[Stable Diffusion] Prompt Sharing and Learning Thread

[fr34ky]

Here's my generated image of the one you posted from

You must be registered to see the links

View attachment 2452366
Very small things can make a difference. I don't have the same version as he. Like I already said he has the old version with the model hash:d8722b4a4d

Your version looks almost exactly like the one posted by Nano999.

Looks like a different seed to be honest.

 

[Mr-Fox]

Your version looks almost exactly like the one posted by Nano999.

Looks like a different seed to be honest.

I have used everything exactly like the op at civitai but because I didn't bother downloading the old version of ned it is not an exact replica. He is using the older version you can find a bit lower on the page with the hash: d8722b4a4d
Now I got curious enough that I'm downloading it to see how close I can get. I'm using xformer and we have no way of knowing if he is running something like this or not, nor if he uses the cuda memory fix by setting split file size.

 

[Nano999]

Yes it is possible but very difficult. You can set clip skip in settings Stable Diffusion at the bottom.

Prompt:beautiful, masterpiece, best quality, hiqcgbody, anime, 1girl, sam yang, halter cut dress, mascara, flirty, blunt bangs, portrait shot, looking at viewer, glossy lipstick, nightclub, intricate details,

Negative prompt: (painting by bad-artist:0.9), easynegative, (worst quality, low quality:1.4), watermark, text, error, blurry, jpeg artifacts, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, artist name, bad anatomy, deformed iris.

You need:

You must be registered to see the links

,

You must be registered to see the links

,

You must be registered to see the links

both lora's with weight: 0.5


and these settings:

Size: 448x640 Seed: 3686603218, Steps: 28, Sampler: DPM++ SDE Karras, CFG scale: 7, Clip skip: 2,
Hires steps: 20, Hires upscale: 1.75, Hires upscaler: Latent (bicubic antialiased), Denoising strength: 0.55

Well I have all of these settings and loras, you saw my result, similar, but sadly not the same.
Could you try to reproduce it? I'm curious of your result.

He is using the old version of NED fyi. You can see that it's the same model hash.

But shouldn't model hash change if the version is different? His model d8722b4a4d and mine d8722b4a4d as well

You must be registered to see the links

download: "NMKD Superscale"
And unpack it inside Stable-Diffusion\stable-diffusion-webui\models\ESRGAN
You need to restart SD for it to show up in the upscaler menu.
Though you would get pretty similar result with SwinIR_4x tbh.

You will also need

You must be registered to see the links

Place it inside Stable-Diffusion\stable-diffusion-webui\models\hypernetworks
go to settings and hypernetworks and select mjv4-hypernetwork and set the weight to 0.3 and press apply.

I'm running my GPU as hard as it will allow me so I hope you have something equivalent or more powerful.

You must be registered to see the links

was already there, I guess upscaler makes these minor differences in result images
But not sure if I set it right, like from here? Should it be in promt field?

Because I don't have hypernetworks in settings:

PS well my GPU is always ag 100% when generating xd

 

[Mr-Fox]

Well I have all of these settings and loras, you saw my result, similar, but sadly not the same.
Could you try to reproduce it? I'm curious of your result.


But shouldn't model hash change if the version is different? His model d8722b4a4d and mine d8722b4a4d as well

You must be registered to see the links

was already there, I guess upscaler makes these minor differences in result images
But not sure if I set it right, like from here? Should it be in promt field?

Because I don't have hypernetworks in settings:

PS well my GPU is always ag 100% when generating xd

Doh! My bad. It's in settings /Extra Networks. I did post my result for that image but I had not bothered getting the old version I'm downloading it now and will post another and we'll see if it makes any difference.

 

[Nano999]

Doh! My bad. It's in settings /Extra Networks. I did post my result for that image but I had not bothered getting the old version I'm downloading it now and will post another and we'll see if it makes any difference.

I am confused, do you mean here? But I can't change weight in here, it generated an image with this setting, but I see no weight control

Although I can type <hypernet:mjv4Hypernetwork_v1:1> in promt without going into setting and change weight there

Look my hash is the same as that guy, but result is still different
My hash: d8722b4a4d

 

[Mr-Fox]

I am confused, do you mean here? But I can't change weight in here, it generated an image with this setting, but I see no weight control

Although I can type <hypernet:mjv4Hypernetwork_v1:1> in promt without going into setting and change weight there

Look my hash is the same as that guy, but result is still different
My hash: d8722b4a4d

View attachment 2452449

There has been a miscommunication. The hypernetwork is for my "waifu" with multiplier 0.3 I was sloppy and called it weight sorry. To replicate the image from civitai you need to turn it off. Then go to settings / Stable Diffusion and change SD VAE to automatic.
I got the head tilt this way. It is still no way near his, wich tells me that I have missed something or he has hid something from us. Possibly both...

 

[Nano999]

There has been a miscommunication. The hypernetwork is for my "waifu" with multiplier 0.3 I was sloppy and called it weight sorry. To replicate the image from civitai you need to turn it off. Then go to settings / Stable Diffusion and change SD VAE to automatic.
I got the head tilt this way. It is still no way near his, wich tells me that I have missed something or he has hid something from us. Possibly both...
View attachment 2452482

no worries, I knew it was about your waifu, I was tryting to figure out multiple things at the same time

btw I had both SD VAE set to automatic and no hypernetworks applied when generating that image
So... the question is still open xD

I think it's something in settings need to change, or some extension I don't have and he has, something which are not shown in the pic info.

 

[Sepheyer]

The zoo called, they want their cougar back.

 

[Mr-Fox]

I had both Additional Networks enabled and <lora:hipoly3DModelLora_v10:0.5> <lora:samdoesartsSamYang_normal:0.5>
in the prompt with VAE-ft-mse840000-ema-pruned enabled. It's still not like civitai op but I doubt we will get a perfect replica.

 

[Nano999]

I had both Additional Networks enabled and <lora:hipoly3DModelLora_v10:0.5> <lora:samdoesartsSamYang_normal:0.5>
in the prompt with VAE-ft-mse840000-ema-pruned enabled. It's still not like civitai op but I doubt we will get a perfect replica.

View attachment 2452584

Should this .safetensors VAE be OK?

You must be registered to see the links

I figured out I don't have it

 

[Mr-Fox]

Should this .safetensors VAE be OK?

You must be registered to see the links

I figured out I don't have it

yes, it's the same as I have.

 

[Nano999]

yes, it's the same as I have.

Have you got it there? My dwnld speed is like 0,01 kb/s

 

[Mr-Fox]

Have you got it there? My dwnld speed is like 0,01 kb/s

You must be registered to see the links

Unpack inside Stable-Diffusion\stable-diffusion-webui\models\VAE
Go to settings/Stable Diffusion and press reload SD VAE at the far right. Chose the new VAE and press apply settings.

 

[Nano999]

You must be registered to see the links

Unpack inside Stable-Diffusion\stable-diffusion-webui\models\VAE
Go to settings/Stable Diffusion and press reload SD VAE at the far right. Chose the new VAE and press apply settings.
View attachment 2452639

Thank you!

Btw I'm trying to figure out how to do XYZ stuff, like I need to set it for each image to uspcale by 1,5 then 1,75 and then 2
Any tips?

 

[Mr-Fox]

Thank you!

Btw I'm trying to figure out how to do XYZ stuff, like I need to set it for each image to uspcale by 1,5 then 1,75 and then 2
Any tips?

It depends what you are trying to do. But lets say you want to compare the amount of steps.
enable xyz plot script, in xtype chose steps and in xvalues write the steps you want to try: 20,24,26,28 etc and then press generate. If you want a specific seed, paste this in the seed box first otherwise SD will chose a random seed that will be used for all images. In other words it will use the same random seed for all images and only change the amount of steps.
You can use xyz plot script to compare many different things. You can try out different samplers, upscalers even lora weight or checkpoints.
SD will generate an image grid that shows the different steps, you can find this in Stable-Diffusion\stable-diffusion-webui\outputs\txt2img-grids
and each individual image Stable-Diffusion\stable-diffusion-webui\outputs\txt2img-images like normal.

Example:

 

[Nano999]

It depends what you are trying to do. But lets say you want to compare the amount of steps.
enable xyz plot script, in xtype chose steps and in xvalues write the steps you want to try: 20,24,26,28 etc and then press generate. If you want a specific seed, paste this in the seed box first otherwise SD will chose a random seed that will be used for all images. In other words it will use the same random seed for all images and only change the amount of steps.
You can use xyz plot script to compare many different things. You can try out different samplers, upscalers even lora weight or checkpoints.
SD will generate an image grid that shows the different steps, you can find this in Stable-Diffusion\stable-diffusion-webui\outputs\txt2img-grids
and each individual image Stable-Diffusion\stable-diffusion-webui\outputs\txt2img-images like normal.

Example:
View attachment 2452704

Well I need to control this parameter:

It seems I can't, because I don't see it in the list.

 

[Mr-Fox]

Well I need to control this parameter:

It seems I can't, because I don't see it in the list.

No you're correct. You can't use upscale multiplier in xyz plot script it seems. But this is something easy to do manually anyway, only time consuming.

 

[Nano999]

No you're correct. You can't use upscale multiplier in xyz plot script it seems. But this is something easy to do manually anyway, only time consuming.

Hm, but can you batch scale images?

e.g. I have 100 images and want to set an order for rescale, but including steps and other stuff from the 1st tab (usual upscale) and not the simple one, saving the original image.

 

[Mr-Fox]

Hm, but can you batch scale images?

e.g. I have 100 images and want to set an order for rescale, but including steps and other stuff from the 1st tab (usual upscale) and not the simple one, saving the original image.

I have never tried to do that kind of batch work. I'm not even sure it can be done with SD. The easiest way would probably be to do it in a software like photoshop. I will have to think about this one. I'll get back to you if I figure something out.

 

[Nano999]

Well you can do it in "Extras" tab, there is a batch functionality.
But it will just upscale the image, while if you upscale on the main tab you might get a completely different results, comparing with non-upscaled image