ltx_24gb.sh

#!/bin/bash

# Provisioning Script for 24GB VRAM (e.g., A5000, RTX A5000, 3090, 4090)
# Uses LTX-2 FP8 Model (~19GB) + Quantized Text Encoder (~6GB)
# Tries to fit within 24GB VRAM, might swap slightly to system RAM or require strict offloading.

source /venv/main/bin/activate
COMFYUI_DIR=${WORKSPACE}/ComfyUI

APT_PACKAGES=(
    "ffmpeg"
    "git"
    "wget"
    "aria2"
)

PIP_PACKAGES=(
    "yt_dlp"
)

NODES=(
    "https://github.com/ltdrdata/ComfyUI-Manager"
    "https://github.com/cubiq/ComfyUI_essentials"
    "https://github.com/Lightricks/ComfyUI-LTXVideo"
    "https://github.com/yuvraj108c/ComfyUI-Whisper"
    "https://github.com/MixLabPro/comfyui-sound-lab"
    "https://github.com/Fannovel16/ComfyUI-VideoHelperSuite"
    "https://github.com/Fannovel16/ComfyUI-Frame-Interpolation"
    "https://github.com/Kosinkadink/ComfyUI-Advanced-ControlNet"
)

# LTX-2 FP8 Model (Best balance for 24GB)
CHECKPOINT_MODELS=(
    "https://huggingface.co/Lightricks/LTX-2/resolve/main/ltx-2-19b-dev-fp8.safetensors"
)

UNET_MODELS=(
)

LORA_MODELS=(
)

VAE_MODELS=(
)

ESRGAN_MODELS=(
    "https://huggingface.co/Lightricks/LTX-2/resolve/main/ltx-2-spatial-upscaler-x2-1.0.safetensors"
    "https://huggingface.co/Lightricks/LTX-2/resolve/main/ltx-2-temporal-upscaler-x2-1.0.safetensors"
)

CONTROLNET_MODELS=(
)

# Text Encoder for LTX-2 (Quantized to save VRAM)
TEXT_ENCODER_MODELS=(
    "https://huggingface.co/city96/t5-v1_1-xxl-encoder-gguf/resolve/main/t5-v1_1-xxl-encoder-Q4_K_M.gguf"
    # LTX-2 Text Encoder: Gemma 3 12B (Required)
    "https://huggingface.co/Lightricks/LTX-2/resolve/main/gemma-3-12b-it-qat-q4_0-unquantized.safetensors"
    # T5 Text Encoder (Often required for standard SD/Flux/LTX pipelines)
    "https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors"
)

# MusicGen / Stable Audio Models (for comfyui-sound-lab)
MUSICGEN_MODELS=(
    "https://huggingface.co/facebook/musicgen-small/resolve/main/state_dict.bin"
)
STABLE_AUDIO_MODELS=(
    "https://huggingface.co/stabilityai/stable-audio-open-1.0/resolve/main/model.safetensors"
)



### DO NOT EDIT BELOW HERE UNLESS YOU KNOW WHAT YOU ARE DOING ###

function provisioning_start() {
    provisioning_print_header
    provisioning_get_apt_packages
    provisioning_get_nodes
    provisioning_get_pip_packages
    
    # ComfyUI Model Structure
    provisioning_get_files "${COMFYUI_DIR}/models/checkpoints" "${CHECKPOINT_MODELS[@]}"
    provisioning_get_files "${COMFYUI_DIR}/models/unet" "${UNET_MODELS[@]}"
    provisioning_get_files "${COMFYUI_DIR}/models/loras" "${LORA_MODELS[@]}"
    provisioning_get_files "${COMFYUI_DIR}/models/controlnet" "${CONTROLNET_MODELS[@]}"
    provisioning_get_files "${COMFYUI_DIR}/models/vae" "${VAE_MODELS[@]}"
    provisioning_get_files "${COMFYUI_DIR}/models/esrgan" "${ESRGAN_MODELS[@]}"
    
    # LTX-2 Text Encoder often needs to be in models/text_encoders
    provisioning_get_files "${COMFYUI_DIR}/models/text_encoders" "${TEXT_ENCODER_MODELS[@]}"

    # MusicGen & Stable Audio
    provisioning_get_files "${COMFYUI_DIR}/models/musicgen" "${MUSICGEN_MODELS[@]}"
    provisioning_get_files "${COMFYUI_DIR}/models/stable_audio" "${STABLE_AUDIO_MODELS[@]}"
    
    # Pre-create folders for other nodes to avoid permissions issues
    mkdir -p "${COMFYUI_DIR}/models/whisper"
    
    provisioning_print_end
}

function provisioning_get_apt_packages() {
    if [[ -n $APT_PACKAGES ]]; then
            sudo apt-get update
            sudo apt-get install -y ${APT_PACKAGES[@]}
    fi
}

function provisioning_get_pip_packages() {
    if [[ -n $PIP_PACKAGES ]]; then
            pip install --no-cache-dir ${PIP_PACKAGES[@]}
    fi
}

function provisioning_get_nodes() {
    for repo in "${NODES[@]}"; do
        dir="${repo##*/}"
        path="${COMFYUI_DIR}/custom_nodes/${dir}"
        requirements="${path}/requirements.txt"
        if [[ -d $path ]]; then
            if [[ ${AUTO_UPDATE,,} != "false" ]]; then
                printf "Updating node: %s...\n" "${repo}"
                ( cd "$path" && git pull )
                if [[ -e $requirements ]]; then
                   pip install --no-cache-dir -r "$requirements"
                fi
            fi
        else
            printf "Downloading node: %s...\n" "${repo}"
            git clone "${repo}" "${path}" --recursive
            if [[ -e $requirements ]]; then
                pip install --no-cache-dir -r "${requirements}"
            fi
        fi
    done
}

function provisioning_get_files() {
    if [[ -z $2 ]]; then return 1; fi
    
    dir="$1"
    mkdir -p "$dir"
    shift
    arr=("$@")
    
    if [[ ${#arr[@]} -eq 0 ]]; then return; fi

    printf "Downloading %s model(s) to %s...\n" "${#arr[@]}" "$dir"
    for url in "${arr[@]}"; do
        printf "Downloading: %s\n" "${url}"
        provisioning_download "${url}" "${dir}"
        printf "\n"
    done
}

function provisioning_print_header() {
    printf "\n##############################################\n#                                            #\n#          Provisioning (24GB VRAM)          #\n#                                            #\n#         This will take some time           #\n#                                            #\n##############################################\n\n"
}

function provisioning_print_end() {
    printf "\nProvisioning complete:  Application will start now\n\n"
}

# Download from $1 URL to $2 file path
function provisioning_download() {
    if [[ -n $HF_TOKEN && $1 =~ ^https://([a-zA-Z0-9_-]+\.)?huggingface\.co(/|$|\?) ]]; then
        auth_token="$HF_TOKEN"
    elif 
        [[ -n $CIVITAI_TOKEN && $1 =~ ^https://([a-zA-Z0-9_-]+\.)?civitai\.com(/|$|\?) ]]; then
        auth_token="$CIVITAI_TOKEN"
    fi
    if [[ -n $auth_token ]];then
        wget --header="Authorization: Bearer $auth_token" -qnc --content-disposition --show-progress -e dotbytes="${3:-4M}" -P "$2" "$1"
    else
        wget -qnc --content-disposition --show-progress -e dotbytes="${3:-4M}" -P "$2" "$1"
    fi
}

if [[ ! -f /.noprovisioning ]]; then
    provisioning_start
fi