Sampler Parameter Packer
Carry your sampler and scheduler as one object
- sampler_params
Samplers and schedulers are two separate dropdowns that always travel together, and Sampler Parameter Packer is the Flux Continuum pack's answer to carrying them as one unit. Pick your sampler, pick your scheduler, and the node bundles both into a single SAMPLER_PARAMS object you can route through one wire. Its companion - Sampler Parameter Unpacker - splits it back out wherever you actually need the individual values.
How it works
The mechanism is trivial in the best way. The node reads ComfyUI's own lists of valid samplers and schedulers at runtime (comfy.samplers.KSampler.SAMPLERS and ...SCHEDULERS), so the dropdowns always match what your ComfyUI build supports - no hardcoded lists to go stale. It then packs a 4-tuple: the sampler object, its name string, the scheduler object, and its name string.
That extra name-string redundancy isn't wasted - the unpacker returns all four, so downstream nodes can consume either the raw object (for a KSampler's sampler/scheduler inputs) or the human-readable name (for anything that logs or displays what ran).
Inputs and outputs
- sampler - dropdown of all samplers ComfyUI knows (44 in the schema), e.g.
euler,euler_ancestral,dpm_2. - scheduler - dropdown of all schedulers (9 in the schema):
simple,sgm_uniform,karras,exponential,ddim_uniform,beta,normal,linear_quadratic,kl_optimal.
Output is one sampler_params (SAMPLER_PARAMS). Pair it with Sampler Parameter Unpacker to recover the individual fields, or pass it to any node that understands the type.
What to actually pick
The defaults don't matter as much as the model family. For Flux Dev the standing advice is Euler with normal or beta - and not Karras, which concentrates denoising in the middle of the trajectory and fights a flow-matching model's straight line. The pack's own workflow ships with sensible Flux defaults and lets you swap them here, which is a nice middle ground between "never touch the sampler" and "hand-edit every KSampler in a 10-module graph."
Installing it
Bundled with robertvoy/ComfyUI-Flux-Continuum:
cd ComfyUI/custom_nodes
git clone https://github.com/robertvoy/ComfyUI-Flux-Continuum
or ComfyUI Manager ("Flux Continuum"), restart. No extra dependencies - it uses only core ComfyUI types.
The catch
It only carries sampler + scheduler; steps, denoise, and guidance stay separate (the pack keeps those on their own sliders). So this isn't a "sampler settings in one box" node in the Efficiency Nodes sense - it's specifically the two dropdowns that live on every KSampler. If that's the pairing you keep setting together, it's genuinely convenient; if you expected full KSampler state, you'll be reaching for the other knobs anyway.
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| sampler | COMBO | 44 options: euler, euler_cfg_pp, euler_ancestral, euler_ancestral_cfg_pp, heun, heunpp2, +38 | |
| scheduler | COMBO | 9 options: simple, sgm_uniform, karras, exponential, ddim_uniform, beta, +3 |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| sampler_params | SAMPLER_PARAMS | — |