Dreamshaper XL Lightning
First time I comment on a model because I would definitely say that we got the model that actually understands our prompts and gives the best results in the first attempt, which definitely saves lot of credits. THIS IS ABSOLUTELY AWESOME.
Well said - I get the best results even with simple prompts (but very often I submit 'crazy' prompts, and Lightning does not let me down). The rendering is absolutely superb! I return to this model time and again when others fail.
I disagree... This model is not suitable for people who are inherently in the LGBT+ community, especially if you are trying to portray people who are different from each other. It does not respect different ages, genders, hairstyles and clothing choices. It chooses ONE of your prompts and makes all people awkwardly look the same. I call it "The Agent Smith Effect"... If you've ever seen The Matrix then you understand why I call it that.
For several days I have been experimenting with a broad range of prompt directions in order to discern the merits and flaws of Lighting. The Merits are piling up and the flaws are few and far between. With sucinct and specific directions Lightning produces what it is asked to without expressing a baked-in style of its own.
I am particularly impressed by what it does with faux-photography, a favorite GAI pursuit of mine. Currently I am discovering the wonderful manners in which it generates detailed fabric and other materials.
A very fine and appreciated tool to have up on the shelf while crafting! Thank you for making this available!
Some notes which may be helpful to others:
If seeking a high quality photographic look, begin at the top of your prompt with
[ professional photography, ]
Lightning responds exceptionally well to brief and clear directions; this is a direction that it knows what to do with.
If you are working to bring out the fine qualities of a particular fabric, precede the specific material with the words "intricately detailed", as in :
[ intricately detailed satin dress, ] or [ intricately detailed leather jacket, ]
Lightning performs wonders on most materials when specifically asked, so also try things like:
[ intrcately detailed ornate jewelry ] or [ intricately deatiled rusty metal armor, ]
Finaly, if you have a result which you especially like you can refine it further. Choose to Evolve the image. Then in 'More Settings' turn off 'use random seed.' & lastly change the Runtime from its default of 'Short' to 'Long'. This costs a full credit but the result will make you smile.
Wonderful tips, @Possibility! I am absolutely sure these tips will help out newbies and others that needs some guidance, generally or specifically re this CKPT. Very kind of you!
I can vouch for using the same seed when you find one that gives specific results! I have some really old ones I re-use as I can get a specific result that I can't obtain elsewhere🌞😊
Thank you
I am happy if these tips are of value to you.
:-)
Thanks
same here - I am digging back into my old SD v1.4 and v1.5 results and run them again with Lightning - absolute great images. Many of my prompts have me as the artist in the modifiers, and Lightning returns some good results that resemble my analog style.
That sounds awesome!👏🦋
Excellent advice. Thank you for taking the time to share.
@Possibility I would agree with all you said but would like to add something extra. The name Dreamshaper is aptly given because the images it does best with are something you would "dream" up - whimsical, stories, fantasies - kind of hard for me to put into words. It works wonders for many of the images I want to generate - images that tend to tell a story, for the most part.
This model also responds very well to including artists' names if you're looking for a style they do. Including a number of different artists, especially if some have widely different styles can give some surprisingly good and unexpected looks. I've kept notes on which artists work well together. (Looking through others' images, you'll soon be able to recognise Gustav Klimt was used!)
What I found it doesn't do well (for me anyways) is realistic animals or insects. Insects and spiders will have far too MANY legs 😱 and animals better suit a fantasy or fairy tale setting. Google Imagen Fast is great at animals.
I notice there are no prompt weights or refiner weights; what does the model assume for these values?
It's a bit different to other models. The prompt weight is actually lower than the standard 0% prompt weight would be. Refiner is not used.
Something else I've noticed about this model is that when I do a prompt with flowers, it tends to give me a wallpaper pattern of flowers instead of a close-up of a few flowers.
This model is literally doing the opposite of everything in my prompts, if I say white girl driving a black car, it gives me a black woman driving a red car. If I say that one lady has a ponytail, it gives me 4 men with ponytails and a bald woman. Why prompt at all?
I have the same problems with every model I've tried except the Dreamshaper line lol
I'm thinking maybe some people just get along better with certain models because of their prompting style
THIS MODEL IS GREAT! Probably the 2nd best one! (Not loras or pro models)
Wow algorithm especially for evolving. Thanks for creating it.
Dreamshaper is an amazing model, which (personally) comes first tied with Stable Core!
I absolutely love working and experimenting with this model. It’s one of my go to when experimenting with prompts.
Love this one! Nice and cheap! Only problem is that it can't do watercolor, but that's okaym
I can see how it is different. I have used it a few times. Sometimes it is ok.
Love this model. I just started using it. I just wish i could get the hands right ...you get a beautiful creation but the hands can let it down ...any ideas would be appreciated..👏👏
A "Fix Hands" tool like the Fix Faces tool would be awesome!
I'm curious about this quick and relatively efficient model. But there are really not many settings here. For example, what sampler method and how many sampling steps does it use? No option for uploading reference image too. Some times images are good but some times it has a big problem with multiple limbs, fingers, tails etc. no matter what your negative prompt is.
Dreamshaper XL Lightning did not give me the cat I had in the prompt for this image and this is not the first time this has happened. I wonder why it refuses to give me the cat I had in the prompt. Strange.
I use this model often, but in some cases, I have a problem - even if I wrote MAN and WOMAN, I get results with two women or men. Sometimes, 10 or 15 times in a row. I change the positions of those words, switch their places, and put them in () or ", but nothing. Any help on this, please?
Question: if an image is generated by NightCafe using "create" and "text-to-image" using THIS model, are the images created free to be used elsewhere, or are they considered under copyright?
This model is fire absolutely amazing ability one of my very favorite ones I learned a lot from using it so good always gives amazing results
This model the color range seems to not be anywhere near at vibrant is the base SDXL. I can clearly tell most of the time when a creation is made with a lightning SDXL model. Constantly gives wallpaper like results. The results are often shaded dark. even at half cost of regular SDXL its not half as good. unless you have unlimited your most likely going to spend more credits to get something good over a normal SDXL model. It does ok on really common stuff that you see over and over again all over the site.
I used this model a short time ago today. I think it’s a little tired. Unfortunately, the young woman in my image has three legs! It’s all good. 😉
currently it is not the same model as before, there is only the same name, at the very end of 2024 they changed, broke this model, from the same prompts, it creates completely different images than before
One of my favorite models to use! I make a big batch of images in v8 until it says I'm creating too fast, delete the ones I don't want, then evolve the best ones to this model to make images of my characters I never would've been able to make without AI. It doesn't follow my prompts quite like v8 does but that's fine because evolving from model to model is a thing. For costing only half a credit, it's absolutely awesome! 💖
🤖 NightCafe
🚀 Features
🔒 Secure Payments
PCI compliant payments powered by Paddle and PayPal and Shopify
Copyright © NightCafe StudioCommunity Standards Acknowledgements Privacy Policy Terms of Service Refund Policy Sitemap
Comments