CivArchive
    Flower lora for sd 15 - v1.0
    NSFW
    Preview 145439219

    🌸 Botanical Flora & Macro Detail LoRA (SD 1.5)

    A lightweight, fine-tuned SD 1.5 LoRA specialized in high-detail macro botanical photography, delicate petal textures, intricate stamen geometry, and vibrant organic coloration.

    Trained on a curated, high-resolution subset of the Oxford Flowers BLIP-captioned dataset with natural language captions to preserve prompt flexibility without burning out the latent space.

    🔍 Weight Scaling & Comparison

    Recommended weight: 0.75 – 0.85 for the best balance of natural petal anatomy and prompt flexibility.

    • 0.0: Baseline SD 1.5 (often lacks macro focus and natural leaf venation).

    • 0.5 – 0.85 (Sweet Spot): Crisp stamen structures, clean lighting, sharp radial symmetry, and vibrant pigmentation.

    • 1.0+: Highly stylized macro saturation with intense floral features.

    • Base Model: Stable Diffusion v1.5

    • LoRA Weight: 0.7 to 0.9 (sweet spot: 0.8)

    • Sampler: DPM++ 2M Karras or Euler a

    • Sampling Steps: 25 – 35

    • CFG Scale: 6.5 – 7.5

    • Clip Skip: 1

    💡 Example Prompts

    Macro Flora:

    macro close up photo of an exotic flower, translucent petals, intricate pollen stamen, dew droplets, sharp botanical photography, soft natural studio lighting, centered, 8k depth of field

    Bioluminescent / Fantasy Concept:

    macro shot of a deep-sea bioluminescent flowering plant, glowing translucent petals, delicate crystal stamen filaments, dark oceanic background, vibrant contrast, ultra detailed

    Negative Prompt:

    blurry, low quality, oversaturated, deformed, plastic look, digital render, noisy, artifacts, bad anatomy

    be sure to share youre results ;)

    Description

    initial release

    LORA
    SD 1.5

    Details

    Downloads
    8
    Platform
    CivitAI
    Platform Status
    Available
    Created
    10/12/2026
    Updated
    10/12/2026
    Deleted
    -

    Files

    pytorch_lora_weights.safetensors

    Mirrors