This workflow is a fast and highly practical Z-Image i2L pipeline that turns a small image set into a usable LoRA and immediately tests it inside the same graph with ControlNet support. Instead of separating training and testing into different workflows, this setup combines both stages into one streamlined process, making it especially useful for creators who want to prototype new styles, characters, objects, or visual identities with minimal friction.
The first part of the workflow focuses on rapid LoRA generation. It uses the RunningHub Z-Image i2L node system to load the pipeline, batch multiple reference images, and generate a lightweight LoRA from those images. In this version, six reference images are used, which is a very practical balance: enough to give the model repeated visual signals, but still lightweight enough for quick experimentation. This makes it suitable for testing a character concept, a fashion look, a creature design, an object identity, or a consistent art direction without preparing a full traditional training dataset.
After the LoRA is created, the workflow immediately saves it and moves directly into the test stage. This is one of the biggest advantages of the setup. You do not need to manually relocate files, rebuild a second workflow, or restart anything. The newly generated LoRA is loaded back into Z-Image Base and tested right away, which dramatically reduces iteration time. For creators working on fast concept validation, this is extremely efficient.
What makes this version even more valuable is the addition of ControlNet testing. Instead of only checking whether the LoRA works in a simple text-to-image scenario, the workflow also evaluates how well it performs under structural guidance. A reference image is passed through DepthAnythingV2Preprocessor to create a depth control map, and that map is applied through Z-Image Fun ControlNet Union. This means you can test whether the LoRA still holds up when composition, depth, and layout are being controlled by an external structure image.
The generation side is also more advanced than a basic one-pass sampler. It uses Z-Image Base together with a structured sampling pipeline that includes CFGGuider, BasicScheduler, SplitSigmas, DetailDaemonSamplerNode, and SamplerCustomAdvanced. This allows the image to be built in a more controlled way, with the sigma schedule divided into structure-building and refinement stages. As a result, the workflow is not only good for fast LoRA prototyping, but also for serious evaluation of prompt compatibility, structural stability, and visual consistency.
In short, this workflow is ideal for creators who want to turn a handful of images into a LoRA quickly, then immediately test whether that LoRA can survive real-world generation conditions such as prompt guidance and ControlNet depth control. If you want to see how the full pipeline is built, how the nodes connect, and how to get the best results from it, make sure to watch the full video tutorial on my YouTube channel.
⚙️ Try the Workflow Online
👉 Workflow: https://www.runninghub.ai/post/2023370355200495617/?inviteCode=rh-v1111
Open the link above to run the workflow directly online and view the generation results in real time.
If the results match your expectations, you can also deploy it locally for further customization.
🎁 Fan Benefits: Register now to get 1000 points, plus 100 daily login points — enjoy 4090-level performance and 48 GB of powerful compute!
📺 Bilibili Updates (Mainland China & Asia-Pacific)
If you are in Mainland China or the Asia-Pacific region, you can watch the video below for workflow demos and a detailed creative breakdown.
📺 Bilibili Video: https://www.bilibili.com/video/BV1qXZMBwEC7/
I will continue updating model resources on Quark Drive:
👉 https://pan.quark.cn/s/20c6f6f8d87b
These resources are mainly prepared for local users, making creation and learning more convenient.
⚙️ 在线体验工作流
👉 工作流： https://www.runninghub.ai/post/2023370355200495617/?inviteCode=rh-v1111
打开上方链接即可直接运行该工作流，实时查看生成效果。
如果觉得效果理想，你也可以在本地进行自定义部署。
🎁 粉丝福利： 注册即送 1000 积分，每日登录 100 积分，畅玩 4090 体验 48 G 超级性能！
📺 Bilibili 更新（中国大陆及南亚太地区）
如果你在中国大陆或南亚太地区，可以通过下方视频查看该工作流的实测效果与构思讲解。
📺 B站视频： https://www.bilibili.com/video/BV1qXZMBwEC7/
我会在 夸克网盘 持续更新模型资源：
👉 https://pan.quark.cn/s/20c6f6f8d87b
这些资源主要面向本地用户，方便进行创作与学习。