Z-Image Turbo + (High / Low VRAM) CivitAI metadata + 4k Upscaling + Detail Daemon — v1.0

Use this workflow to generate Z-Image Turbo images from text, adjust detail with Detail Daemon, and upscale images with SeedVR2. It takes text and either a standard or GGUF model, and produces generated or upscaled images.

Image generation uses a standard LoRA pipeline instead of hooks.

For CivitAI uploads, metadata nodes record the model, sampling settings, and LoRAs used, so you do not have to enter that generation information by hand.

Model files listed are ae.safetensors, ema_vae_fp16.safetensors, qwen_3_4b.safetensors, seedvr2_ema_7b_sharp_fp16.safetensors, z_image_bf16.safetensors, z_image_turbo_bf16.safetensors, and z_image_turbo-Q8_0.gguf.

The workflow also lists these node packs: cg-use-everywhere, comfyui_essentials, comfyui_layerstyle, comfyui_lg_samplingutils, comfyui-custom-scripts, comfyui-detail-daemon, and comfyui-easy-use.

Other listed node packs are comfyui-gguf, comfyui-image-saver, comfyui-impact-pack, comfyui-kjnodes, rgthree-comfy, seedvr2_videoupscaler, and was-ns.

If you do not need CivitAI upload metadata, remove its group and enable the Save image node for standard image saving.

Suggestion · not verified

If you plan to use only one model format, remove the other model; the workflow supports standard and GGUF models.

Suggestion · not verified

Yellow nodes contain inputs and settings you can change; red nodes contain instructions and helpful notes.

A reuse license is not specified here; check the CivitAI page for its terms.

Suggestion

Before using the GGUF option, check that your ComfyUI setup has comfyui-gguf and z_image_turbo-Q8_0.gguf available.

Suggestion · not verified