HOME | DD

Description
I've had a load of people message me over the past few weeks about how I'm creating my images so I thought I'd make another one of these as I have no problem sharing that information with you all. I'd love it if it meant more people could participate in creating content.

Next: Tips for AI Image Generation! | 2 of 2

Also, note that any images I create for captions are uploaded on my alt account and are complete free for any other caption creator to use: www.deviantart.com/johnlavrend…

I also see the irony of making a post about creating my own images and then not doing so for this post 😅
Original Artist: BeriPeri | original, original character, white hair / OC 3 - teacher compilation 1 - pixiv

---

Johnlav’s Tips for AI Image Generation.

What software do you use?
I use a software called Automatic1111 to make my images, it’s quite complex but it’s free to use and the only real requirement is to have a fairly powerful computer with an Nvidia graphics card.

What do I need apart from A1111?
You can use a website called Civitai to find various models to use. The main thing you will need is a Checkpoint model.

This Checkpoint model is used to define the overall style of the image. There are a variety of options available depending on your preferences, some more anime-inspired and some more photo-realistic and many others in between. I’d recommend selecting one of the more top-rated models that you think looks good.

ReV Animated or Realistic Vision are probably good options to start with but feel free to experiment. Each model has a handful of settings for optimal use so make sure to read the description of the model and match the settings in A1111, it may seem overwhelming but you don’t necessarily need to know what these do.

Settings to take note of
Width and height should be set to 512x512 for a square image, 768x512 for landscape or 512x768 for portrait.

CFG Scale is how closely it will follow the prompt, a higher number will be more accurate but will increase the likelihood of hallucinations. A value between 4-7 is often recommended.

Sampling steps add detail so the higher the better but also, the longer it will take to render and at a certain point there will be no benefit so it’s best to just follow the settings suggested by your model’s description.

As for the Seed, most of the time you can leave this on random (-1) but if you create an image that is close to what you want, you can use its seed to try to recreate something very similar to it.

Description
I've had a load of people message me over the past few weeks about how I'm creating my images so I thought I'd make another one of these as I have no problem sharing that information with you all. I'd love it if it meant more people could participate in creating content.

Next: Tips for AI Image Generation! | 2 of 2

Also, note that any images I create for captions are uploaded on my alt account and are complete free for any other caption creator to use: www.deviantart.com/johnlavrend…

I also see the irony of making a post about creating my own images and then not doing so for this post 😅
Original Artist: BeriPeri | original, original character, white hair / OC 3 - teacher compilation 1 - pixiv

---

Johnlav’s Tips for AI Image Generation.

What software do you use?
I use a software called Automatic1111 to make my images, it’s quite complex but it’s free to use and the only real requirement is to have a fairly powerful computer with an Nvidia graphics card.

What do I need apart from A1111?
You can use a website called Civitai to find various models to use. The main thing you will need is a Checkpoint model.

This Checkpoint model is used to define the overall style of the image. There are a variety of options available depending on your preferences, some more anime-inspired and some more photo-realistic and many others in between. I’d recommend selecting one of the more top-rated models that you think looks good.

ReV Animated or Realistic Vision are probably good options to start with but feel free to experiment. Each model has a handful of settings for optimal use so make sure to read the description of the model and match the settings in A1111, it may seem overwhelming but you don’t necessarily need to know what these do.

Settings to take note of
Width and height should be set to 512x512 for a square image, 768x512 for landscape or 512x768 for portrait.

CFG Scale is how closely it will follow the prompt, a higher number will be more accurate but will increase the likelihood of hallucinations. A value between 4-7 is often recommended.

Sampling steps add detail so the higher the better but also, the longer it will take to render and at a certain point there will be no benefit so it’s best to just follow the settings suggested by your model’s description.

As for the Seed, most of the time you can leave this on random (-1) but if you create an image that is close to what you want, you can use its seed to try to recreate something very similar to it.