Z-Image-Turbo

8-Step Fast Generation Z-Image-Turbo generates high-quality images in just 8 NFEs (Number of Function Evaluations) using Decoupled-DMD distillation technology. This breakthrough efficiency delivers sub-second generation on enterprise H800 GPUs while maintaining photorealistic output quality that rivals models requiring 50+ steps. Bilingual Text Rendering Z-Image-Turbo excels at rendering complex text directly in generated images—supporting both English and Chinese with exceptional accuracy. Create posters, logos, and graphics with legible typography that most AI image generators struggle to produce correctly. Photorealistic Image Quality Powered by the S3-DiT (Scalable Single-Stream DiT) architecture, Z-Image-Turbo produces photorealistic images with accurate lighting, shadows, and details. The DMDR framework combines distribution matching with reinforcement learning to enhance semantic alignment and visual aesthetics. Consumer GPU Compatible Z-Image-Turbo fits comfortably within 16GB VRAM, enabling professional-quality AI image generation on consumer-grade GPUs. No expensive enterprise hardware required—run the full 6B parameter model on your own machine with optimized memory efficiency. Prompt Enhancement Z-Image-Turbo's built-in Prompt Enhancer empowers the model with reasoning capabilities, transcending surface-level descriptions to tap into underlying world knowledge. Generate images that accurately interpret your creative intent with robust instruction adherence. Open Source (Apache-2.0) Z-Image-Turbo is fully open-source under the Apache-2.0 license. Access the complete model weights on Hugging Face and ModelScope, customize for your workflows, and deploy without licensing restrictions. Commercial use permitted.