Nodes/lzits Nodes/Bernini Setup Commands
ComfyUI Node

Bernini Setup Commands

A node that prints the setup commands you were about to Google

By lzitser23·Created 9 months ago·Updated 3 months ago· 0
Bernini Setup Commands
    • commands
    target_dir/root/Bernini
    model_dir/root/Bernini-R-Diffusers
    install_requirementstrue

    Getting Bernini running means cloning a repo, installing its requirements into the right Python, and downloading a 14B model. That's three commands you'll mistype once and then have to untangle. Bernini Setup Commands just prints them for you, copy/paste ready. That's the entire job - it does not run anything, because the setup has to happen in a dedicated Python 3.11 CUDA environment, not ComfyUI's.

    The inputs

    • target_dir - where to clone bytedance/Bernini (default ~/Bernini).
    • model_dir - where to download ByteDance/Bernini-R-Diffusers (default ~/Bernini-R-Diffusers).
    • install_requirements - whether the printed commands include pip install -r requirements.txt inside the cloned repo.

    What it outputs

    A single commands string, roughly:

    # Bernini-R setup commands (run in a Python 3.11 + CUDA 12.4 environment)
    git clone https://github.com/bytedance/Bernini.git /path/to/Bernini
    cd /path/to/Bernini
    pip install -r requirements.txt
    pip install -U huggingface_hub
    hf download ByteDance/Bernini-R-Diffusers --local-dir /path/to/Bernini-R-Diffusers
    # Then set Bernini Model Config's bernini_repo_dir/config to those two paths.
    

    Read it off the node, or wire the string into a Show Text-style display node and keep it in your graph. The node is an output node, so it renders on its own.

    How to use it, in context

    1. Add the node, adjust the two paths, run it, copy the commands.
    2. Run them in a Python 3.11 + CUDA 12.4 environment (the upstream repo recommends a strong GPU - H100-class hardware is the official ask, though people run it on far less).
    3. Point Bernini Model Config's bernini_repo_dir and config at those two paths.
    4. Wire Model Config + Case Builder + Generation Settings into Bernini Run Single GPU and go.

    One thing the node can't help with: it prints commands for the renderer (Bernini-R). If you want the full model with the MLLM planner, that's ByteDance/Bernini and a different, heavier setup - and the community's honest read is that the planner never got a proper ComfyUI path, so nearly everyone runs the renderer-only version this node targets.

    Installing this node

    One of the 16 nodes in the lzits Nodes pack (Lior Zitser, MIT). Install via ComfyUI Manager (search "lzits Nodes") or:

    cd ComfyUI/custom_nodes
    git clone https://github.com/lzitser23/lzits-nodes.git
    

    Restart ComfyUI. Zero dependencies - it only builds a string.

    Categorylzits nodes/Bernini

    Inputs (3)

    NameTypeDefaultDescription
    target_dirSTRING/root/Bernini
    model_dirSTRING/root/Bernini-R-Diffusers
    install_requirementsBOOLEANtrue

    Outputs (1)

    NameTypeDescription
    commandsSTRING