Z-Image-Omni-Base
Strategic evolution of Z-Image-Base, featuring omni pre-training for unified image generation and editing, avoiding complexity and performance loss of task switching.
Overview
Z-Image-Omni-Base marks a strategic shift from the original 'Base' model towards an 'omni' (omnipotent) pre-training architecture. It unifies image generation and editing/inpainting tasks within a single framework using the Scalable Single-Stream Diffusion Transformer (S3-DiT). This omni pre-training allows for seamless transitions between generating new images and editing existing ones without the need for separate specialized models, offering higher parameter efficiency and flexibility for developers.
Features
- Omni Pre-training for Unified Generation and Editing
- Seamless Task Switching without Performance Loss
- Scalable Single-Stream DiT (S3-DiT) Architecture
- Supports Cross-Task LoRA Adapters
- 6B Parameter Efficiency
- Superior Performance on Complex Multimodal Tasks
Images
Related Links
Popular Tools
Explore our most popular creative tools
Z-Image Edit
Upload image, transform with one sentence
Creative Engine
One sentence, AI provides infinite prompt creativity.
Image Parse
Upload image, retrieve prompt instantly.
Prompt Library
Discover thousands of high-quality AI prompts.
Z-Image LoRA
Combine multiple LoRA models to create unique AI artwork
Z-Video
Generate creative videos from text or images with AI.
AI Image Generator
Turn your text into stunning images instantly.
Style Library
Explore curated artistic styles for your creations.
Remove Background
Instantly remove backgrounds from images with precision.
Image Upscaler
Enhance image resolution up to 4K/8K.
Image Reframe
Expand images to any aspect ratio with outpainting.