- Jun 7, 2021
- 105
- 75
Hi Nickohoma, hi all,
First at all, I'm happy to see that this thread has a kind of fan-base. This game is cool, models are great, so let's try our very best to make it live as long as possible.
Then regarding AUTOMATIC1111, I need your expertise. I followed the installation instruction, but when I try to generate images, well... As you put your generation parameters within the image, I tried to import one of them. New example with the last one :
Then I used the Send to txt2img to import the same parameters. Then I generated an image, and surprise :
Well, it is more or less the same pictures (or not). Could you please help me to understand what's happening ?
And yes, I know you propose to DM you, but I prefer to ask here, in case of someone has the same issue than me.
Last but not least, if an image has some small errors, such as number of fingers, missing hairs... Do you use a specific tool such as Photoshop ? Or do you re-generate the image with more parameters ?
Best regards,
First at all, I'm happy to see that this thread has a kind of fan-base. This game is cool, models are great, so let's try our very best to make it live as long as possible.
Then regarding AUTOMATIC1111, I need your expertise. I followed the installation instruction, but when I try to generate images, well... As you put your generation parameters within the image, I tried to import one of them. New example with the last one :
Then I used the Send to txt2img to import the same parameters. Then I generated an image, and surprise :
Well, it is more or less the same pictures (or not). Could you please help me to understand what's happening ?
And yes, I know you propose to DM you, but I prefer to ask here, in case of someone has the same issue than me.
Last but not least, if an image has some small errors, such as number of fingers, missing hairs... Do you use a specific tool such as Photoshop ? Or do you re-generate the image with more parameters ?
Best regards,
- Jan 11, 2022
- 53
- 207
Hi Shuyaourt, and welcome.
Glad to help you get set up in Auto1111. Biggest things you need are probably the Pony checkpoint and the Lora's.
You must be registered to see the links
You must be registered to see the links
You must be registered to see the links
You must be registered to see the links
The checkpoint goes in \stable-diffusion-webui\models\Stable-diffusion, and the checkpoints go in \stable-diffusion-webui\models\Lora.You may need to relaunch Auto111 after you have those in place.
I'd also recommend pulling a different image into PNG info, like one of the bj pics from my first post. The Laura/Ashley pic you pulled uses regional prompting, which might not be installed by default. The pov bj pics are all more straightforward.
Those two things should get you much closer.
- Jan 16, 2023
- 3
- 7
I plan to post the rest of Laura's training soon, I already have some additional ideas to continue this arc of Laura's training ☕
- Jan 11, 2022
- 53
- 207
Yeah sometimes you get a near-perfect image but then have some extra fingers or something. First thing I try is to regenerate with the same prompt, same seed, but a variation seed, with low variation strength.
I prefer that method because then you I can still regenerate the image later with the exact same parameters.
If I have no luck with that, I toss the image over to the inpaint tab. You can select and regenerate just a portion of the image that way. It's a little tricky but usually does the job.
Some people definitely do touch-ups in Photoshop but I have zero actual talent so Photoshop is a mystery to me.
Last edited:
Much much better.
2 additional questions :
- In your parameters, you have this lora : lora:Expressive_H-000001:1. Should I also download it ? On civitai.com, it seems there are a lot of lora.
- How many time it takes you to generate an image ? The system needed 10 minutes to generate this one, but I think it is more related to my setup (RTX 2060).
2 additional questions :
- In your parameters, you have this lora : lora:Expressive_H-000001:1. Should I also download it ? On civitai.com, it seems there are a lot of lora.
- How many time it takes you to generate an image ? The system needed 10 minutes to generate this one, but I think it is more related to my setup (RTX 2060).
- Jan 11, 2022
- 53
- 207
Yep, much closer.
ExpressiveH is another style Lora. I usually only use it on actual sex action shots, since it doesn't seem to do much outside of those. 100% optional if you'd want to include it in your stuff. I also use a penis size slider Lora most of the time there's a penis in the shot.
If Hires. fix is turned on when you import that image, you can probably turn that off. I only run that for my "final" versions of images. It upscales them makes them look nicer imho but totally not necessary and probably triples generation time.
ComfyUI is also significantly faster than Auto1111. I started with it but switched to Auto1111 when I couldn't figure a few things out. ComfyUI is much closer to "bare metal" than Auto1111, so you need to understand more about what's going on to get the most out of it.
From what I understand, the biggest factor in how long an image takes to generate is how much video ram you have. The SDXL/Pony model is a little over 8GB, so if you have > 8GB, it can load the whole model in memory and run much faster.
On my 3060 / 9900 PC, those images take about 40 seconds with Hires. fix off, just for reference.
Last edited: