Z-Image is an open-weights AI image generation model suite built for fast, high-quality, and developer-friendly visual creation. The collection includes turbo and base text-to-image models, image-to-image generation, inpainting, ControlNet-guided generation, LoRA inference, and LoRA training workflows, giving creators and developers flexible tools for scalable image production.
Built around an efficient 6B-parameter image generation architecture, Z-Image is designed for strong photorealistic quality, fast generation speed, bilingual text rendering, and reliable prompt adherence. It is suitable for portraits, product visuals, posters, concept art, social media assets, e-commerce images, character styles, design iteration, and customized visual generation.
Core Model Capabilities
Fast Text-to-Image Generation:
Use Z-Image Turbo to generate high-quality images from natural-language prompts with fast inference, strong prompt understanding, clean composition, and practical production performance.
Base Image Generation:
Use Z-Image base models for higher-flexibility image generation, creative exploration, fine-tuning, and downstream development workflows.
Image-to-Image Generation:
Transform input images into new visual outputs while preserving key composition, subject structure, and style direction.
Inpainting:
Edit or regenerate selected areas of an image using prompts and masks, making it useful for object replacement, background fixes, visual cleanup, and creative retouching.
ControlNet Guidance:
Use edge, depth, pose, or structure-based guidance to control composition, layout, character posture, and scene geometry more precisely.
LoRA Inference:
Apply custom LoRA weights to generate consistent styles, characters, products, or brand-specific visual identities across image outputs.
LoRA Training:
Train custom LoRA adapters from reference images to support personalized styles, consistent characters, and repeatable creative production workflows.
Open-Weights Creative Workflows:
Build image generation pipelines with open-weights flexibility, scalable API access, and production-ready deployment on WaveSpeedAI infrastructure.
Z-Image Models on WaveSpeedAI give creators and developers fast access to open-weights AI image generation, image editing, ControlNet guidance, LoRA customization, and LoRA training tools with flexible pricing, scalable APIs, and production-ready visual quality.