Training custom LoRA models for Z-Image Turbo requires a specific approach due to its distilled architecture. This guide covers the de-distillation adapter technique developed by Ostris, training parameters, and deployment steps. Quick Start: If you prefer a web-based solution, use our LoRA training tool to train custom Z-Image Turbo LoRAs directly in your browser—no local GPU required. (View Highlight)
ParameterValueNotesSteps2,000-5,000Start with 3,000 for testingLearning Rate1e-4 to 5e-5Lower for fine detailsLoRA Rank (r)8-16Higher = more capacityResolution1024×1024Match base modelBatch Size1-2Adjust based on VRAM (View Highlight)
Z-Image Turbo is a step-distilled model—it achieves fast 8-step generation through a distillation process. Standard LoRA training breaks this distillation quickly, resulting in:
• Loss of speed benefits (no longer 8-step capable)
• Unpredictable quality degradation
• Artifacts in generated outputs (View Highlight)
Adapter generation: Thousands of images were generated using Z-Image Turbo at various sizes and aspect ratios
Controlled distillation breakdown: A LoRA was trained on these images at low learning rate (1e-5), allowing distillation to break down in a controlled manner
Training on top: Your custom LoRA trains on top of this adapter, learning only your new content
Adapter removal: At inference time, remove the training adapter—your LoRA preserves distilled speed
This approach lets you train style, character, or concept LoRAs while maintaining 8-step Turbo generation. (View Highlight)
Download Training Adapter:
Use V2 of the adapter for refined results.
Quality training data is essential for good LoRA results.
• Count: 5-15 images for characters, 15-25 for styles
• Resolution: 1024×1024 minimum (matches Z-Image Turbo’s native resolution)
• Format: PNG or JPEG
• Variety: Include different angles, lighting, and contexts
Create a text file for each image with the same name:
Caption format example:
Use a consistent trigger word that will activate your LoRA during generation. (View Highlight)
Two options in AI Toolkit:
Z-Image Turbo W/ Training Adapter
• Preserves 8-step Turbo speed
• Best for shorter runs (< 5,000 steps)
• Remove adapter at inference
Z-Image De-Turbo (De-Distilled)
• No adapter needed at inference
• Suitable for longer training runs
• Slightly slower generation
For most use cases, option 1 (with training adapter) is recommended.
Adjust paths and parameters based on your dataset and hardware. (View Highlight)
Based on user reports:
Enable Low VRAM mode if you encounter memory errors on 12GB cards.
Copy LoRA file to ComfyUI/models/loras/
Add “Load LoRA” node to workflow
Connect to Z-Image Turbo model loader
Set LoRA strength (0.5-1.0)
Remember to include your trigger word in prompts.
• Verify file path is correct
• Check LoRA was trained on Z-Image Turbo (not FLUX or SD)
• Increase LoRA strength
• Training too long (distillation breakdown)
• Learning rate too high
• Reduce steps or use V2 training adapter
• Enable Low VRAM mode in AI Toolkit
• Reduce batch size to 1
• Use gradient checkpointing
• Remove training adapter if using Method 1
• Check LoRA strength (reduce if artifacts appear)
• Verify training completed without errors
Start small: Test with 1,000 steps before committing to longer runs
Caption carefully: Good captions improve LoRA quality significantly
Use V2 adapter: The refined V2 training adapter produces better results
Monitor checkpoints: Save every 500 steps to find optimal training point
Test incrementally: Generate samples at each checkpoint to avoid overtraining (View Highlight)
