "LoRA" in the positive Prompt? Is the placement important?

@RB-ns2nn Жыл бұрын

I've worked with this model using orangemix and I honestly couldn't tell the difference between vaes. I would love to get more information on vaes.

@OlivioSarikas Жыл бұрын

🔥😍

@ArmoredAnubis Жыл бұрын

how to get the dark user interface? mine is white. thanks

@krisstone2562 Жыл бұрын

can you make a tutorial on how to revert the new UI for automatic1111, It is not user friendly at all and messy

@Yosh1d44 Жыл бұрын

I'm sorry, I have a question, what is the difference of VEA between "MSE" and "EMA", because anyone cant say me what is the difference is or in which cases it is better to use one or the other

@II-nb6ko Жыл бұрын

The order seems to also affect the quality of the image and whether or not the keyword is actually used or not. Somtimes I've struggled to get a certain feature to appear (a pendant for example) and it wouldn't ever show up until I moved the keyword before or after something. It takes a lot of trial and error to get it to generate a pendant consistently. Is there a doc that explains what needs to go be for what for best results or is it totally arbitrary?

@OlivioSarikas Жыл бұрын

yes, good point. I should have mentioned that

@CASS1__ 10 ай бұрын

it's a really great AI model, but I almost never get one character in a generated image. Always two, or more, any tips? I have (1girl, sole_female, solo_focus) in positive prompts and (2girls, multiple_girls) in negative promps, doesn't do anything at all

@rafaelbelethian3176 Жыл бұрын

Where do I put the easynegative file?

@seifergunblade9857 Жыл бұрын

please make a video about vae. can i run without vae?

@mickeytjr3067 Жыл бұрын

bad-artist-anime, is not available any more the link goes no a hugging face page that is 404. Is there another alternative to this or can you put it up for download?

@OnlineTabletopcom Жыл бұрын

It is very pretty, but it doesn't work well with my Lora. When I use my Lora, the images look like splatters/rough oil painting

@Philson Жыл бұрын

VIDEO ON VAEs PLEASE!

@marksutherland774 Жыл бұрын

great video, is it possible to use this with a custom model e.g a model of myself and add this, if so how, thank you

@OlivioSarikas Жыл бұрын

i haven't tried it, but loras should work. try the lora with a weight of 0.8 and lower the weights in the negative prompt maybe

@marksutherland774 Жыл бұрын

@@OlivioSarikas Thanks, just getting started with all this but have stable diffusion installed locally and I see a tab for lora in extensions, thanks again for your reply.

@uwubeats Жыл бұрын

What do you think about the new Automatic1111 UI Update?

@OlivioSarikas Жыл бұрын

Didn't notice many changes yet, but i will look into it

@footballfanatic170 Жыл бұрын

Its showing Remacri Upscaler missing for me. pls help

@767corp Жыл бұрын

Maybe I'm wrong but even tho it has urban meaning "absurdres" I doubt it does anything in SD as it's assigned 2 tokens 1 for absur and other for dres which on their own shouldn't do anything but take up your token space. There's a huge list somewhere on git, that contains all SD embeddings so you can cross check certain terms but I think with tokenizer is easiest. One word should be one token as I understand , if it splits then prompt is using both meaning individually. Maybe I'm wrong tho , still not super clear on all priorities. But I've read somewhere for example if you use word photorealistic for SD won't actually mean nothing like that because it's assign 2 tokens so photo will be one prompt and realistic 2nd one even tho to us is one word.

@OlivioSarikas Жыл бұрын

hm... there are a lot of myths around prompts and what works. Some say 4k does nothing or naming a camera doesn't work. Then again you see things used in prompts with stunning results and when it's not used it doesn't look as good. If it works, use it. No matter if it makes sense or not ;)

@OmantisO Жыл бұрын

More VAE's more info 😆

@MrFreak-ev6pc Жыл бұрын

Skin pores also work great for skin

@OlivioSarikas Жыл бұрын

Thank you!

@suyogsavle4719 Жыл бұрын

Hi Olivio, can you help me with this please?... I have rtx 3060 6gb laptop...It is able to generate a batch of 4 images in 15-20 seconds at 512x512. I am ready to wait 5 or 10 minutes too, but want to generate images at 1024x1024. I get the error Out of memory CUDA error. 768x768 is the max resolution at which I don't get this error sometimes.

@terraincognitagaming Жыл бұрын

There are 5 fingers and a thumb on the hand example.

@OlivioSarikas Жыл бұрын

but it still looks better that most hands we have seen for a long time

@lpnielsen1001 Жыл бұрын

Does the empty prompt between 'Insanely detailed' and ((masterpiece)) mean anything, or is it just some left over noise?

@OlivioSarikas Жыл бұрын

sometimes it helps

@lpnielsen1001 Жыл бұрын

@@OlivioSarikas regarding your ultra sharp upscale video, i've found that the NMKD Super Scale, works even better than the one you showed in the video, if you want to give it a try. At least for realism pictures. Exact name: 8X_NMKD-Superscale_150000_G

@gionasser Жыл бұрын

Can control net be utilized here?

@OlivioSarikas Жыл бұрын

Yes, it's a 1.5 trained model

@EmilioNorrmann Жыл бұрын

where do I have to save those embedding files to work ?

@OlivioSarikas Жыл бұрын

Into the folder called "embeddings" in your A1111 folder

@sandexhshrestha2562 Жыл бұрын

Is there any way we can train character and get custain charater and outfit in result ?

@OlivioSarikas Жыл бұрын

try Lora training

@toututu2993 Жыл бұрын

The huge problem about these portrait is that the expression looks dead like "I wanna be gone already" expression

@USS_Daedalus Жыл бұрын

Which Model to use for Landscape? I would like Design a Beautiful Beach⛱️😊

@OlivioSarikas Жыл бұрын

I want to find a good one as well

@evelynannrose Жыл бұрын

I'd like a vid on vae

@moon47usaco Жыл бұрын

More info on VAE. What it is in more detail, how best to use them? Should I use them during dreambooth/koya training? Etc... =] Thank you for all the excellent info Olivio.