Vani — a consistent young South Asian / Indian female character LoRA for Flux1‑D.

Trained on 50 captioned reference photos with:

• Trainer: ostris/flux-dev-lora-trainer

• Steps: 2000

• LoRA rank: 32

• Learning rate: 4e-4

• Optimizer: adamw8bit

• Batch size: 1

• Resolutions: 512, 768, 1024

• Caption dropout: 0.05

• Final loss: ~0.30 (clean convergence)

Trigger word: ananya

Recommended settings:

• Base: Flux.1 D

• CFG: 3.5

• Steps: 28

• LoRA strength: 1.0

• Sampler: Euler

Good at: consistent identity across studio portraits, casual outdoor shots, indoor candids, close-ups, and lifestyle photography.

Not intended for: extreme stylization or non-photorealistic outputs.