Hi all,
HassanSDXL LORA's are in Huggingface ready for access requests.
The 13gb checkpoint also uploaded.
NSFW Model Release: Starting base model to improve Accuracy on Female Anatomy
I am excited to announce the release of our SDXL NSFW model! This release has been specifically trained for improved and more accurate representations of female anatomy. As always, our dedication lies in bringing high-quality and state-of-the-art models to our users so this model is fa...
2023-10-05 15:02:53 +0000 UTC
View Post
Howdy all,
Since the last update I've been toying around with SDXL NSFW training.
Here's what I've found so far and there's some examples of the results below:
LEARNINGS
1: SDXL training learns NSFW concepts much better than SD2.1. It takes less images and less time for SDXL to learn the concepts which means the base model already has a good understanding of most of the anatomy with the exception of specific granular details
2: The resources need...
2023-09-14 10:55:34 +0000 UTC
View Post
similar to the tiktok one,
downloads X amount of photos from X amount of instagram profiles
Extracts faces from the downloaded photos, resizes to minimum 768x768
Saves the extracted faces to disk separated by each person detected
2023-05-03 22:23:51 +0000 UTC
View Post
This is helpful to pull faces for training out of multiple tiktok videos automatically
2023-05-03 20:06:12 +0000 UTC
View Post
Hey all,
It's been a while since we've had a progress update on SD2.1
What's been going on?
I've been training various concepts (12 different types of datasets) over the last couple of months. SD2.1 has proven quite difficult to train and I've been experimenting across many different training configs to try find the sweet spot of training parameters and convergence of my datasets.
How was the training results?
Some concepts...
2023-04-22 13:04:37 +0000 UTC
View Post
Hi all,
I want to provide an update on the 2.1 nsfw model.
I've had it in testing locally the last few weeks, testing all the concepts I trained and further training some concepts that are a struggle. The model is good at certain concepts but still there are flaws with things like male anatomy, penetration, dildos and some female anatomy isn't coming out perfectly.
I could release it in its current state but I don't think it's good enough to release and will cause more ...
2023-04-10 07:13:07 +0000 UTC
View Post
Hassan Fantasy 1.5
Fantasy.ai is the official and exclusive hosted AI generation platform that holds a commercial use license for HassanBlend you can use their service at https://fantasy.ai/
Please report any unauthor...
2023-03-15 19:50:40 +0000 UTC
View Post
While I am still testing SD2.1 NSFW (It's really close!)
I welcome you patrons to an early access Patreon only phase for a new model I am releasing.
Hassan Fantasy 1.5
This model is SD1.5 finetuned on thousands of fantasy style art and realistic illustrations at 768px resolution combined with my NSFW content thrown in to add some spice. It was called Hassan Fantasy 1 until I added more NSFW content into it to balance it out to 1.5
It's not a pure NSFW mod...
2023-03-13 20:17:14 +0000 UTC
View Post
I'm running a bot in my discord group, patrons are welcome to use it to generate images in Stable Diffusion directly from Discord. Whether it's going to be kept or not I'm not sure yet depending on the cost as right now I'm renting a runpod server to allow patrons to have the functionality.
Only patrons can use it currently.
Just go to the #sdbot
2023-03-12 18:15:49 +0000 UTC
View Post
My 2.1 NSFW model is a lot closer now. I've had good success over the last week specifically in training. I have it locally in my own test runs in batches and xy grids, also running some side training on runpod one or two specific concepts with new tailored datasets to polish up what I'm finding in the test runs.
There are still some concepts that are difficult but I'm considering releasing and allowing these to be "iterations" we improve on with patches etc.
My batch tests will ru...
2023-03-11 08:55:51 +0000 UTC
View Post
Hi all,
While I'm continuing to finetune SD2.1, The folks over at Waifu Diffusion have released an SD2.1 model.
This model is in beta but it has photorealism NSFW
This is an amazing upgrade to what existed today. I will continue finetuning and will incorporate this into my model too but in the meantime I thought you all can enjoy this
h...
2023-03-08 23:28:39 +0000 UTC
View Post
Hey folks, it's been a couple of weeks so I thought I'd share an update.
I've encountered some issues training SD2.1 with specific types of datasets that the model just isn't learning well and it's causing distortions and poor results.
I spent the last couple of weeks siphoning through to figure out the root cause.
Training is resuming again, results are coming through and it's looking good so far.
Those who have attempted to train SD2.1 on new NSF...
2023-03-02 17:52:19 +0000 UTC
View Post
Control Net is a way to lock in an image you are generating, tieing it to a specific structure you preset via an image input combined with a prompt.
Edit: Updated guide with correct extension repo link
2023-02-14 23:14:31 +0000 UTC
View Post
Hi all,
been finetuning SD2.1 for a good while now, it's a very very slow process to ensure it's done right.
I'm following the same approach I took in the small dataset sample I released a few weeks ago but this time at a larger scale and a slower learning rate. This means it's taking way longer to train but I think it's worth it in terms of quality.
Here's a sample of how it's looking today, still not ready but I'll highlight some pro's and con's so far. Even s...
2023-02-05 22:02:41 +0000 UTC
View Post
Some of you folks asked about captioning, I don't think there's a need for a full guide on captioning however here's a post about it
Finetuning:
I used to use CLIP but for proper training a model I don't like the result of the various CLIP models available. So I manual caption images. I batch create text files for each image I'm using and I manually type the caption out. I find it quicker to type on an empty text file than to edit part CLIP captions for hundreds...
2023-01-20 22:27:55 +0000 UTC
View Post
SD1.5, image samples are also SD1.5 with this embedding
sample prompt
high detail photo of gothic jenortxyz , provocative sexy pose, perfect face, stunning, (hair bangs and braided pig tails:1.5), (wearing sexy lingerie:1.5), pale skin, perfect hands
2023-01-20 20:37:36 +0000 UTC
View Post
a photo of sexy billxyz-8000 (wearing provocative outfit:1.3), accurate face, perfect eyes, large huge breasts, symmetrical eyes
SD1.5 - 8k steps - no captions
2023-01-20 14:12:17 +0000 UTC
View Post
CONTEXT
So as you know from a previous update, I've run a test of training NSFW content into SD2.1 and it worked well on a small dataset of 300 images across 6 different types of content.
Since the last update I have been gathering more datasets, captioning and here's where I'm at:
CURRENT STATE
- I've started training SD2.1 again, this time the "real" training, not a test run.
- I'm training one dataset at a time, ensuring...
2023-01-18 23:10:15 +0000 UTC
View Post
Free
Trained on SD1.5
Dataset included
2023-01-18 20:15:46 +0000 UTC
View Post
Rihanna trained on SD1.5
photo of rhiaxyz123 wearing sexy tight gym leggings , provocative pose , in a gym, perfect face, perfect eyes, from front
Images attached with metadata
2023-01-17 10:57:06 +0000 UTC
View Post
Photo of hollywillxyz-5000 standing wearing a beautiful (revealing lingerie:1.2), high quality, perfect eyes, perfect face
SD1.5
All the way up to 20,000 steps included
2023-01-16 16:27:46 +0000 UTC
View Post
Trained on SD1.5
Sample Prompt from this image:
perfect high detail photo of emiliahassan woman (wearing lingerie:1.2), perfect eyes, perfect face, bleached blond hair
Negative prompt: anime, cartoon, fake, drawing, illustration, boring, 3d render, long neck, out of frame, extra fingers, mutated hands, ((monochrome)), ((poorly drawn hands)), 3DCG, cgstation, red eyes, multiple subjects, extra heads, text ,watermarks...
2023-01-15 14:19:01 +0000 UTC
View Post
I've been finetuning SD2.1 with various NSFW images as a test run to see the potential. The level of detail you can get is awesome but it is harder to finetune as it lacks proper concepts of anatomy and positions related to NSFW.
Here is how it looks so far, as usual, patrons will gain access first based on tiers to test this model. Early access this time will be different to give higher tiers more time to spend playing with it.
Although these images are not too bad, it's ab...
2023-01-13 00:38:22 +0000 UTC
View Post
HassanBlend1.5.1.2 is released, It's an iteration and improvement from 1.5 with a removal of a merge and some additional tweaks made.
https://civitai.com/models/1173/hassanblend-15-and-previous-versions
https://huggingface.co/hassanblend/HassanBlend1.5 (in progress uploadi...
2023-01-08 20:19:25 +0000 UTC
View Post
1: Focus on more datasets to finetune into StableDiffusion1.5 and SD2.1, I currently have high quality datasets for the following to start training:
- BDSM
- BJ
- All sort of breasts data, close up, nipple macro, skin texture on breasts etc
- Buttplugs
- Closeups of assholes and vaginas (different muff types etc too)
- Dildo/Vibrator
...
2023-01-07 15:32:30 +0000 UTC
View Post
HassanBlend1.5 is now released!
Thank you for your support and patience. The newest version is live. It doesn't stop there, I'm already looking into improvements for V1.6 and I have a bunch of high quality datasets ready. I will be creating a Poll here and in discord for the community to help provide suggestions on datasets to include. In the meantime I'll continue creating guides/hypernetworks/embeddings and finetuning the model.
About this version<...
2023-01-06 20:26:06 +0000 UTC
View Post
https://discord.com/channels/1026788486818320405/1060138000333406269
You need to link your discord to patreon first, then you get a role that says you are subscribed on patreon and gain access to a hidden channel https:/...
2023-01-05 20:57:23 +0000 UTC
View Post
https://discord.com/channels/1026788486818320405/1060137796209213460/1060534918779764786
You need to link your discord to patreon first, then you get a role that says you are subscribed on patreon and gain access to a hidden channel
2023-01-05 12:32:22 +0000 UTC
View Post
Here's a summary of the hypernetworks I've released here on patreon when combined with HassanBlend1.5 being released.
chavgirlshassan,
hassans-detailed-skin ,
hassan-face-enhancer ,
hassans-eye_skin_enhancer ,
longexposurehassan ,
nudewomenposes ,
2023-01-05 00:16:58 +0000 UTC
View Post
Trained a new hypernetwork on 200 images of skin macro shots, some showing goosebumps, hair standing up, imperfections, stretch marks, wrinkles etc.
I'm sharing all the hypernetworks from that training batch but my recommendation is to use the 40,000 step at a strength around 0.5,
If you use the 10/20k step version, I'd recommend lower than 0.5 before it starts to have too strong an effect.
Some text it was trained with:
Wrinkled Skin
2023-01-04 22:41:08 +0000 UTC
View Post
