Z Image Base LoRA Trainer

Z Image Base LoRA Trainer

Playground

Z-Image Base LoRA Trainer – train custom image LoRA models from your own dataset, with zip uploads, auto-tuned defaults and fast iteration for brand, character or IP looks. Ready-to-use REST inference API, best performance, no cold starts, affordable pricing.

Features

Z-Image Base LoRA Trainer

Z-Image Base LoRA Trainer is a high-performance custom model training service for the Z-Image text-to-image generation model. It allows you to train lightweight LoRA (Low-Rank Adaptation) adapters for personalized styles, characters, and concepts — bringing your custom visuals into AI-generated images.

Why Choose This?

• Efficient training Train custom adapters specifically optimized for Z-Image’s fast diffusion architecture.

• Compact and portable Produces lightweight LoRA files that are easy to share and deploy.

• Plug-and-play compatibility Trained LoRAs work directly with Z-Image Base LoRA and Z-Image Turbo LoRA models.

• Preserves base model speed Your custom styles inherit Z-Image’s fast generation capabilities.

Training Process

1. Data Upload Prepare and upload a ZIP file containing your training images. Include 10-20 high-quality, diverse images for best results.

2. Configure Trigger Word Set a unique trigger word (e.g., “p3r5on”) that will activate your trained style or character in prompts.

3. Adjust Training Parameters

   • steps — Total training iterations (default: 1000)
   • learning_rate — Training speed (default: 0.0001)
   • lora_rank — Adapter capacity (default: 16)

4. LoRA Training The system runs a tailored LoRA optimization loop:

   • Freezes the base model weights
   • Trains only the low-rank adapter layers
   • Applies Z-Image optimized settings

5. Model Export After training completes, you receive a LoRA adapter file (.safetensors) compatible with:

   • Z-Image Base LoRA
   • Z-Image Turbo LoRA

Parameters

How to Use

1. Prepare your images — collect 10-20 high-quality, diverse images of your subject.
2. Create a ZIP file — package all images into a single ZIP archive.
3. Upload your data — drag and drop or provide a public URL to your ZIP file.
4. Set trigger word — choose a unique word that won’t conflict with common terms.
5. Adjust parameters (optional) — modify steps, learning_rate, and lora_rank as needed.
6. Run — submit and wait for training to complete.
7. Download — receive your LoRA adapter file for use with Z-Image models.

Pricing

Billing Rules

• Base price: $1.25 per 1,000 steps
• Total cost = $1.25 × (steps / 1,000)

Best Use Cases

• Character LoRAs — Train on character images to maintain identity across generations.
• Brand Styles — Create custom visual styles for consistent marketing imagery.
• Art Styles — Capture specific artistic aesthetics for creative projects.
• Product Photography — Train on product photos for consistent visual presentations.

Pro Tips

• Use 10-20 high-quality, diverse images of your subject for best results.
• Choose a unique trigger word that won’t conflict with common words (e.g., “m1style” instead of “style”).
• Higher lora_rank (32-64) captures more detail but increases training time and file size.
• Lower learning_rate (0.00005) is more stable but requires more steps.
• Start with default settings, then adjust if needed.

Try Your LoRA

After training, use your LoRA with these models:

• Z-Image Base LoRA — Base model with LoRA support.
• Z-Image Turbo LoRA — Fast generation with LoRA support.

Guidance

• How to Train your LoRA?
• How to Use your LoRA?

Notes

• Minimum recommended: 4 images, optimal: 10-20 images.
• Training time scales with the number of steps configured.
• Higher parameter values (steps, lora_rank) will increase training time and cost.
• For faster iterations, start with lower settings and increase gradually.

# Submit the task
curl --location --request POST "https://api.wavespeed.ai/api/v3/wavespeed-ai/z-image/base-lora-trainer" \
--header "Content-Type: application/json" \
--header "Authorization: Bearer ${WAVESPEED_API_KEY}" \
--data-raw '{
    "trigger_word": "p3r5on",
    "steps": 1000,
    "learning_rate": 0.0001,
    "lora_rank": 16
}'

# Get the result
curl --location --request GET "https://api.wavespeed.ai/api/v3/predictions/${requestId}/result" \
--header "Authorization: Bearer ${WAVESPEED_API_KEY}"