sd_hassan

HassansGuides - Getting started with WebUI On Google Colab

simple guide covering how to get up and running with the WebUI on Google Colab, an alternative option to Runpod and Local which also has a paid/free tier

Chavgirl Creator Hypernetwork 89k Steps

Trained on around 600 images of ChavGirls up to 89k steps (89k steps for this tier), this hypernetwork will convert a normal "standard" person to ChavGirl style.

Those unfamiliar with ChavGirls can read more here but they're known for dressing provocative, having a certain stereotype about them and mostly based in UK.

The prompt for the following was a simple: 2022-12-23 23:42:52 +0000 UTC View Post

ChavGirl Creator Hypernetwork 30k steps

Trained on around 600 images of ChavGirls up to 89k steps (30k steps for this tier), this hypernetwork will convert a normal "standard" person to ChavGirl style. 

Those unfamiliar with ChavGirls can read more here but they're known for dressing provocative, having a certain stereotype about them and mostly based in UK.

The prompt for the following was a simp...

2022-12-23 23:39:39 +0000 UTC View Post

Landscape LongExposure hypernetwork

This hypernetwork was trained on around 500 long exposure images

The effect it has on your image is:
- If it's a night landscape, it will try adding more stars and star trails
- If there's water it will try to add long exposure effect
- If there's a lot of elements like forests/cities/sky/water together, it will try blending the image to be a little more consistent like a long exposure shot.

It's not always recommended to use strength 1.0 and you may not actually get a b...

What content to work on next for you?

I appreciate all your support folks.

After the last poll where many of you said you wanted more tutorials/guides , I wrote a bunch of guides and made some videos teaching methods I follow for using and training etc. I hope you find these helpful

What are some things you'd like to learn or have me create and share moving forward?

Hassans Guides - Getting started using Automatic1111 WebUI on Runpod

simple couple of pages to help you get started on runpod with Automatics WebUI

C.A.M (Trinity) embedding - up to 10k steps

10k steps for this tier

Trained on SD1.5 up to 19k steps for other tiers

Sample prompt:

portrait photo of beautiful 25yo  carrie-xyz0-moss-19000 wearing sexy lingerie, perfect eyes, perfect face

C.A.M (Trinity) embedding - all steps

Trained on SD1.5 up to 19k steps

Sample prompt: 

portrait photo of beautiful 25yo  carrie-xyz0-moss-19000 wearing sexy lingerie, perfect eyes, perfect face

Hassans Guides - #6 Img2Img Inpainting guide - Basic

8 page guide to how I use inpainting to touch up an image

Hassans Guides - #5 - Txt2Img guide, my approach- Basic

7-8 pages covering my process from starting a prompt to completing a txt2img generation

Hassans Guides - #4 Understanding Hypernetworks/Embeddings - Basic

2 page quick overview on what they are and how we use them, more advanced info on training them in another guide

Hassans Guides - #3 Main scripts/extensions I use - Basic

quick 4-5 pages covering the extensions/scripts I use, there's way more out there that I have installed but not that I actively use as a day2day

Hassans Guides - #2 Webui My Customizations - Basic

Quick 4 page pdf going over the customizations I use

Hassans Guides - #1 Getting Started - Basic

Simple 10 page document I created for getting started using the WebUI

HassanBlend 1.5 progress

Hey all,

As you know from the previous update, the finetuning that I performed over 4-5 days last time wasn't efficient enough, mostly down to captioning. Since that last update what I've been doing is:

• Manually writing captions for batches of the dataset, there's almost 7k images in total so I'm doing batches of 3-500 at a time
• Finetuning each batch one at a time, so it's an iterative approach to finetuning the model instead of doing it all in one go
• Testi...
  2022-12-14 11:12:59 +0000 UTC View Post

J.A Embedding

example prompt:

photo of beautiful 25yo  jennifer-xyz0123-10000 in (sexy lingerie:1.2), perfect eyes, perfect skin

SD1.5 - up to 30,000 steps all included in the zip

T.H Embedding

13k steps

SD1.5

What type of content you want released next?

M.K Embedding - up to 10k steps (Boosted tier)

Trained on sd1.5 up to 30,000 steps, 10k steps available at this tier

Sample prompt:

high quality photo of milakun2022-10000 wearing lingerie, with beautiful real eyes ,beautiful smile

M.k embedding - all steps to 30k

Trained on sd1.5 up to 30,000 steps

Sample prompt: 

high quality photo of milakun2022-30000 wearing lingerie, with beautiful real eyes ,beautiful smile

M.F Embedding

She's a weird one to train because her eyes aren't as human as AI as used to lol

You may not get perfect results but here's my first run at an XY prompt using 3 of the embeddings, there are 12 embeddings included up to 35k steps


Example prompt: high quality photo of megx312fox-33000 wearing lingerie, with beautiful real blue eyes ,beautiful smile


Trained on SD1.5, should be fine to use on any model that works with SD1.5 datasets


2022-12-09 12:11:29 +0000 UTC View Post

Finetuning HassanBlend1.5 update

Quick update on the finetuning,   I've been testing it today after a few hundred thousand steps, and although I've had some great success in some area's I've had poor success in other areas.
I'm trying to keep this as a well balanced model that can be iterated on each version and also that others can finetune whatever way they want after too.  I don't think in it's current state it's balanced enough.

I want to re-edit all the captions...

HassanBlend model training suggestions

If you haven't seen already, you can make suggestions on what I add to the model next. 

The process will be : 

1 : Collecting examples from discord on what users want to see added to the model

2: Out of the list, pick the top 10 or so suggestions

3: Create a poll on Patreon for my subscribers to determine out of the list what to focus on first

Reference to discord: 2022-12-01 10:26:27 +0000 UTC View Post

Finetuning HassanBlend update

Most of you are on the discord I imagine but for those who aren't, the finetuning is going well. Training hassanBlend1.4 on thousands of NSFW variations including:

-Training dataset is based on male penis images, selfies etc with their -penis and closeups of penis 

- vagina closeups, various poses of women with vagina's spread 

- bdsm rope data etc 

- upskirt photos, low angle butt/ vagina shots 

- blowjob images 

- various interco...

Hassans Eye + Skin Enhancer

This is another version of the enhancer. The previous one focused on eyes and portraits and skin and hair, this one focuses mainly on eyes and the surrounding skin around the eyes and top of the nose etc. 

Really adds a nice effect to make it more realistic. 

18,500 steps

2022-11-29 23:48:32 +0000 UTC View Post

Face enhancer 60k steps

higher steps than the previous one, recommended to use around 0.5 strength but use this guide to determine the output you prefer, 1 being full effect

Try it on mostly shots with a face/eyes. Will upload new one when it's completed training.

Just download the .pt file and drop it in your hypernetwork folder

prompt:

a very beautifu...

2022-11-29 20:30:28 +0000 UTC View Post

New Eyes / Face / Portrait enhancer hypernetwork

This is still being trained so this is more of an early access release for Supershot patrons, don't consider it the "final" release but it may be fun for you to test and play with in the meantime while I train the rest of it.

It's around 45k steps so far. I don't recommend 1.0 full strength just yet until the full training is complete,. I think a strength of 0.4-0.6 is good.

Try it on mostly shots with a face/eyes. Will upload new one when it's completed training. ...