Skip to content
Meshive GPU Cloud Docs

QuickStart - Deploying ComfyUI 2

Load a workflow template inside ComfyUI, download missing models via SSH, and generate your first output.

  1. Load Workflow in ComfyUI

    Section titled “Load Workflow in ComfyUI”

    Select Template inside ComfyUI.

    "Select Template" button inside the ComfyUI interface ComfyUI template gallery overlay

    Check the missing nodes in the modal.

    "Missing nodes" modal listing required custom nodes and Copy URL buttons

    Don’t close this popup — you should click [Copy URL] next to each model.

    Via SSH, download the models with the commands below

    Section titled “Via SSH, download the models with the commands below”
    SSH terminal connected to the Pod

    Download each file into the folder ComfyUI reads that model kind from, under /workspace/ComfyUI/models/. Putting a file anywhere else means ComfyUI will not find it and the node stays red.

    1) Create the folder if it doesn’t exist:

    Terminal window
    mkdir -p /workspace/ComfyUI/models/vae
    wget -c -O /workspace/ComfyUI/models/vae/ae.safetensors \
      "https://huggingface.co/Comfy-Org/z_image_turbo/resolve/main/split_files/vae/ae.safetensors?download=true"
    

    mkdir -p /workspace/ComfyUI/models/text_encoders
    wget -c -O /workspace/ComfyUI/models/text_encoders/qwen_3_4b.safetensors \
      "https://huggingface.co/Comfy-Org/z_image_turbo/resolve/main/split_files/text_encoders/qwen_3_4b.safetensors?download=true"

    2) Skip mkdir if the folder already exists:

    Terminal window
    wget -c -O /workspace/ComfyUI/models/diffusion_models/z_image_bf16.safetensors \
      "https://huggingface.co/Comfy-Org/z_image/resolve/main/split_files/diffusion_models/z_image_bf16.safetensors?download=true"
    Terminal showing wget progress for downloading models

  2. Generate Image or Video

    Section titled “Generate Image or Video”
    Generated image output from the ComfyUI workflow

Next Steps

Section titled “Next Steps”
  • Write Prompt — input the positive and negative prompts. Z-image Turbo allows for high-quality generation even with simple descriptions.
  • Modify Sampler Settings — adjust parameters such as steps, CFG scale, and seed to fine-tune the generation process.
  • Apply LoRAs — connect LoRA nodes to apply specific styles or characters to your Z-image Turbo workflow.