New Models Ideal For Low-VRAM GPUs (Patreon)

Video - https://youtu.be/5Wae2H3ures

There have been a few new models recently, so I figured I'd take a look at what they have to offer!

1. Probably the best of them is Lumina 2.0 which features some impressive prompt comprehension in a 2B model. The styles are a bit hit and miss, but at least it has some ;)

Example Lumina 2.0 image

2. TinyBreaker is a prototype with a model weight of just 3 GB, mixing both Pixart and SD. Being Pixart based it's great at more "rough" looking things such as sketches, drawings and certain painting styles. Super fast too!

Example TinyBreaker image

3. Animagine XL 4 provides both base and optimised models fine-tuned from SDXL to provide anime style images. Being SDXL, you can use all the usual bells and whistles such as WaveSpeed.

Personally, I've been liking the crisp and clean outputs from the Euler Ancestral Dancing sampler. If you've not got the extra samplers installed already, you can install via manager. Check out https://github.com/Clybius/ComfyUI-Extra-Samplers for more information. For more information on WaveSpeed, see https://www.patreon.com/posts/119802818

Example Animagine image

Smaller can mean faster too, so even higher-end GPU owners can benefit :)

Links

Links to the various models shown in the video:

Lumina 2.0 Github - https://github.com/Alpha-VLLM/Lumina-Image-2.0

Comfy Example workflow - https://comfyanonymous.github.io/ComfyUI_examples/lumina2/

Repackaged Lumina Files - https://huggingface.co/Comfy-Org/Lumina_Image_2.0_Repackaged/tree/main

Tiny Breaker - https://github.com/martin-rizzo/ComfyUI-TinyBreaker

Animagine XL 4 -https://huggingface.co/cagliostrolab/animagine-xl-4.0

Workflows

Workflows are numbered in order the appear in the video. 1 is Lumina, 2 is TinyBreaker and 3 is Animagine.