Simple ZImageTurbo workflow for LoRA use.

Use this ZImageTurbo workflow to apply a light iterative latent upscale pass to a normally generated base image that uses one or more LoRAs and a chosen LoRA strength; it can produce stronger LoRA style or character expression, sharper fine details and textures, and better facial

Generate the base image normally in 8–12 steps, then connect it to Iterative Latent Upscale from Impact Pack.

Inputs include a ZImageTurbo base image, the model used to create it, one or more LoRAs, and the LoRA strength.

The pass can produce sharper fine details and textures, better facial features and micro-details, and stronger style or character expression without major composition changes.

The listed node packs are comfyui_essentials, comfyui-impact-pack, and rgthree-comfy.

The listed model files are ae.safetensors, qwen_3_4b.safetensors, Soyjak 2 z.safetensors, and z_image_turbo_bf16.safetensors.

Set the upscale factor to 1.1–2, use 1–3 upscale steps, and set denoise to 0.22–0.49.

Use the same model, LoRA or LoRAs, and strength for the base generation and upscale pass.

Before use, verify that ae.safetensors, qwen_3_4b.safetensors, Soyjak 2 z.safetensors, and z_image_turbo_bf16.safetensors are available.

Suggestion · not verified

The listing does not specify a license, so check applicable usage terms before reuse.

Suggestion · not verified