Formerly Turbo Diffusion XL
v1.91
It seems the problem was ZavyChromaXL..
Primary: Moxie Diffusion XL v1.8
Secondary: Zaxious_XL v4.0
v1.9
I am fairly adamant to use only trained checkpoints when creating these checkpoints. I was excited to see with Zaxious now has a trained checkpoint. I'm pretty curious to play more with this. So far, I'm liking this version quite a bit.
Primary: Moxie Diffusion XL v1.8
Secondary: Zaxious_XL v4.0
v1.8
I have typically liked ProtoVision and DynaVision more, and used NightVision sparingly. However, lately I like NightVision the best. It seemed like time to incorporate it since there were new releases of NightVision and Zavy (Zavy being one of my long time favorites).
Primary: Moxie Diffusion XL v1.7
Secondary: NightVision XL v9.0.0
Tertiary: ZavyChromaXL v8.0
v1.7
Fully SDXL version...
Primary: ZavyChromaXL v7.0
Secondary: RealVisXL v4.0
Tertiary: ProtoVision XL v6.6.0
Lightning Versions
v1.6
Crystal clear has been in my top 3 almost continuously since I found it. I was excited when I saw they posted a Lightning version.
Primary: Moxie Diffusion XL v1.51
Secondary: Crystal Clear Lightning v1.0
v1.51
This is a low percentage merge. I don't find a large difference between 1.50 and v1.51. However, the comparison images seem to be a little sharper and more realistic. I decided to do this merge after playing with the donor checkpoints.
Primary: Moxie Diffusion XL v1.50
Secondary: AI Real Lightning v2.0
Tertiary: epic Realism XL - Zeus
v1.5
As I have been playing more with v1.4, I am finding that I am not as happy with the realism as I initially thought. I resetting it by merging checkpoints of some of my favorite trained checkpoints.
Primary: Juggernaut v9 RDPhoto2-Lightning_4S
Secondary: DreamShaperXL Lightning DPM++ SDE
Tertiary: Copax Timeless v11 Lightning
As I have been playing more with v1.4, I am finding that I am not as happy with the realism as I initially thought. I resetting it by merging checkpoints of some of my favorite trained checkpoints.
Primary: Juggernaut v9 RDPhoto2-Lightning_4S
Secondary: DreamShaper XL vLightning
Tertiary: Copax Timeless v11 Lightning
v1.4
Primary: Turbo Diffusion XL v1.32 (now Moxie Diffusion XL)
Secondary: Copax Timeless v11 Lightning
Tertiary: RealVisXL v4.0
v1.32
This version is a complete ground up re-make using SDXL Lightning checkpoints.
v1.30 unpublished
Primary: Juggernaut v9 RDPhoto2-Lightning_4S (unpublished)
Secondary: WildcardX-XL Lightning
Tertiary: Blue Pencil XL v5.0
v1.31
Primary: Turbo Lightning v1.30 (unpublished)
Secondary: RealVisXL v4.0
Tertiary: PixelWaveLightning v01
v1.32
Primary: Turbo Lightning v1.31 (unpublished)
Secondary: DreamShaper XL vLightning
This has a baked in VAE.
Sampling Method: DPM++ SDE Karras
Sampling steps: 7
CFG Scale: 1
Clip Skip: 1 or 2
Turbo Versions
v1.2
The anime in this version seems to be more of a realism anime.
Primary: Turbo Diffusion XL Turbo v1.12 (this checkpoint)
Secondary: DreamShaperXL v2 Turbo DPMpp SDE
Tertiary: Hyperreal Surrealist - Turbo Edition v1.0
v1.12
A re-merge of 1.0 using Vibrant 2.0 instead of 1.0. Everything else remains the same.
Primary: Turbo Diffusion XL Turbo v1.0 (this checkpoint)
Secondary: Vibrant Horizon - Turbo XL v2.0
Tertiary: Sleipnir
v1.1
Primary: Turbo Diffusion XL Turbo v1.0 (this checkpoint)
Secondary: Vibrant Horizon - Turbo XL v1.0
Tertiary: Sleipnir
v1.0
This checkpoint is a Stable-Diffusion merge between the follow 3 SDXL Turbo models.
Primary: DreamShaper XL
Secondary: Artium
Tertiary: Vibrant Horizon - Turbo XL
This has a baked in VAE.
Sampling Method: DPM++ SDE Karras
Sampling steps: 8
CFG Scale: 2 > 3.5
Clip Skip: 1 or 2
Description
v1.2
A re-merge of 1.0 using Vibrant 2.0 instead of 1.0. Everything else remains the same.
Primary: Turbo Diffusion XL Turbo v1.12 (this checkpoint)
Secondary: DreamShaperXL v2 Turbo DPMpp SDE
Tertiary: Hyperreal Surrealist - Turbo Edition v1.0
FAQ
Comments (11)
I'm having a very difficult (impossible?) time getting this checkpoint to generate full body images, it really wants to focus on either the face or the upper body, waist and up.
Why don't you post your prompt to this comment, and let me try it?
Positive: extreme detail, (standing in a photography studio), photorealistic photograph of girl posing, wearing a pink sweetegl lolita dress with white lace trim and floral embroidery, very full skirt, (extremely tight restrictive fitted bodice), detailed face and eyes, ((detailed tights and (shoes))), (professional studio lighting), (full body shot:1.6), <lora:sweetlolitaxlv4:1.0>
Negative prompt: ugly, deformed, a featureless color shader, 3d max, anime, cartoon, upper body only, cowboy, close up
@moxie1776 Posted. I did find that I was able to use controlnet to get full body shots, but the regular way I'm used to doing it by simply calling (full body shot) in the positive prompt wasn't working for me. Could be the sampler I suppose, I think I was using DPM++ 3M SDE Karras.
@MysticDaedra make full body or full shot your second or third prompt.
what i found out is, "detailed eyes" makes it much harder to get full shots, and the lora could also be a reason
@diaserikcg114 I have seen this, but never tried it. I think I will check it out :)
Can you please release SFW version. Thanks a lot
I apologize, the way I do these checkpoints doesn't give me control over sfw/nsfw. If you find issues with it generating nswf images, you should be able to add nsfw to the negative prompt to alleviate the problem.
I don't generate nsfw images, and while I don't test for that, I don't think I have ever had this checkpoint generate an nsfw image. I have no idea if it is sfw or nsfw. :/
@moxie1776 Thanks for reply. Great model.