Halcyon Drift

Model Name: Halcyon Drift — Desert World Series

Trigger word: Halcyon Drift

Base model: Krea2-Turbo (Flux-based)

What this set is:
Three photorealistic sci-fi scenes sharing a consistent desert/dune-world setting: robed travelers on foot, a large starship as the visual anchor of each frame, and warm dusk-to-golden-hour lighting throughout. The three images cover different moments of the same world — a dramatic vertical launch from a marketplace, two figures walking toward a landed ship across open grassland, and a coastal spired city with a starship hovering above the harbor.

Method used to build the captions:

1. Each image was written as one full sentence rather than a tag list, in the order: lighting → foreground subject(s) with clothing/texture detail → ship or main sci-fi element with material/mechanical detail → background/setting → atmosphere → camera/lens specs → style tags.

2. Consistent visual language was used across all three — "robed travelers," "warm dusk light," "starship" — so the LoRA learns a coherent world rather than three unrelated scenes.

3. The trigger word Halcyon Drift opens each caption as "A Halcyon Drift scene, cinematic photograph..." so it's learned as a scene-level descriptor rather than a disconnected tag.

Files:

• image.jpg + matching image.txt per pair — standard LoRA training format (Kohya/OneTrainer compatible)