I use civitai to download loras and checkpoints if I need them. I use comfyUi and a1111 localy to generate pictures and videos)
- Thread starter aDDont
- Start date
-
- Tags
- 2dcg ahegao ai cg anal sex animated bdsm bukkake cheating corruption creampie dilf exhibitionism female domination female protagonist footjob group sex handjob humiliation internal view interracial male domination masturbation mobile game multiple penetration netorare oral sex pov romance sex toys spanking stripping swinging teasing vaginal sex voiced
There is only one thing dev have to change please is the text size some times it didn't understand what dialogue written please add a text size so player can adjust the text size
Probably not very soon - many people asked to stack up more content between releases, so I probably stack up a couple more scenes first (I guess something like 5 at least). For now i've finished 0.5/5
- I wanted to train a lora for wan but failed miserably - it's came out to be a lot harder then I expected. Then I tried VACE rendering in hope to overcome the missing Lora limitations, and failed again, broke my workflow because of some nodes or dependencies I installed, then spend lots of time repairing my workflow, managed to make it even better, understood that I am mistaken, and finaly came back to work in more or less the same state as I was before the whole Lora training experiment. Lots of time and nerves wasted So it mean your working husband sharing Storyline where Mila kiki and the other Girl involved?
Thankyou for replying So Fast you are a really humble personProbably not very soon - many people asked to stack up more content between releases, so I probably stack up a couple more scenes first (I guess something like 5 at least). For now i've finished 0.5/5
I will work on this route later, for now my only focus is the NTS route)
Thanks for the information.
Is there a good side to were KI generates an NSFW story with pictures?
I have no idea tbh)
The same I've been working on in 1.5.x versions - Bob's slut (sub to everyone).
Joshy92
Message Maven
- Mar 25, 2021
- 13,983
- 35,488
- 967
TheresiaW
Member
- Jan 1, 2020
- 228
- 493
- 134
Do you have a good step-by-step guide for beginners with no experience?
I downloaded ComfyUI, installed ComfyUI Manager, and downloaded a few models from Civitai that I think look good.
Nevertheless, when I try to create something, I either get error messages or rubbish.
TheresiaW
Well I handle the errors by asking chatGpt what tf does that mean and what can I do to fix it It helps... Sometimes) It's hard to say what is wrong without context, and the fastest and easiest way to fix technical problems is to send logs to chatgpt. Most of the time it's either dependencies, either versions conflicts/missing nodes.
With comfy you can drag and drop generated images into your comfyUI window and it converts it into workflow graph, which supposedly will lead to the exact same image if you hit the button, but there are some weak points. For example my images won't work that way, because i don't use comfy for pictures (I use a1111, the oldest UI), and also because I generate them in iterations fixing stuff I don't like in photoshop. But you can find lots of pictures on civitai that will have the metadata. Also most of the models previews and model's info has the description about the prompt and params.
I would say just copy someone elses prompt and then fix the stuff you don't like. If it's not working - then go from the easiest part - there are templates in comfyUI (on the left). Find there txt2img template for sdxl and try running it as is. If that give you rubbish too - look at the log and send to chatgpt with the description what you've done and what is the problem)
Well I handle the errors by asking chatGpt what tf does that mean and what can I do to fix it
It helps... Sometimes) It's hard to say what is wrong without context, and the fastest and easiest way to fix technical problems is to send logs to chatgpt. Most of the time it's either dependencies, either versions conflicts/missing nodes.With comfy you can drag and drop generated images into your comfyUI window and it converts it into workflow graph, which supposedly will lead to the exact same image if you hit the button, but there are some weak points. For example my images won't work that way, because i don't use comfy for pictures (I use a1111, the oldest UI), and also because I generate them in iterations fixing stuff I don't like in photoshop. But you can find lots of pictures on civitai that will have the metadata. Also most of the models previews and model's info has the description about the prompt and params.
I would say just copy someone elses prompt and then fix the stuff you don't like. If it's not working - then go from the easiest part - there are templates in comfyUI (on the left). Find there txt2img template for sdxl and try running it as is. If that give you rubbish too - look at the log and send to chatgpt with the description what you've done and what is the problem)
Check the checkpoints you downloaded, 9 times out of 10 if you are new, you are missing either a clip last layer restriction, or an activation prompt (if it's a LoRA).
I've noticed that most from civitai use -1 or -2, if you miss this (or another small parameter tweak) you will end up with just latent noise being thrown back at you.
Best suggestion is to view someone else's metadata on civitai, and copy the prompts and values over, and use the same linked checkpoints and loras identified.
If all else fails, use the pre-baked in templates, and make sure those run as-is, and then start tweaking those one value or parameter at a time to see what breaks.
e.g. Swap the model(checkpoint) only to get a new art style, change the values in the Ksampler (cfg, steps, sampler, scheduler, etc.) and see how that impacts the output, and go from there!
Actually knowing and learning what all the things do helps when you want to fine tune and tweak things, or deal with errors, so it helps to poke around.
Vanilla099
Well-Known Member
- Sep 14, 2017
- 1,567
- 139
- 211
Privet friend, I have a couple questions
1. Have you tested Lora vs Dreambooth to train character consistency?
2. Have you tested Flux vs SD? (and which sd version do you use, 1, 1.5 or SDXL?)
3. Have you tested Qwen?
4. Any tips you can share like which Lora's I should definitely use for making nsfw drawn stuff like this consistently? Or prompts?
Thank you in advance
1. Have you tested Lora vs Dreambooth to train character consistency?
2. Have you tested Flux vs SD? (and which sd version do you use, 1, 1.5 or SDXL?)
3. Have you tested Qwen?
4. Any tips you can share like which Lora's I should definitely use for making nsfw drawn stuff like this consistently? Or prompts?
Thank you in advance
1. I am using character Loras for most of the characters in game
2. Flux is llm model. Sometimes it makes something cool. Completely uncontrollable, bad with inpainting. I don't like it. I use sdxl based custom chekpoints
3. I heard of it wanted to test it, but it didn't work from the box and I didn't want to spend to much time on it. People say it's good)
4. You need a good checkpoint and understanding how prompts work. There is no nsfw Lora, there are nsfw checkpoints, they are large and mostly understand stuff you need in nsfw game. Try illustrious family or pony family for starters)
