Nodes/ComfyUI-Config-Presets/Preset Settings Saver
ComfyUI Node

Preset Settings Saver

Save your exact sampler settings before you forget them

By pooria98·Created 3 months ago·Updated 2 months ago· 0
Preset Settings Saver
      preset_namemy-generation-preset
      width512
      height512
      cfg7.0
      steps20
      sampler_nameeuler
      schedulernormal
      hires_steps10
      hires_cfg7.0
      hires_denoise0.50
      hires_upscale_by2.00
      hires_modelNone found
      is_latenttrue

      The Preset Settings Saver is the write-half of this pack's two-node trick. The Loader reads a named preset from a JSON file; this node lets you create one from the UI instead of hand-editing JSON. You found a configuration that works - 832x1216, CFG 5, euler, 25 steps, that hires recipe you keep tweaking - and you want it saved so the dropdown has it next time.

      It's a standalone output node that does exactly one thing: write settings to configs/generation_presets.json inside the pack folder.

      How it works

      Drop it onto an empty canvas (it's an output node, so it can't chain into anything), fill in the values, and hit the blue execute/queue button. It reads the existing JSON, replaces any preset with the same name, appends the new one, and writes the file back. There's nothing to wire - it has no outputs at all.

      Two details are worth knowing because they come straight from ComfyUI itself. The sampler and scheduler dropdowns are pulled live from ComfyUI's own lists (comfy.samplers), so you can't typo a sampler that doesn't exist - you get the 44 samplers and 9 schedulers ComfyUI actually supports. And hires_model is populated from your upscale_models folder via folder_paths; if you have no upscalers installed, it reads "None found."

      The inputs you'll actually touch

      • preset_name - whatever you want to call it, e.g. "dreamshaper 8" or "karras 512". This is what shows up in the Loader's dropdown.
      • width, height, cfg, steps - the core generation settings.
      • sampler_name, scheduler - pick from the dropdowns.
      • hires_denoise, hires_upscale_by, is_latent - for a hires-fix second pass. is_latent toggles latent upscaling (upscale in latent space, no model needed) versus a model-based upscale through hires_model. The default is true, which is the "free" path.

      Then save, refresh the page, and the Loader node sees it.

      Installing

      Same as its sibling. ComfyUI Manager ("ComfyUI Config Presets"), or:

      cd ComfyUI/custom_nodes
      git clone https://github.com/pooria98/ComfyUI-Config-Presets.git
      

      Restart ComfyUI. No model downloads, and no dependencies beyond what the pack already needs.

      Gotchas

      • Refresh after saving. The README is blunt about this: you must refresh the page before the Loader dropdown updates. If you save and think "it didn't work," it usually did - you just haven't refreshed.
      • Your presets live inside the install folder. The JSON is written into custom_nodes/ComfyUI-Config-Presets/configs/. Delete the folder to reinstall or move machines and your presets go with it - once you've built a collection you like, copy that file somewhere safe.
      • It's a saver, not a validator. A preset_name that already exists is silently overwritten, which is convenient - that's how you update a preset in place. Just don't expect a warning about it.

      The take

      This node exists to make the JSON approachable. Editing that file by hand is genuinely fiddly - one stray comma and the Loader quietly finds nothing. The Saver gives you a form that can't corrupt the file, and that's basically its whole job. Not glamorous, but if you collect presets, it beats a text editor.

      CategoryConfig Presets

      Inputs (13)

      NameTypeDefaultDescription
      preset_nameSTRINGmy-generation-preset
      widthINT51264–8192
      heightINT51264–8192
      cfgFLOAT7.00–100
      stepsINT201–10000
      sampler_nameCOMBOeuler44 options: euler, euler_cfg_pp, euler_ancestral, euler_ancestral_cfg_pp, heun, heunpp2, +38
      schedulerCOMBOnormal9 options: simple, sgm_uniform, karras, exponential, ddim_uniform, beta, +3
      hires_stepsINT101–10000
      hires_cfgFLOAT7.00–100
      hires_denoiseFLOAT0.500–1
      hires_upscale_byFLOAT2.000.1–10
      hires_modelCOMBONone found1 options: None found
      is_latentBOOLEANtrue

      Outputs (0)

      No outputs