Waiting for new models to train LoRA components.

Waiting for new models to train LoRA components.

I expect major developments in LoRA training systems in the future; in the meantime, I’m posting my images here (even though the current component is for Z Immage Turbo).

My images are generated with MindCraft Studio for Mac > https://themindstudio.cc/mindcraft

Qwen

Qwen-Image-2512 (6-bit) - QWEN - (~31GB) Strong prompt understanding and world knowledge.

Qwen-Image 2.1 (4-bit) - QWEN 2.1 - (~11.6GB) Bilingual prompts, native transparent (RGBA) output, and in-context editing with up to 10 reference images.

https://civitai.com/models/2711540/v67-qwen-image

https://civitai.red/models/2711540/v67-qwen-image

Ideogram 4

Ideogram 4 (FP8) - IDEOGRAM - (~27GB) Extremely high-quality text-to-image model with native spelling and graphic design capabil

https://civitai.com/models/2790003/v67-ideogram-4

https://civitai.red/models/2790003/v67-ideogram-4

Lens Turbo

Lens Turbo (6-bit) - LENS - (~17GB) 4-step text-to-image model from Microsoft

https://civitai.com/models/2710942/v67-lens-turbo

https://civitai.red/models/2710942/v67-lens-turbo

MindCraft Studio for Mac

It is a tool of mine that updates very frequently and is constantly adding new models—keeping pace with official releases—and it will soon replace my Grok subscription once it gains video generation capabilities.

Angelo Maiota