Z-Image Base LoRA Trainer: Lora Support

Train custom LoRA adapters for the Z-Image Base checkpoint from a zip dataset. Auto-tuned defaults, fast iteration, pay-per-step pricing on Muapi.

📝

Overview

About this model

Z-Image Base LoRA Trainer fine-tunes the full Z-Image Base checkpoint directly from your own image dataset, giving you a custom LoRA for a specific character, brand identity, or artistic style built on the base (non-distilled) Z-Image model. Upload a .zip archive of training images with optional .txt captions, and the trainer handles auto-tuned learning rates and step scheduling so you don't need to hand-tune a training run. Steps, learning rate, and LoRA rank are all still adjustable for users who want finer control, and training completes in minutes rather than hours.

1Z-Image Base Customization: Fine-tune the full Z-Image Base checkpoint for a targeted visual style or domain.

2Brand & Style Adaptation: Teach the model a specific brand identity, color palette, or artistic style.

3Object & IP Training: Create a targeted LoRA adapter for a proprietary product, icon, or piece of intellectual property.

💰

Pricing & Value

Cost analysis

Provider	Cost	Notes
muapiapp	$2.50 per 1,000 steps	Pay per training step with auto-tuned defaults and zero setup overhead.
Fal.ai	Not available	Z-Image Base LoRA Trainer is not hosted on Fal.ai.
Replicate	Not available	Z-Image Base LoRA Trainer is not hosted on Replicate.

muapiapp$2.50 per 1,000 steps

Pay per training step with auto-tuned defaults and zero setup overhead.

Fal.aiNot available

Z-Image Base LoRA Trainer is not hosted on Fal.ai.

ReplicateNot available

Z-Image Base LoRA Trainer is not hosted on Replicate.

* Competitor pricing is estimated based on similar model architectures and usage tiers.

⚙️

Technical Details

Configuration schema

Parameter	Type	Description	Default
Dataset Zip URL	string	URL to zip archive with images. Try to use at least 4 images in general the more the better. In addition to images the archive can contain text files with captions.	`undefined`
Trigger Word	string	Optional trigger word. If a caption file exists, it is prepended when not already present.	`example`
Steps	int	Number of steps to train the LoRA on.	`1000`
Learning Rate	float	Learning rate for Z-Image Base LoRA training.	`0.0001`
LoRA Rank	int	LoRA rank dimension.	`16`

Dataset Zip URLstring

URL to zip archive with images. Try to use at least 4 images in general the more the better. In addition to images the archive can contain text files with captions.

Default Valueundefined

Trigger Wordstring

Optional trigger word. If a caption file exists, it is prepended when not already present.

Default Valueexample

Stepsint

Number of steps to train the LoRA on.

Default Value1000

Learning Ratefloat

Learning rate for Z-Image Base LoRA training.

Default Value0.0001

LoRA Rankint

LoRA rank dimension.

Default Value16

📖

Implementation Guide

Developer documentation

How to Use Z-Image Base LoRA Trainer API

Prepare your dataset: Create a .zip archive with at least 4 high-quality training images. Optionally include .txt caption files sharing the same base filename as each image.
Set training parameters: Customize trigger_word, steps (default 1000, range 1,000-10,000), learning_rate (default 0.0001), and lora_rank (default 16, range 1-64).
Submit the request via API:

curl -X POST "https://api.muapi.ai/api/v1/z-image-base-lora-trainer" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "data": "https://example.com/my-dataset.zip",
    "trigger_word": "my_custom_style",
    "steps": 1000,
    "learning_rate": 0.0001,
    "lora_rank": 16
  }'

Fetch trained weights: Once status reaches completed, the response's outputs array contains a download URL for your trained .safetensors file.

❓

Common Questions

Frequently asked

What is Z-Image Base LoRA Trainer?

Z-Image Base LoRA Trainer is an asynchronous API that fine-tunes custom LoRA adapters for the Z-Image Base checkpoint from a zipped image dataset, with auto-tuned defaults and customizable hyperparameters.

How is this different from the standard Z-Image LoRA Trainer?

This trainer targets the Z-Image Base checkpoint rather than the standard Z-Image model, so LoRAs it produces are meant to be paired with Z-Image Base generations for best results.

How should I structure the dataset zip archive?

Upload a `.zip` file containing image files (PNG/JPEG/WebP). To include custom captions, add `.txt` files with the exact same base filename as each corresponding image.

What is a trigger word and how is it used?

A trigger word is a unique token prepended to training captions. Including this trigger word in a later generation prompt activates the specific learned character, object, or style.

How many steps should I use for training?

The default is 1,000 steps, which works well for most datasets. You can adjust training steps between 1,000 and 10,000 depending on your dataset size and desired fidelity.

What output format is generated upon completion?

Once training is complete, the API output contains a URL to download the trained LoRA weights file, ready for inference against the Z-Image Base checkpoint.