Nodes/ComfyUI_mittimiLoadPreset/(testing)LoadAndSettingParameters02
ComfyUI Node

(testing)LoadAndSettingParameters02

The experimental node that replaces your whole loader stack

By mittimi·Created 2 years ago·Updated 2 years ago· 1
(testing)LoadAndSettingParameters02
    • ckpt
    • stop_at_clip_layer
    • vae
    • positive_prompt
    • negative_prompt
    • Latent
    • Steps
    • CFG
    • sampler_name
    • scheduler
    ◄preset▾►
    ◄checkpoint▾►
    ◄ClipNum-1►
    ◄vae▾►
    ◄PosPromptA—►
    ◄PosPromptC—►
    ◄NegPromptA—►
    ◄NegPromptC—►
    ◄Width512►
    ◄Height512►
    ◄BatchSize1►
    ◄Steps20►
    ◄CFG—►
    ◄SamplerName▾►
    ◄Scheduler▾►

    The "(testing)" in the display name is doing honest work. LoadAndSettingParameters02 is the author's experiment in taking the preset idea all the way: instead of handing values to your existing loaders, it loads the checkpoint itself, encodes your prompts, builds the latent, and hands KSampler everything it needs in one go.

    What it replaces

    A normal txt2img stack - CheckpointLoaderSimple, CLIPSetLastLayer, a VAE loader, two CLIPTextEncode nodes, EmptyLatentImage - collapses into this single node. Its outputs are ckpt (MODEL), a CLIP, a VAE, positive and negative CONDITIONING, a Latent, plus Steps / CFG / sampler_name / scheduler.

    Naming quirk worth knowing before you wire anything: the second output is labeled stop_at_clip_layer, but it's actually the CLIP model with your ClipNum already applied. Don't be fooled - that's what goes into CLIPTextEncode's clip input.

    How it works

    It loads the checkpoint via ComfyUI's load_checkpoint_guess_config, applies the ClipNum to the CLIP, and loads the VAE - or uses the checkpoint's merged one if the vae dropdown is set to the special "Use_merged_checkpoints" option. It builds a blank latent of [BatchSize, 4, Height//8, Width//8], concatenates PosPromptA + ", " + PosPromptC into a single prompt, and encodes it. That's also why this is the only node in the pack with Width / Height / BatchSize widgets: it's the one constructing the latent, so it needs them.

    The layout runs preset dropdown on top, then checkpoint, ClipNum, vae, the four prompt boxes, and the dimensions/step/sampler controls. Same preset mechanism as the rest of the pack - selecting a preset rewrites the widgets and tweaks stick, but because it auto-selects a preset on creation, rebuilding a workflow from a dropped image re-applies it and clobbers your values.

    Where it fits

    Honestly, it's a quick test rig. One node to switch model, hit go, and A/B a checkpoint - great for that. But it's a beta the README doesn't even bother describing, and hiding the loader stack behind one node makes problems harder to debug. If something looks off, suspect this node first. For anything you'll maintain, the v1 non-testing nodes - or the author's successor pack, ComfyUI_mittimiLoadPreset2 - are the saner choice; the README says this whole repo won't be updated.

    Install

    Via ComfyUI Manager (search "ComfyUI_mittimiLoadPreset"), or manually:

    cd ComfyUI/custom_nodes
    git clone https://github.com/mittimi/ComfyUI_mittimiLoadPreset
    

    Restart ComfyUI. Only dependency is toml, no model downloads. Presets live in the pack's presets/ folder and are scanned at startup - add a .toml, restart, and remember to back them up before updating the pack.

    CategorymittimiTools

    Inputs (15)

    NameTypeDefaultDescription
    presetCOMBO3 options: (testing)preset.toml, PRESET TEMPLATE.toml, [Sample] PonyRealismSetting.toml
    checkpointCOMBO0 options:
    ClipNumINT-1-10–-1—
    vaeCOMBO1 options: Use_merged_checkpoints
    PosPromptASTRING—
    PosPromptCSTRING—
    NegPromptASTRING—
    NegPromptCSTRING—
    WidthINT5128–16384—
    HeightINT5128–16384—
    BatchSizeINT11–999999—
    StepsINT201–999999—
    CFGFLOAT—
    SamplerNameCOMBO34 options: euler, euler_cfg_pp, euler_ancestral, euler_ancestral_cfg_pp, heun, heunpp2, +28
    SchedulerCOMBO9 options: normal, karras, exponential, sgm_uniform, simple, ddim_uniform, +3

    Outputs (10)

    NameTypeDescription
    ckptMODEL—
    stop_at_clip_layerCLIP—
    vaeVAE—
    positive_promptCONDITIONING—
    negative_promptCONDITIONING—
    LatentLATENT—
    StepsINT—
    CFGFLOAT—
    sampler_nameeuler,euler_cfg_pp,euler_ancestral,euler_ancestral_cfg_pp,heun,heunpp2,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_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,ddim,uni_pc,uni_pc_bh2—
    schedulernormal,karras,exponential,sgm_uniform,simple,ddim_uniform,beta,linear_quadratic,kl_optimal—