Model Exclusive for image generaion?
Hi there,
I’m curious whether you have (or plan to release) a model focused exclusively on image generation, specifically for Qwen Image 2509 or 2512, separated from the edit/inpainting model.
I’ve been testing several models available on CivitAI, and so far none of them come close to the prompt adherence and consistency I see in your work.
Most models struggle with:
Accurate prompt following
Consistent variation in human height
Body proportions and overall anatomy diversity
This made me wonder if separating concerns could help:
a pure image-generation model, independent from the editor, might allow better specialization and potentially deliver the best of both worlds — high-fidelity generation plus a strong editing pipeline.
I’d really appreciate your thoughts on whether this approach makes sense from a training and architectural standpoint.
Thanks in advance, and sincere congratulations on the quality of your work — it’s genuinely impressive.
Nope. Edit only. Being able to input images as part of workflows it just too useful of a feature to do without. Why have an image generation model that only takes text when you can have it take text and optionally images?
you can give a blank or a solid colour image as input and describe the generation in prompt to be able to create standalone images, although I'm not one to guarantee the quality of that compared to using qwen image