<h1 id="midcentury-dream-qwen">Midcentury Dream Qwen </h1>Midcentury Dream Qwen is a Qwen Image LoRA trained on hundreds of cropped photos and illustrations from real magazine scans from the 1950s and 1960s. The source images were largely high quality. This LoRA is a little constrained by Qwen itself (I find it a little soft compared to other models), but it still might come in handy for those willing to put in a little effort with their prompts. In my testing it seems to work well with both photo and illustration prompts. If you get grainy images try adding quality keywords like "DSLR" and such, it seems that Qwen is suggestible in that way, even when prompting for illustrations it seemed to help avoid grainy and noisy images.The gallery has some example workflows, please change the output directories and number of samplers to your liking, this was a quick and dirty workflow test.

Midcentury Dream Qwen

Trained for 1500 steps instead of 2500, less of the style but maybe more stable for wider range of uses.

Files with SHA256: 776657778f6aaffebdae2180aae29522c2b0a138b4fcd89562c63a82e5c426c1