KSampler (Extras)
The Sampler That Speaks WebUI — A1111 Euler, AYS, beta57, GITS
- model
- positive
- negative
- latent_image
- LATENT
The pack author moved from A1111 to ComfyUI and refused to give up their sampler vocabulary. KSampler (Extras) is that refusal made into a node: a drop-in KSampler with extra samplers and schedulers you won't find in the stock dropdown. The headline additions are the A1111-style euler_a_a1111 sampler (the Euler-ancestral behavior WebUI users know and love), the automatic_a1111 scheduler, AYS schedules for SD1 and SDXL, GITS, and a local implementation of the RES4LYF beta57 preset.
What's actually in the dropdowns
Everything a normal KSampler has - model, seed, steps (20), cfg (8), positive/negative, latent, denoise (1) - plus extras on the sampler and scheduler lists:
euler_a_a1111(sampler): k-diffusion-style Euler ancestral, reimplemented to match A1111's behavior. If your reference images came with "Euler a" in their settings and you want to reproduce that look, this is the port.AYS SD1/AYS SDXL(schedulers): the "Align Your Steps" noise-level schedules, which hit good quality at low step counts on their respective model families. A legitimately useful quick-win if you're on SD1.5 or SDXL.GITS(scheduler): a newer schedule from the GITS paper line - worth a test, especially at low steps.beta57(scheduler): the RES4LYF beta schedule (alpha 0.5, beta 0.7), reimplemented locally. This is the schedule that 2026 anime and flow-matching workflows name as a default, and it's the reason people install RES4LYF without always knowing they have. Having it inside a stock-feeling KSampler is handy if you don't want the full RES4LYF pack.automatic_a1111(scheduler): approximates A1111's default sigma schedule.
The node keeps Comfy's normal seed handling, partial-denoise behavior, and progress callbacks - so it drops into existing workflows without changing your habits.
Who this is for
Honestly? Mostly WebUI refugees, and people copying settings from shared workflows that name these schedulers. If you never used A1111, euler_a_a1111 vs plain euler_a is a subtle fidelity difference you may never notice, and the extra scheduler names are just more knobs to be confused by - the stock dropdown plus RES4LYF's own pack covers the flow-matching crowd. But if you've got a workflow that says "Euler a, 28 steps" and want exactly that, this node is the cleanest way to get it without installing a second pack.
One framing from the KB worth keeping: samplers split into converging and ancestral families, and the architecture matters more than the name - Karras-style schedules that shine on SD1.5/SDXL can fail outright on flow-matching models. beta57 and the AYS schedules are where the interesting low-step behavior lives.
Install
Ships in the SimpleSyrup pack. ComfyUI Manager → search SimpleSyrup → Install → restart, or:
Set-Location ComfyUI\custom_nodes
git clone https://github.com/Artificial-Sweetener/SimpleSyrup.git
Set-Location SimpleSyrup
..\..\venv\Scripts\python.exe -m pip install -r requirements.txt
Restart ComfyUI; current build required (v3 extension API).
Common issues
- "Euler a" result doesn't match my old WebUI render - exact pixel matches between implementations are rare; settings transfer, seeds don't.
- Schedule names missing - you're looking at a plain KSampler, not the Extras node. The extras only exist in this one.
- Node missing entirely - update ComfyUI first.
It's the sampler for people who left home but kept the keys.
Inputs (10)
| Name | Type | Default | Description |
|---|---|---|---|
| model | MODEL | Diffusion model used to denoise the input latent. | |
| seed | INT | 00–18446744073709550000 | Seed used to create sampling noise. Reusing it with matching settings makes results repeatable. |
| steps | INT | 201–10000 | Number of denoising steps. More steps can add refinement but take longer. |
| cfg | FLOAT | 8.00–100 | Prompt guidance strength. Higher values follow the positive prompt more strongly but can look overcooked. |
| sampler_name | COMBO | Sampling algorithm. It affects the image's look, speed, and stability. | |
| scheduler | COMBO | Noise schedule used during sampling. It changes how quickly structure and detail form. | |
| positive | CONDITIONING,CONDITIONING_BATCH | Positive conditioning that guides what the sampler should add. | |
| negative | CONDITIONING,CONDITIONING_BATCH | Negative conditioning that guides what the sampler should avoid. | |
| latent_image | LATENT | Latent input whose samples will be denoised. | |
| denoise | FLOAT | 1.000–1 | Sampling strength. Lower values preserve the input more; higher values allow larger changes. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| LATENT | LATENT | Denoised latent for VAE decode or more latent processing. |