Nodes/ComfyUI-MaxedOut/LoadLatent_WithParams
ComfyUI Node

LoadLatent_WithParams

Resume a saved Wan 2.2 latent with its original prompt and settings

By Maxed-Out-99·Created about a year ago·Updated 21 days ago· 9
LoadLatent_WithParams
    • shift
    • positive
    • negative
    • samples
    • steps
    • cfg
    • sampler_name
    • scheduler
    • end_at_step
    • filename_prefix
    • trim_latent
    • high_workflow
    latent
    run_folderfalse
    refresh_before_runfalse

    The single-latent loader from Maxed Out's save/load family, and the piece that makes the Wan 2.2 two-stage workflow actually survivable. Generate a bunch of high-noise starts with the first model, save the ones you like, then come back later and load one with its prompt and sampler settings still attached. The "WithParams" is the whole point: the parameters come back with the latent.

    Wan 2.2 splits generation between a high-noise model (motion, composition) and a low-noise refiner (detail). The community standard is to run the cheap first pass many times, then spend the expensive second pass only on the drafts worth it. Doing that by hand means keeping notes on which seed, steps, and CFG each start used. This node kills the notes.

    How it works

    Save Latent nodes in this pack write .latent safetensors files under ComfyUI/input/latents/, with the source workflow and KSampler parameters embedded as metadata. This loader opens the file you pick, reads that metadata, and returns the tensor plus every setting you saved with it. There's also the SD3-style shift value on the way out, which Wan 2.2 workflows use for the CFG/prompt scaling knob.

    The inputs and outputs that matter

    One input: latent, a dropdown of every .latent file under input/latents, grouped by folder. Outputs:

    • samples - the LATENT to feed your sampler.
    • positive / negative - the saved prompt text (this is the text-prompt flavor; the VACE 2.2 variant returns CONDITIONING instead).
    • steps, cfg, sampler_name, scheduler, end_at_step, shift - the settings from the original run.
    • filename_prefix - comes out as the file's folder + name, so re-saving keeps things tidy.

    Wire samples and the settings into your low-noise sampler and you're effectively resuming a run. For finishing multiple starts at once, use the batch version instead.

    Installing it

    It's in the ComfyUI-MaxedOut pack. ComfyUI Manager search Maxed Out, or:

    cd ComfyUI\custom_nodes
    git clone https://github.com/Maxed-Out-99/ComfyUI-MaxedOut.git
    

    Restart ComfyUI. No extra Python dependencies - the pack has no requirements.txt - and nothing to download.

    The honest caveat

    This name is a legacy key. Maxed Out has been renaming this family of latent nodes as the pack matures, and the older WithParams entries are kept alive so previously saved workflows still load. If you're setting up fresh on Wan 2.2, the VACE 2.2 variants are the current flavor. The other thing to know: the dropdown refreshes at ComfyUI startup, so a latent written by a different workflow while ComfyUI is running won't show up until you restart it.

    CategoryMXD/Latents

    Inputs (3)

    NameTypeDefaultDescription
    latentCOMBO0 options:
    run_folderBOOLEANfalseWhen enabled, hitting Queue Prompt auto-queues every latent in this file's folder, one after another, instead of just the selected file.
    refresh_before_runBOOLEANfalseWhen run_folder is on, re-scan the latents folder for new files right before the queuing loop starts, instead of using the dropdown list as of whenever it was last populated. Use this when another workflow is still writing latents into this folder as you queue this one.

    Outputs (12)

    NameTypeDescription
    shiftFLOAT
    positiveCONDITIONING
    negativeCONDITIONING
    samplesLATENT
    stepsINT
    cfgFLOAT
    sampler_nameeuler,euler_cfg_pp,euler_ancestral,euler_ancestral_cfg_pp,heun,heunpp2,exp_heun_2_x0,exp_heun_2_x0_sde,dpm_2,dpm_2_ancestral,lms,dpm_fast,dpm_adaptive,dpmpp_2s_ancestral,dpmpp_2s_ancestral_cfg_pp,dpmpp_sde,dpmpp_sde_gpu,dpmpp_2m,dpmpp_2m_cfg_pp,dpmpp_2m_sde,dpmpp_2m_sde_gpu,dpmpp_2m_sde_heun,dpmpp_2m_sde_heun_gpu,dpmpp_3m_sde,dpmpp_3m_sde_gpu,ddpm,lcm,ipndm,ipndm_v,deis,res_multistep,res_multistep_cfg_pp,res_multistep_ancestral,res_multistep_ancestral_cfg_pp,gradient_estimation,gradient_estimation_cfg_pp,er_sde,seeds_2,seeds_3,sa_solver,sa_solver_pece,ddim,uni_pc,uni_pc_bh2
    schedulersimple,sgm_uniform,karras,exponential,ddim_uniform,beta,normal,linear_quadratic,kl_optimal
    end_at_stepINT
    filename_prefixSTRING
    trim_latentINT
    high_workflowSTRING