Please do share your creations if you make something that you like. We can all learn from each other.
For best results:
Start with a resolution of 1024x768 I recommend using the DPM++ SDE or DPM++ 2S a sampler with a CFG scale between 1 and 2.5, and 5-15 steps. Clip skip: 2. And then experiment.
I've seen users make beautiful pics with other settings, so it may be worthwhile to experiment to see what suits your desire.
VAE is included. Generally speaking, adding steps increases detail but if you want to use any sort of LoRA or hiresfix, a higher step count may fry your image. Lower cfg tends to follow the prompt better; higher cfg tends to give more contrast.
2.2 - Hopefully more realistic, though colors are less punchy. Definitely itterative.
1.9 SPO - better colors, and composition, now generates at 1288x1288
1.8 in addition to finer dark elements, now makes coherent pictures up to 1152x1152 without the aid of controlnet.
1.6 improvements to contrast and composition.
1.5 improves color and detail.
1.4 merged in some of my favorite LoRAs.
1.2 Is more creative and tends to handle hands better.
Version 1.0 is functionally the same as 0.9t, the only notable difference being more life-like and saturated colors.
0.9t represents a huge leap over the previous iteration. I've merged in some of the realness of Juggernaut and the responsiveness of OpenDalleV1.1 and crammed it back down to Turbo(merge) speed.
Description
Experimental Build
FAQ
Comments (36)
Could you please add 6G pruned version too? Thanks.
Looks great for a turbo model, any chance of getting a pruned fp16 with baked VAE version?
Dude, please, make a SDXL inpaint NSFW checkpoint!
Honestly, it's a very nice model. I like the results with fast generation
This is now my main turbo model :)
This has become my absolute no. 1 go-to SDXL model for almost everything. Thank you for the awesome work! 🙏
What's your process for upscaling, currently I seem to be losing detail in the upscale. Does TURBO change the upscaling process we should use? Would you mind sharing your workflow (if your using comfyui)?...... PS I love ultraspice! :)
This model is truly amazing.
Until now, I've viewed Turbo models as a cheap shortcut with a compromise on quality. This checkpoint has changed that ignorant opinion. The quality is simply amazing.
This model is excellent, it's the best SDXL series model I've encountered on Civitai so far, and it has great quality with minimal restrictions. Keep up the good work!
Excellent model. I have a simple question:
Any way to reduce the number of steps needed while keeping the details? Loving the model so far, but if I reduce the steps too much I lose detail and faces became mangled. My current setup (dpmpp sde + karras) 4 setps, cfg 2 produces bad faces that are sometimes mangled. This is slightly faster than my sd1.5 setup.
Thank you so much! I think this is my main now
Could you list what checkpoints and loras you've merged into this checkpoint?
The pictures with 1.5 look sooo good! Again even better than 1.3 and 1.4. I am having some problems with hands though. Not a big deal if you create enough pictures. ;-)
This is the best of all sdxl, i tested allll them. I hope he make a inpaint version
I love this checkpoint, and it just keeps getting better. Love how you can push it from realism, to vibrant, to illustration and it all looks fantastic.
Absolutely fantastic model!!! Don't try to stop!!!
What are the setting recommendations for hires fix? Want to try to push some images to the next level.
I've fallen in love with this model, it does everything I want for now and I keep wanting to try new things!
Question... Can you enable the onsite gen doodah please?
Great and versatile model, from art, fantasy to realistic photos....
Bit sensitive of CFG, so .5 can make dramatic change, but, it is fun to experiment. Excellent!
It looks good
Upgraded from 1.5 to 1.8 and, finally, the model does a good job on animal fur (I mainly do photorealistic animal stuff). The only drawback is that many animals still are created with surplus legs. Models such as Dreamshaper or Juggernaut do better in that respect, so there is still room for improvement. Apart from that, Ultraspice is great and has become one of my favorite models.
For commercial purpose,
Is there any way you can make a NON-TURBO SDXL model version of this?
This is the best model I know and having issues finding anything comparable that is only SDXL (not turbo).
I'm going to guess v1.9 is not SD 1.5 as it says on the base model, because it's not loading for me.
I noticed a tremendous improvement in how it's rendering hands in v1.9. Great job!
Some prompts require CFG of 1.0 to look normal.
Other prompts a CFG of 2.5 is fine and looks normal.
v1.9 does have better results than v2,
but I'm loving the models and work into them. i would back the sharpens off as sharp image is trigger prompt on all models and its harder to get a softer image, ie soft image is not a trigger and needed slightly blurred that costs details.
i like most of the versions of this ckpt!! i think it handled lower steps and prompt guidance really well, and it handled the prompts well whether i had prompt enhance on or not and it worked well with the creativity feature that mage added. it also handled multiple people in images quite well, with only minor issues. good job on the ckpt versions!!
EULA ancestral sgm uniform is best,
karras is my favorite but due to sharp training can not be used, I'd recommend softer training, as mentioned before can use sharp image in prompt to make sharp images, but soft images its not possible.
For releasing a new version like every 5 seconds so that poor old me can't keep up, you only get cat's now... which wasn't planned at all... kinda.
'this is the coolest model, this is exactly what I was looking for, thanks a lot to the author!!!!!
hard to keep up with the versions :o
Yo, just wanna let you know that i uploaded your model onto Pixai with due credits:
AI Art Model: Ultraspice V 2.3 | PixAI
If you wish for me to remove it then feel free to message me, alternatively you can create an account there and claim it for yourself.
Tried ultraspice for the first time a few days ago (v2.3) and I am really impressed.
I've never found an SDXL model that runs so smoothly and quickly without hiccups.
Amazing model for random short prompts giving fantastic images. Prompt like you're high and you'll be amazed. Does stunning faces if you ease off on the CFG