imageLoRA training
Z Image
6B-parameter Turbo text-to-image with 8-step inference plus ControlNet, i2i, inpaint, and LoRA tiers.
Studio is in early access; availability resolves per project after sign-in.
- Endpoints
- 12
- Eligible
- 12
- Input schemas imported
- 12 / 12
- Delivery
- fal.ai
- Weights
- Not stated
- Developer
- Not stated in source
Overview
Z-Image Turbo compresses text-to-image inference to 8 steps maximum, configurable down to 1, on a lean 6B-parameter base. Trading model size for raw speed suits rapid prototyping, variation testing, and high-volume asset generation. ControlNet, image-to-image, inpaint, and LoRA tiers extend the family.
Capabilities
- 8-step max inference
- 6B lean footprint
- ControlNet guided i2i
- LoRA adaptation tiers
Best for
- Rapid prototyping
- Content variation testing
- High-volume assets
Use cases
- Speed-first pipelines
- Controlled i2i edits
- Custom LoRA styles
Endpoints
12 eligible. Cataloged is not the same as executable: Studio resolves which endpoints a project can run.
| Endpoint | Task | Catalog status | Inputs |
|---|---|---|---|
fal-ai/z-image/base | image generation | Eligible | 11 fields · requires prompt |
fal-ai/z-image/base/lora | lora training | Eligible | 12 fields · requires prompt |
fal-ai/z-image/turbo | image generation | Eligible | 10 fields · requires prompt |
fal-ai/z-image/turbo/controlnet | image generation | Eligible | 15 fields · requires prompt, image_url |
fal-ai/z-image/turbo/controlnet/lora | lora training | Eligible | 16 fields · requires prompt, image_url |
fal-ai/z-image/turbo/image-to-image | image to image | Eligible | 12 fields · requires prompt, image_url |
fal-ai/z-image/turbo/image-to-image/lora | image to image | Eligible | 13 fields · requires prompt, image_url |
fal-ai/z-image/turbo/inpaint | image editing | Eligible | 16 fields · requires prompt, image_url, mask_image_url |
fal-ai/z-image/turbo/inpaint/lora | image editing | Eligible | 17 fields · requires prompt, image_url, mask_image_url |
fal-ai/z-image/turbo/lora | lora training | Eligible | 11 fields · requires prompt |
fal-ai/z-image/turbo/tiling | image generation | Eligible | 16 fields · requires prompt |
fal-ai/z-image/turbo/tiling/lora | lora training | Eligible | 17 fields · requires prompt |
Questions
Step range?
1 to 8, far below standard diffusion.
Which tiers?
Turbo, ControlNet, image-to-image, inpaint, LoRA.