TRELLIS 2 just dropped
https://github.com/microsoft/TRELLIS.2
From my experience so far, it can't compete with Hunyuan 3.0, but it gives a nice run for the money for all the other closed-source models.
It's definitely the #1 open source model at the moment.
https://redd.it/1podq9a
@rStableDiffusion
https://github.com/microsoft/TRELLIS.2
From my experience so far, it can't compete with Hunyuan 3.0, but it gives a nice run for the money for all the other closed-source models.
It's definitely the #1 open source model at the moment.
https://redd.it/1podq9a
@rStableDiffusion
GitHub
GitHub - microsoft/TRELLIS.2: Native and Compact Structured Latents for 3D Generation
Native and Compact Structured Latents for 3D Generation - microsoft/TRELLIS.2
After a couple of months learning I can finally be proud of to share my first decent cat generation. Also first one to compare.
https://redd.it/1pocq6e
@rStableDiffusion
https://redd.it/1pocq6e
@rStableDiffusion
Difference between ai-toolkit training previews and ComfyUI inference (Z-Image)
https://redd.it/1pofru6
@rStableDiffusion
https://redd.it/1pofru6
@rStableDiffusion
Z-Image-Turbo-Fun-Controlnet-Union-2.1 available now
2.1 is faster than 2.0 because of a bug in 2.0.
Ran a quick comparison using depth and 1024x1024 output:
2.0: 100%|██████| 15/15 [00:09<00:00, 1.54it/s\]
2.1: 100%|██████| 15/15 [00:07<00:00, 2.09it/s\]
https://huggingface.co/alibaba-pai/Z-Image-Turbo-Fun-Controlnet-Union-2.0/tree/main
https://redd.it/1pot3yw
@rStableDiffusion
2.1 is faster than 2.0 because of a bug in 2.0.
Ran a quick comparison using depth and 1024x1024 output:
2.0: 100%|██████| 15/15 [00:09<00:00, 1.54it/s\]
2.1: 100%|██████| 15/15 [00:07<00:00, 2.09it/s\]
https://huggingface.co/alibaba-pai/Z-Image-Turbo-Fun-Controlnet-Union-2.0/tree/main
https://redd.it/1pot3yw
@rStableDiffusion
huggingface.co
alibaba-pai/Z-Image-Turbo-Fun-Controlnet-Union-2.0 at main
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
Apple drops a paper on how to speed up image gen without retraining the model from scratch. Does anyone knowledgeable know if this truly a leap compared to stuff we use now like lightning Loras etc
https://x.com/jiqizhixin/status/2001132022971105335?s=19l
https://redd.it/1pov7ab
@rStableDiffusion
https://x.com/jiqizhixin/status/2001132022971105335?s=19l
https://redd.it/1pov7ab
@rStableDiffusion
X (formerly Twitter)
机器之心 JIQIZHIXIN (@jiqizhixin) on X
What if you could speed up a cutting-edge AI image generator without retraining it from scratch?
Apple presents a new method that unifies two major AI image models (diffusion & flow matching) under a simple statistical rule. This lets them directly apply…
Apple presents a new method that unifies two major AI image models (diffusion & flow matching) under a simple statistical rule. This lets them directly apply…