Photon-P2.1

Overview

Photon-P2.1 is a Text-to-Image LoRA fine-tune of Z-Image-Turbo, our high-quality image generation model. It was trained to push the base model toward richer scenes, more coherent lighting, and stronger composition — while keeping the fast, few-step generation behavior that Z-Image-Turbo is known for.

What's New in P2.1

Photon-P2.1 is the first refinement step of the Photon-P2 series. Compared to the earlier Photon releases, it delivers:

More natural lighting — softer falloff, better color grading, and more believable atmosphere across a wide range of scenes.
Improved scene diversity — from arctic interiors and beaches to pirate ships and tea parties, P2.1 holds up across very different subjects and moods.
Stronger composition — better framing and depth, with subjects that sit naturally in their environment instead of floating on top of it.
Cleaner details — crisper textures on fur, feathers, water, foliage, and architecture without over-sharpening.

The sample grid above shows generations spanning landscapes, wildlife, interior scenes, and whimsical settings — all produced with the same LoRA, demonstrating how consistently P2.1 adapts to different prompts.

About Photon

Photon is our line of LoRA adapters built on top of Z-Image-Turbo. Each Photon release focuses on making the base model feel more like a photographer and less like a renderer: intentional lighting, grounded compositions, and a painterly yet realistic finish.

Usage

Use Photon-P2.1 with any pipeline that supports Z-Image-Turbo. Load the LoRA weight (Photon-P2.1.safetensors) on top of the base model and generate as you normally would — no special prompting tricks required.

License

Photon-P2.1 is released under the Apache 2.0 license. You are free to use, modify, and distribute this model, including commercially.