Lumina 2.0 is a solid base model that fulfills our expectations for SD3/3.5. It’s truly open-source with an Apache 2.0 license, which is a significant advantage.
In terms of comparison with Flux, Lumina 2 excels in two key areas: it’s better at generating illustrations and concept understanding. For instance, it’s the only model that comprehends the concept of Left vs. Right.
It’s Apache 2-licensed model. Flux and Stable Diffusion, while claiming to be open-source, are not truly free. Additionally, Nvidia’s SANA was recently relicensed as Apache 2, which could potentially create a robust ecosystem. Community should prioritize supporting Lumina and SANA since they are genuinely free. Flux, on the other hand, is a deceptive model that we can never truly own. Unfortunately, video models are all burdened by restrictive licenses, such as LTX-1 and Hunyuan. Hopefully, when one of these models is released with an Apache 2 license, the rest of the ecosystem will be compelled to adapt.
https://github.com/Alpha-VLLM/Lumina-Image-2.0?tab=readme-ov-file