madmasyhur/sd-batik-dreambooth

SD Batik DreamBooth - Indonesian Traditional Patterns

This is a fine-tuned Stable Diffusion v1.5 model trained using DreamBooth technique on 20 traditional Indonesian batik motifs. The model can generate authentic Indonesian batik patterns using specific trigger tokens for each of the 20 different motifs.

Model Description

This model was trained to understand and generate traditional Indonesian batik patterns, preserving the cultural heritage and artistic characteristics of each specific motif. Each batik style has been carefully learned through the DreamBooth fine-tuning process.

Usage

from diffusers import StableDiffusionPipeline
import torch

# Load the model
pipe = StableDiffusionPipeline.from_pretrained(
    "madmasyhur/sd-batik-dreambooth",
    torch_dtype=torch.float16,
    safety_checker=None,
    requires_safety_checker=False
)
pipe = pipe.to("cuda")

# Generate batik patterns
# use both unique token and motif name for better result 
prompt = "a photo of balibtk batik bali pattern, traditional indonesian textile, high quality, detailed"
negative_prompt = "blurry, low quality, distorted, text, watermark"

image = pipe(
    prompt,
    negative_prompt=negative_prompt,
    num_inference_steps=50,
    guidance_scale=7.5,
    height=512,
    width=512
).images[0]

image.save("batik_pattern.png")

Supported Batik Motifs & Tokens

Motif	Token	Region/Style	Description
Bali	balibtk	Balinese traditional	Traditional Balinese ornamental patterns
Betawi	betbtk	Jakarta traditional	Jakarta's indigenous cultural motifs
Cendrawasih	cenbtk	Papua bird motif	Bird of paradise inspired patterns
Ceplok	cepbtk	Central Java geometric	Geometric circular and floral motifs
Dayak	daybtk	Kalimantan traditional	Borneo indigenous tribal patterns
Garutan	garbtk	Garut regional style	West Java regional motifs
Gentongan	genbtk	Traditional dyeing technique	Classic dyeing method patterns
Jumputan	jumbtk	Tie-dye technique	Traditional tie-dye resist patterns
Kawung	kawbtk	Yogyakarta classic	Four-circle geometric royal pattern
Keraton	kerbtk	Royal court style	Noble palace traditional designs
Lasem	lasbtk	Chinese-Javanese fusion	Chinese-influenced coastal patterns
Megamendung	mgmdbtk	Cirebon cloud pattern	Stylized cloud and sky motifs
Nitik	nitbtk	Fine dot pattern	Delicate small dot decorations
Parang	parbtk	Diagonal knife pattern	Diagonal sword/blade motifs
Sasirangan	sasbtk	Banjarmasin traditional	South Kalimantan traditional patterns
Sekar	sekbtk	Floral pattern	Various flower and plant motifs
Sidoluhur	sidlbtk	Javanese noble pattern	High-status traditional design
Sidomukti	sidmbtk	Javanese prosperity pattern	Prosperity and fortune symbols
Sogan	sogbtk	Brown natural dye	Natural brown-colored traditional dye
Tambal	tambtk	Patchwork pattern	Mixed patchwork design motifs

Example Prompts

Classic Patterns

"a photo of kawbtk batik kawung pattern, traditional indonesian textile, geometric design, high quality"
"a photo of parbtk batik parang pattern, diagonal lines, traditional javanese, detailed, masterpiece"
"a photo of kerbtk batik keraton pattern, royal court style, intricate details, traditional"

Regional Styles

"a photo of mgmdbtk batik megamendung pattern, cloud motif, cirebon style, intricate details"
"a photo of cenbtk batik cendrawasih pattern, bird of paradise, papua traditional, vibrant colors"
"a photo of daybtk batik dayak pattern, tribal design, kalimantan traditional, authentic"

Natural Motifs

"a photo of sekbtk batik sekar pattern, floral design, traditional indonesian, beautiful flowers"
"a photo of sogbtk batik sogan pattern, natural brown dye, traditional coloring, organic texture"
"a photo of lasbtk batik lasem pattern, coastal design, chinese javanese fusion, elegant"

Technique-based

"a photo of jumbtk batik jumputan pattern, tie dye technique, traditional resist dyeing, colorful"
"a photo of genbtk batik gentongan pattern, traditional dyeing method, authentic technique, detailed"
"a photo of nitbtk batik nitik pattern, fine dots, delicate pattern, traditional craftsmanship"

Training Details

• Base Model: izzudd/sd-batik-llava
• Training Method: DreamBooth
• Training Steps: 8000
• Learning Rate: 2e-6
• Training Resolution: 256x256
• Inference Resolution: 512x512
• Dataset: 20 traditional Indonesian batik motifs
• Mixed Precision: no
• Optimizer: AdamW

Model Performance

The model has been trained to maintain the authentic characteristics of each batik motif while being capable of generating variations and combinations. The training process focused on:

• Preserving traditional pattern structures
• Maintaining cultural authenticity
• Generating high-quality textile textures
• Understanding motif-specific characteristics

Cultural Considerations

This model represents traditional Indonesian cultural heritage. Please use it respectfully and acknowledge the cultural significance of batik patterns in Indonesian society. Batik is recognized by UNESCO as a Masterpiece of Oral and Intangible Heritage of Humanity.

Technical Notes

• Best results with guidance scale 7.5-12.0
• Recommended inference steps: 50-100
• Works well with negative prompts to avoid unwanted artifacts
• Can be combined with other prompts for creative variations

Citation

If you use this model in your research or creative work, please consider citing:

@misc{sd-batik-dreambooth,
  title={SD Batik DreamBooth - Indonesian Traditional Patterns},
  author={madmasyhur},
  year={2025},
  url={https://huggingface.co/madmasyhur/sd-batik-dreambooth}
}

Acknowledgments

• Traditional batik artisans and cultural heritage preservationists
• Indonesian Ministry of Culture for cultural documentation
• Stability AI for the base Stable Diffusion model
• HuggingFace for the diffusion pipeline infrastructure