[Stable Diffusion] Prompt Sharing and Learning Thread

[Mr-Fox]

Thank you very much, you are very kind.
I have one last question for the moment (I hope).
When you say to focus on one concept at a time, do you mean that I have to reorder the prompt by categories (eg: quality / appearance / action and pose / body part / related body part), or I have to generate an image first and then update it with new prompt? (in this last case how should I do?)

I meant focus on generating images with only one concept to begin with. After you get good result with one, go to the next concept and focus on this. After you have started to get some consistent results with one concept per image, you can start to try more than one concept per image. To learn and to see what effects what, you need to exclude as many variables as possible. Start with trying to get a good image of the legs spread, then when you are getting good results, go with the next concept. One concept is difficult enough for even an experienced SD user, no need to make it more difficult than it needs to be. Also read the advice I gave to fustylugss.

 

[FallingDown90]

You are absolutely right -.-" thank you very much

 

[Mr-Fox]

General tips for basic prompting. Keep it simple.. Avoid a ton of different tags, saying the same thing. It doesn't help anything.
Don't use too many additions like Lora's or Ti's. Getting one to work correctly is hard enough. Use simple clear phrases and descriptions. SD "understands" some technical language but most of the time assume it doesn't and describe it instead. It's not real AI, there is no intelligence here. No "singularity". This is machine learning (deep learning) and algorithms. SD is language to image generation and built on "machine learning" language models. In order for SD to "know" anything, someone must have trained it with an image and a complimentary prompt or description that it then "learn" to associate with the concept of this image.

General tips in general.. No1 get a decent checkpoint model that is good at the style that you wish to achieve, it can't do anything it hasn't been trained to do though. No2 the prompt is the most powerful tool we have, the negative prompt in particular. No3, if the checkpoint and prompt doesn't achieve the desired result now is the time to add Lora or Ti. First try with only checkpoint and prompt though.

 

[FallingDown90]

You have been really clear and very helpful, thank you very much. Leaving aside the perverse part, with your advice I think I will be able to make the most of SD also to create references for my work. Let's hope so... in the meantime, if one day I can fix Bea I'll show you the results

 

[Mr-Fox]

You have been really clear and very helpful, thank you very much. Leaving aside the perverse part, with your advice I think I will be able to make the most of SD also to create references for my work. Let's hope so... in the meantime, if one day I can fix Bea I'll show you the results

You're welcome. Sounds great, I'm looking forward to it.

 

[Sharinel]

72 hours isn't much though,

Pfft, rookie numbers

 

[Mr-Fox]

And beachbody of the year goes to..

Spoiler: More

You don't have permission to view the spoiler content. Log in or register now.

 

[Mr-Fox]

Pfft, rookie numbers

I really didn't mean anything by it..

 

[FallingDown90]

A stupid question... from the render previews it looks like it achieved the goal... but i think it crashed and didn't save the image... is there a way to figure out if. Am I wasting time waiting for something to move? Or is there something that. can I do to save what can be saved?

Spoiler

You don't have permission to view the spoiler content. Log in or register now.

 

[fustylugss]

This is great. Welcome to the gang. 72 hours isn't much though, if you meant that you just started. It's a good start though. Continue to experiment, this is how you learn. Read guides and watch tutorials. On the first page in this thread you can find links to some of the guides. Some of the info might be old, incorrect or made redundant at this point though. You can also simply use the search bar for this thread and filter for members, use my name Mr-Fox or Sepheyer, devilKKW, Jimwalrus, Schlonborn, Dag00th, me3, Sharlotte. These are some of the most active members in this thread (though I might forget someone) and we have all made guides or posted tips. You can learn from all of us.

You must be registered to see the links

and

You must be registered to see the links

are some of the most popular and active posters on youtube. They might not be the most knowledgeable or most correct but they are great at introducing you to a new tool or technique and show you the basics.

Thank you for the comment. I followed your steps and the youtubers you mentioned. I also took one pose from Dag00th.
Honesty, I am having trouble with making LoRa models . I need to do more research and make my own. Cheers !

 

[fustylugss]

Pfft, rookie numbers

We all gotta start somewhere! Although the devianart link is killer! Thanks!

 

[onyx]

A stupid question... from the render previews it looks like it achieved the goal... but i think it crashed and didn't save the image... is there a way to figure out if. Am I wasting time waiting for something to move? Or is there something that. can I do to save what can be saved?

Spoiler

You don't have permission to view the spoiler content. Log in or register now.

What size are you using (Width/height)? If you have it too high you're likely running out of memory.

 

[FallingDown90]

I use a preset from a picture "resize: from 512x768 to 1024x1536"
Now I try to decrease, thanks

 

[me3]

I'm getting closer at least, still struggling to make the actual body being fire. Whoever keeps saying that AI images is just throwing some words together really have no clue...

(at least she's hot...)

The downside of creating large HQ images is that you need to compress them to be able to post...which i guess kinda defeats the purpose.
Outpainting seemed to have not gone too badly

 

[Sharinel]

The downside of creating large HQ images is that you need to compress them to be able to post...which i guess kinda defeats the purpose.
Outpainting seemed to have not gone too badly
View attachment 2802093

Quake skins have come a long way. (Sorry that's the first thing it reminded me, I r old)

 

[Mr-Fox]

Thank you for the comment. I followed your steps and the youtubers you mentioned. I also took one pose from Dag00th.
Honesty, I am having trouble with making LoRa models . I need to do more research and make my own. Cheers !

Ambition is really great but get proficient with SD first before you get into training a Lora. You risk burning yourself out of interest if you get into too many things at once. It really takes a while before one gets a handle of SD, even now that there are so much good information about it. Training a good Lora or Ti is yet another can of worms to get into. Don't misunderstand me, I'm not trying to discourage you. I only want to help you not bite off too much at once. When you decide that you want to give it a go, watch this

You must be registered to see the links

by Aitrepeneur to get setup, read the awesome post by Schlongborn, and this

You must be registered to see the links

is the best info source, it gets updated on the regular. Don't use the settings by Aitrepeneur, copy the settings from the Rentry instead.

 

[Mr-Fox]

I use a preset from a picture "resize: from 512x768 to 1024x1536"
Now I try to decrease, thanks

What is your GPU and amount of Vram? This is the limiting factor for wich resolution you can use generating images.

 

[FallingDown90]

What is your GPU and amount of Vram? This is the limiting factor for wich resolution you can use generating images.

How can i check the amount of vram?
Anyway i use parallel desktop on a macbook, i know it's crazy in fact in the future i will use my pc but sometimes it's necessary.

Thi is my Mac info:

Spoiler

You don't have permission to view the spoiler content. Log in or register now.

In any case, even if slowly and with the difficulties you know (beyond various blackouts in the city) I obtained a satisfactory result.

I hope you appreciate it and that you can give me suggestions on how to improve the bugs you notice and the various inconsistencies with the prompts.
Also I would be grateful if you can also recommend me some useful extensions.

I leave you the render and some settings I used:

Render (sexual contents):

Spoiler

You don't have permission to view the spoiler content. Log in or register now.

Settings and my Extension:

Spoiler

You don't have permission to view the spoiler content. Log in or register now.

Once again, thanks for everything

 

[Mr-Fox]

How can i check the amount of vram?
Anyway i use parallel desktop on a macbook, i know it's crazy in fact in the future i will use my pc but sometimes it's necessary.

Thi is my Mac info:

Spoiler

You don't have permission to view the spoiler content. Log in or register now.

In any case, even if slowly and with the difficulties you know (beyond various blackouts in the city) I obtained a satisfactory result.

I hope you appreciate it and that you can give me suggestions on how to improve the bugs you notice and the various inconsistencies with the prompts.
Also I would be grateful if you can also recommend me some useful extensions.

I leave you the render and some settings I used:

Render (sexual contents):

Spoiler

You don't have permission to view the spoiler content. Log in or register now.

Settings and my Extension:

Spoiler

You don't have permission to view the spoiler content. Log in or register now.

Once again, thanks for everything

The name of your GPU is Radeon Pro 560 4 GB. Notice the 4 GB, this is the amount of vram your card has. This means that you have low vram and this is very limiting. If you have a pc with a better card, use this instead.
Regarding the image and your settings.
To be honest I think you should choose a more simple image with a more simple concept. Example, only a character standing with the beach backgound.
Start without any Lora and only work with the prompt and checkpoint. Remember only one concept at a time to begin with. Structure the prompt. I usually go with: image style, image composition, description of the subject and clothing, then light and color etc, if any Lora is added put it on the end so it's easier to overview the prompt. Only use the positive for the things you want to see, don't use "no shirt etc in positive it belongs in negative. In the negative add the things you don't wish to see, remove the word "no" and add shirt and knepad etc here. Then when you have achieved something good, move forward. This is when you introduce the pose and work with the editor to refine it. Then introduce the masturbation, you might need a lora for the masturbation. Try to get the body with muscles etc with only the prompt at first. Only add the muscle Lora if you have to.
Don't use easy negative Ti or easy prompt etc, try to get a good result with the prompt first. Those are only bad shortcuts for lazy impatient people that have no clue imo. Maybe a bit harsh but I have never had any good result with these things.

 

[me3]

With

You must be registered to see the links

being out they've at least reduced the size of the file to about half of 0.9
There's claims you need at least 12gb vram to use it, which we already know is a massive lie considering comfyui already ran it fine (and fairly fast) on far less. On my 6gb it used less than 5gb and didn't use too much longer than other models on the same image size etc.

Testing in a1111 is another story though. To LOAD the model you need a lot of RAM, not vram, without settings you probably need close to 16gb, for me it used close to 32, but that's probably a case of "take as much as is free".
Having OOM with generating, while i can load it fine, even with --medvram it OOM for me so far. Maybe some settings can be tweaked, but hopefully they can optimize things in the code as it clearly should work on much less vram usage (which the case of comfy proves).
--lowvram lets me generate images, but speed is a major issue....3,5-4min on a 1024x image...funny enough it only uses 2.9gb during so in theory you should be able to use it on pretty small cards...with a lot of patience

(updated: have gotten it to run with --medvram so that's possibly with just 6gb vram too)