💖 Support & Share the Magic!
Love these embeddings? Like, add to collection, and tag this resource in your creations—I’d adore to see what you make! 🎨✨
👉 Looking for amazing anime model? Check out JANKU by yours truly, you won't be disappointed. It is highly tested with my embedding! It also has ONSITE generation support!
👉 More goodies on my profile! Follow for updates!
👉 Join me on our discord server (https://discord.gg/tWrKk3jcXx)! Share art, ask for help, or just chat. We have CivChan bot plus games there!
☕ Support future updates: Buy me a coffee or tip buzz! Every bit helps. Thank you for fueling the art! 🚀
Want to improve your images with zero effort even more? ✨
TrendCraft is the only detailer LoRA trained on the last 30 days of trending CivitAI images —just plug into Pony XL , Illustrious , SDXL, or Noob Vpred workflows for instant, viral-ready upgrades. No prompts, no work. 🪄🔥
🎨 Lazy Embeddings: For Creators Who Hate Typing (But Love Results) 🛋️
Why waste time typing prompts? 💤
Lazy Embeddings = pre-loaded magic for Illustrious/NoobAI/SDXL/Pony XL. Just drop lazypos + lazyneg and go . No copy-pasting, no tears.
⚡ How to Be Lazy (Illustrious/Noob/SDXL):
Positive Prompt: Start with
lazypos(orembedding:lazyposfor ComfyUI).Negative Prompt: Start with
lazyneg(orembedding:lazyneg). If you are using onsite generation, make sure to use lazyneg v3.1 or higher.Done. Seriously. 🙃
Pony XL Users:
Positive:
lazyup(orembedding:lazyup).Negative:
lazydn(orembedding:lazydn).Pro Tip:
lazympos+lazyup= ✨ next-level quality ✨
🔥 If Hands Suck:
Add lazyhand to your NEGATIVE prompt. Too lazy? Just use lazyhand + lazyneg (boom , fixed).
Wet and Shiny Skins:
Add lazywet to NEGATIVE prompt to remove wetness and water drops from skin. It will reduce/remove shiny skin as well! It is a must have in all my prompts when doing SFW.
🙈𝐍𝐒𝐅𝐖:
Block not safe for work with lazy𝐍𝐒𝐅𝐖 in negative prompt. If you want more use it in positive prompt!
Mature characters: Use lazy𝐥𝐨** in negative to make characters more mature and don't get flag on CivitAI!
Realism/Anime: Use lazyreal in negative if you want only anime output. In positive, if you want realistic people output.
🌟 Why Lazy Wins:
⚡ Zero Effort: Skip the prompt grind.
🛠️ Universal: Works with any Illustrious/Noob/Pony/SXDL model.
🤖 Battle-Tested: Trained on prompts real artists actually use .
💡 Peak Laziness: lazypos + lazyneg = flawless art with 2 words . Check the examples—it’s easier than scrolling TikTok .
Stay Lazy. Generate Smarter. 🚀
(Because life’s too short for repetitive typing.)
Protip: Embeddings don't cost any extra buzz for using. If anything, it will save you buzz since you won't generate bad images!
🛠️ Can’t See Lazy Embeddings for Onsite Generation?
Enable Advanced Mode:
Go to Account Settings → Toggle “Advanced Mode” → Now you can select Lazy Embeddings for any model.
🚀 ComfyUI Users:
Drop embeddings into your
embeddingsfolder.Positive Prompt:
embedding:lazyposNegative Prompt:
embedding:lazynegPro Tip: Use ComfyUI-Custom-Scripts for autocomplete magic! (Though
lazypos/lazynegare short enough to remember 😎)
💫Examples done with my model, JANKU:
Here is what they are doing under the hood so you don't have to type it! Please note I'm always taking feedback and updating these embeddings. I use them daily!
Illustrious XL / NoobAI / SDXL - see below for Pony XL
Positive prompt (lazypos v2.0):
masterpiece, newest, absurdres, best quality, amazing quality, very aesthetic, ultra-detailed, highly detailedNegative prompt (lazyneg v3.0):
low quality, worst quality, lowres, username, sketch, censor, blurry, distorted, bad anatomy, signature, watermark, patreon logo, artist nameProtip: Use lazyhand embedding in NEGATIVE if you want to fix the hands for the seed!
Pony XL
Positive prompt (ponyup):
score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_upponyup7 only includes score_9, score_8_up, score_7_up
ponyup6 only includes score_9, score_8_up, score_7_up, score_6_up
Negative prompt (ponydn):
score_4, score_5, score_6Other versions:
lazympos
masterpiece, best quality, amazing qualityDescription
Fixed color issues with onsite generation. If you are having color issues for onsite gen, put lazyneg last. Note this is only for onsite generation, local generation has no issues!!
Reduced tokens.
FAQ
Comments (26)
lazyfeet when?
😏
+1
For what?that model work good with legs and toes
@DOLPH1NARU 任何模型，当我想看到可以舔的美味玉足的时候😋
@xiaoyinbi Owwww, okay then
GREAT!!!!
Lazyhand v2.0 released!
* Fixed color issues for onsite gen. If you are having color issues still, put lazyhand first and always put lazyneg last! This is only for onsite gen.
* Lower token usage for lazyhand, should work better or the same.
Make sure you download the latest version as I had to re-upload twice due to bug.
Thank U!
It works really well <3
Is there a folder with examples? I seriously used several different embeddings, I ran through AI works with examples, and it's ALWAYS a placebo.
There’s no real benefit to using embeddings if you’re already including the right quality tokens in your prompt (see the section below). Embeddings are essentially stored prompt fragments. Their main purpose is to save tokens or preserve phrasing consistency. They mattered more in SD 1.5, since prompts were truncated after 75 tokens, but with SDXL’s larger context window they’re usually unnecessary unless you’re using very long or complex prompts.
LoRAs, on the other hand, actually involve training and can modify how the model (and its CLIP encoder) interprets or weights certain concepts, which makes them much more impactful than embeddings.
As for it being placebo if you WEREN'T using them... depends on the mix you're using. Some of the base models (Pony, Illustrious, etc.) are extremely finicky to even get working properly without the right prompts. If you message me with what model you're using along with an example prompt I could help you troubleshoot a little more.
@jhininabin Not specifically one. I have over 20 models and 400+ Loras for different jobs. I can't even find which artworks in my work have built-in embeddings and which don't. But when I tested different models, I didn't notice any difference. (The difference between them is like different seed.) I even showed my images to three different AIs, and they couldn't tell the difference either. The authors of Embeddings also never show at least a dozen examples for comparison.
@Twostep Right, but remember that if you're already following the prompt guidelines for those models (masterpiece, best quality etc for positives and low quality, worst quality, etc for negatives) then the embeddings here will make no difference because they are just those prompts packed into a single prompt for convenience.
@jhininabin I disagree that there is no real benefit to using embeddings in SDXL. The disconnect here is the term "embedding". Bundling a series of tags and or prompts into a single trigger is best described as a "runtime-merged embedding". This differs greatly from a "dataset trained embedding", which is trained on a dataset of images and caption/tag files, similar to how a LoRA is trained. The difference between a LoRA and a dataset trained embedding is a LoRA effects a checkpoint's U-Net, the mechanics that do the drawing, and many times the text encoders as well, the mechanics that interpret the text prompt, whereas a dataset trained embedding only effect the text encoders.
Dataset trained embeddings are far more effective than runtime merged embeddings. They can have a similar effect as a LoRA for narrow concepts like a specific face, character, object, ect.
我想请问，这个embedding的作用是将大量提示词合成为一个提示词，只占用一单位的令牌，还是说仅仅是一共文字宏，是输入大量提示词的快捷指令，仍然占用大量令牌？
On of my favorite embeds tbh
хмммм...
who is lazy ember?
Jan the absolute goat❤️
No, you are the goat 💪
This is an amazing embedding I really love it.
However, after switching to SwarmUI recently, I’ve been having a bit of trouble getting it to work properly.
If I want to use this embedding in SwarmUI, should my prompt be written as:
1. <embed:POS/janxd/lazypos/lazypos>,lazypos,
2. <embed:POS/janxd/lazypos/lazypos>,
3. lazypos,
Which of these is the correct way for it to take effect?
I’ve been stuck on this issue for quite a while, and I’d really appreciate it if anyone could help clarify.
it should be like this embedding:lazypos because swarmui uses comfyui under the hood
option 2 is what you use for SwarmUI
FLUX USERS: It's worth it using lazyneg but not lazypos with flux prompting. I made a couple of comparisons with a fixed seed and the result is always better with lazyneg - much sharper, much better backgrounds, sometimes also the close plane - however - it shines in the backgrounds. Tested both with Nepotism, Cyberrealistic, Timeless & NSFW Mix. With lazypos, it is sometimes better, sometimes worse, depends on composition.
Of course, you'll get a typical "size mismatch" error and information that the embedding will be ignored but the trigger words are applied anyway, which actually helps a lot and makes it easier - more lazy - as always.
MY CURRENT STRATEGY:
lazypos + lazyneg --> for backgrounds detailers
lazneg (only) --> for people detailers, because lazypos makes things "anime-ish" - with FLUX, it results in making body details blurry - I am assuming that it is the issue with anime being popular as input data so - it overprioritizes the anime-ish feeling about the bodies.
Cheers.
Tell me, can lazyhand be used for hints in ADetailer?
LazyNegative's v3.0 and v3.1 have the same hash, but what's the difference?
Details
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.