Skip to content

iXimNet/Z-ImageStudio

Repository files navigation

Z-Image Studio

轻量级的 Z-Image / Z-Image Turbo Web 端工作台,提供现代化界面与参数控制。
A lightweight web studio for Z-Image and Z-Image Turbo with a modern UI and controls.

最近更新 / Recent Updates (2026-01-31)

  • 历史卡片布局优化,内容顶对齐并保持同列高度一致。
    History cards keep consistent height per row with top-aligned content.
  • 预览占位与历史加载逻辑优化,删除当前预览时清空参数与占位。
    Preview placeholder updates based on history; clearing on delete resets metadata.
  • 预览大图支持 1:1、拖拽查看、边缘圆角提示与滚轮缩放;新增全屏预览。
    Modal preview adds 1:1 view, drag-to-pan with edge cues, wheel zoom, and fullscreen mode.
  • 运行时可配置推理 dtype(ZIMAGE_DTYPE),自动选择更兼容的类型。
    Runtime dtype selection (ZIMAGE_DTYPE) to improve hardware compatibility.

功能 / Features

  • 提示词控制台(负向提示词、引导尺度、步数、随机种子)。
    Prompt console with negative prompt, guidance scale, steps, and seed controls.
  • 文生图:支持 Z-Image Turbo 与 Z-Image,一键切换。
    Text-to-image with Z-Image Turbo or Z-Image (toggle in UI).
  • 切换模型时应用推荐参数默认值。
    Model-aware defaults apply recommended steps/guidance on switch.
  • 可选:切换模型时自动卸载旧模型以降低显存占用。
    Optional auto-unload of previous models to reduce VRAM usage.
  • 生成历史画廊(预览、元数据、删除)。
    History gallery with preview, metadata, and delete.

快速开始 / Quick Start

  1. 安装依赖 / Install dependencies:

    pip install -r requirements.txt
  2. 运行应用 / Run the app:

    python app.py
  3. 打开页面 / Open the UI:

    http://localhost:7860
    
  4. 可选:从示例生成 .env / Optional: create .env from example:

    copy .env.example .env

说明 / Notes

  • 宽高必须能被 16 整除。
    Width and height must be divisible by 16.
  • Z-Image Turbo 建议 9 步左右、低引导尺度。
    Z-Image Turbo runs best around 9 steps with low guidance values.
  • Z-Image 建议 28-50 步、引导尺度 3-5。
    Z-Image is tuned for 28-50 steps with guidance around 3-5.
  • 输出保存到 generated_images/,上传保存到 uploads/,历史保存到 history.json
    Outputs are saved to generated_images/, uploads to uploads/, history to history.json.

配置 / Configuration

应用会读取 .env(不提交到仓库)作为运行时配置:
The app reads a .env file (not committed) for runtime options:

  • ZIMAGE_DEVICE: cuda / mps / cpu(留空自动检测)。
    cuda / mps / cpu (blank = auto).
  • ZIMAGE_DTYPE: bf16 / fp16 / fp32(留空自动选择)。
    bf16 / fp16 / fp32 (blank = auto).
  • ZIMAGE_MPS_UNET_FP32: MPS 下将 UNet/transformer 上采样到 fp32(1 启用,0 关闭)。
    Upcast UNet/transformer to fp32 on MPS (1 enable, 0 disable).
  • ZIMAGE_MPS_NAN_FALLBACK: MPS 下出现 NaN/黑图告警时自动回退到 fp32 并重试(1 启用,0 关闭)。
    Retry with fp32 UNet/transformer on MPS when NaN/black-image warnings appear (1 enable, 0 disable).
  • ZIMAGE_CPU_OFFLOAD: CUDA 下启用 CPU Offload(1 启用,0 关闭)。
    Enable CPU offload on CUDA (1 enable, 0 disable).
  • ZIMAGE_KEEP_MODELS: 是否保留多模型(1 保留,0 切换时卸载)。
    Keep multiple models loaded (1) or unload others on switch (0).

MPS 稳定性与性能提示:
MPS stability & performance tips:

  • 若出现黑图/NaN 告警,保持 ZIMAGE_MPS_NAN_FALLBACK=1,必要时开启 ZIMAGE_MPS_UNET_FP32=1
    If you see black images/NaN warnings, keep ZIMAGE_MPS_NAN_FALLBACK=1 and optionally set ZIMAGE_MPS_UNET_FP32=1.
  • ZIMAGE_DTYPE=bf16 仅在 MPS 支持 bf16 时可用;若报错请回退到 fp16/fp32
    ZIMAGE_DTYPE=bf16 requires MPS bf16 support; fall back to fp16/fp32 if it errors.

模型来源与致谢 / Model Sources & Acknowledgements

本项目使用的 Z-Image 与 Z-Image-Turbo 模型由通义实验室(Tongyi-MAI)在 ModelScope 发布。感谢模型作者与社区的开源贡献与支持。
Z-Image and Z-Image-Turbo are published by Tongyi-MAI on ModelScope. We respect and appreciate the authors and the community for their open-source contributions.

模型链接 / Model links:

https://www.modelscope.cn/models/Tongyi-MAI/Z-Image
https://www.modelscope.cn/models/Tongyi-MAI/Z-Image-Turbo

About

A lightweight web studio for Z-Image Turbo and Z-Image with a mordern-style front end.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors