Like the work I do and want to say thanks? Buy me a coffee or Support me on Patreon for exclusive early access to my models and more!
DynaVision XL was born from a merge of my NightVision XL model and several fantastic LORAs including Sameritan's wonderful 3D Cartoon LORA and the Wowifier LORA, to create a model that produces stylized 3D model output similar to computer graphics animation like Pixar, Dreamworks, Disney Studios, Nickelodeon, etc.
DynaVision XL is capable of both SFW and NSFW output. As with all of my other models, tools and embeddings, DynaVision XL is easy to use, preferring simple prompts and letting the model do the heavy lifting for scene building.
WARNING - DO NOT USE SDXL REFINER WITH DYNAVISION XL
The SDXL refiner is incompatible and you will have reduced quality output if you try to use the base model refiner with DynaVision XL.
Follow me here by clicking the heart ❤️ and liking the model 👍, and you will be notified of any future versions I release. I also need your help with feedback, please please please post your images and your honest feedback below, I will use your feedback and your output to help guide future revisions!
Description
Did a rebuild and rebalance to try to better improve eyes / teeth / hands, and swapped in some updated LORAs in the process. New Alpha3 is a little different than the previous version, but overall I think it has improvements to the above and better lighting. Also seems to be more face varied. I will post the changelog to the comments on the main DynaVisionXL listing. Enjoy!
FAQ
Comments (20)
ALPHA 0.3.6 CHANGELOG 08/08/23
I'm probably going to switch to beta releases here soon and slow down a bit, as I'm liking where the style is right now. I may still need to do one more "ground up" rebuild to remove Wowifier from the merge, as it continues to sometimes cause issues with armor and robotic arms showing up occasionally unprompted. - Please give me your feedback here or on Threads or Discord, I take all feedback seriously!
Changes
- This is a breakdown and rebuild of the formulas from Alpha 0.2, so I could update LORA versions.
- - Reduced the strength of Wowifier XL, as it was causing robotic arms and armor to appear unprompted much too frequently
- - Increased the strength and updated version of SamDoesArt SDXL LORA to compensate
Known Issues
- Still seeing armor pop up occasionally, though much less often than before. I'm considering removing the Wowifier XL lora completely if it continues to be an issue.
- Hands are still hit or miss, though they're starting to become more hit than miss thankfully.
- Eyes at a distance are mostly fixed now, but can still sometimes look wonky.
- pupil and iris shapes can still sometimes be off.
@socalguitarist Fantastic model, lots of fun! Is there any way to train the model to handle 2:3 aspect ratios better? I'm struggling to generate images at 1024x1536 - only one out of every three images is usable, the others are stretched or have duplicated features. Don't get me wrong, I love a girl with four eyes, but not when it's not in my prompt! ❤️
@zer0TF Ah, if you want to do 2:3, run it at 896 x 1152 - that's a native resolution for SDXL and you should get minimal goofiness, then you can upscale it in img2img/extras to whatever final reso you like.
Here are the native resolutions SDXL supports. While it's definitely possible to train at that resolution, in the meantime, try using these:
1024, 1024 # 1:1 square
1152, 896 # 3:2 Landscape
896, 1152 # 2:3 Photo
1216, 832 # 5:4 PanNScan
832, 1216 # 4:5 Portrait
1344, 768 # 16:9 Widescreen
768, 1344 # 9:16 Vertical
1536, 640 # 21:9 Cinematic
640, 1536 # 9:21 VerticalSlice
@socalguitarist Thank you! 😘 I will give these a try!
Close up effect even stronger than in the previous version.
After comparing some images side by side ... the old alpha 020 is the clear winner (at least imho).
No doubt that Dynavison XL is far ahead compared to all the other XL models and so is the new version ... but for now I'll keep both versions ;-).
Thanks for sharing your amazing creation with us.
amazing~!!!!
damn, feeling like i found revAnimated all over again!
very nice. i may have to run my whole library through this model... :) interesting style. thanks for posting.
It gives me this error if I try to generate: RuntimeError: All tensors were expected to be on the same device, but at least two devices were found, cpu and cuda:0! (while checking the index argument in the wrapper_CUDA__index_select method).
Most likely you tried to generate when something was already going on, sometimes it does that. Simply restart your session. (If in colab then restart runtime and runall.)
Beautiful images. Extremely well done - great work. However, can't seem to force it to create a full-body nude no matter what.
Needs more pixar pussy in the training set. 🍑
New update coming soon that will help fix this - stay tuned 😉
这是XL本身的特点之一，全身=半身，加权重几乎无效
try adding (((nicely shaped detailed legs))) and (((high heels))) to the prompt
i posted some platinum blonde examples for your viewing pleasure. i get about 2-3 full body shots in 9. enjoy
@tedbiv
@wyxzddsjj919 glad it worked for you... :)
@wyxzddsjj919 i get about 6-8 out of 9 full body images with this prompt with latest version of model. they really are nice...
raw photo of a fully naked 20 yr old stunningly beautiful 1girl, standing dance pose, hands in the air, fully nude, standing on a pedestal, full body portrait, nice shapely legs, dark tan, black high heels, ankle bracelet, perky large breasts, erect nipples, detailed long realistic hair, brown eyes, detailed realistic skin, ultra sharp focus, high contrast, 8k, 4k, hd
neg = teeth, clothes, bra, panties, body cover, pants, shorts, shirt,
Details
Files
dynavisionXLAllInOneStylized_alpha036FP16Bakedvae.safetensors
Mirrors
dynavisionXLAllInOneStylized_alpha036FP16Bakedvae.safetensors
dynavisionXLAllInOneStylized_alpha036FP16Bakedvae.safetensors
dynavisionXLAllInOneStylized_alpha036FP16Bakedvae.safetensors
dynavisionXLAllInOneStylized_alpha036FP16Bakedvae.safetensors
dynavisionXLAllInOneStylized_alpha036FP16Bakedvae.safetensors
dynavisionXLAllInOneStylized_alpha036FP16Bakedvae.safetensors
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.