ComfyUI Node
⚗ Alchemica KSampler Pro (two-phase prompt, previews)
AlchemicaKrea with two-phase prompts (scene/detail), composition previews, refinement variants and a profile chosen by model + resolution.
⚗ Alchemica KSampler Pro (two-phase prompt, previews)
- model
- positive
- latent_image
- positive_detail
- negative
- refiner_model
- opts
- vae
- latent
- sigma_plot
- report
- used_seed
- image
◄seed0►
◄profile▾►
◄modebalanced►
◄two_stagetrue►
◄redraw-1.00►
◄detail0.12►
◄variety0.60►
◄denoise1.00►
◄phase_switch_sigma-1.00►
◄phase_blend0.00►
◄previews0►
◄pick0►
◄refine_variant0►
◄preview_methodlatent2rgb (fast)►
◄degrid0.0►
Category⚗ AlchemicaMente/sampling
Inputs (23)
| Name | Type | Default | Description |
|---|---|---|---|
| model | MODEL | — | |
| positive | CONDITIONING | SCENE prompt (or your only prompt) | |
| latent_image | LATENT | At the final resolution (⚗ Resolution node) | |
| seed | INT | 00–18446744073709550000 | BASE seed. With previews keep it on 'fixed' |
| profile | COMBO | (auto) = profile calibrated on this model at the closest resolution | |
| mode | COMBO | balanced | 3 options: fast, balanced, max detail |
| two_stage | BOOLEAN | true | — |
| redraw | FLOAT | -1.00-1–0.95 | Hand-off sigma between the two stages. -1 = from the profile |
| detail | FLOAT | 0.120–0.25 | Micro-relief. 0.12 validated; above 0.2 coloured speckles come back |
| variety | FLOAT | 0.600–1.2 | Noise that creates texture. 0 = deterministic |
| denoise | FLOAT | 1.000–1 | 1.0 = from scratch. <1 = img2img (no two stages, no previews) |
| phase_switch_sigma | FLOAT | -1.00-1–0.99 | Sigma below which the DETAIL prompt takes over. -1 = where composition is locked (from the profile, ~0.87). Lower = detail weighs less. 0 = off. Acts on skin, fabric, patterns, small objects; NOT on colours or global style (measured: those are decided in the first step together with composition) |
| phase_blend | FLOAT | 0.000–0.4 | Width of the cross-fade between the two prompts. 0 = hard switch (no cost). >0 = gradual blend, costs one extra model call per step inside the band |
| previews | INT | 00–16 | N > 0: generates N low-resolution previews (seed, seed+1, ...), up to ~30 s each. Measured: pose and framing match the final image (0.87-0.90); clothing may change. Then set 0 again and pick the number in 'scelta' (choice) |
| pick | INT | 00–255 | With previews = 0: completes preview number k (uses seed + k) |
| refine_variant | INT | 00–9999 | Changes ONLY the second-stage noise: same pose and framing, but face, lettering, jewellery and small objects change. 0 = standard |
| positive_detailopt | CONDITIONING | DETAIL prompt (from the ⚗ Two-phase prompt or ⚗ Text node). Disconnected = behaves like AlchemicaKrea | |
| negativeopt | CONDITIONING | — | |
| refiner_modelopt | MODEL | Stage 2 with a different model | |
| optsopt | ALCHEMICA_OPTS | — | |
| vaeopt | VAE | Optional: the final image is decoded and SHOWN in this node, and returned on the 'image' output (like the Efficiency KSampler) | |
| preview_methodopt | COMBO | latent2rgb (fast) | Live preview while sampling. latent2rgb = instant, a bit soft; taesd = sharper (needs the TAE model in models/vae_approx, otherwise latent2rgb); 'ComfyUI setting' = whatever ComfyUI is set to (often none) |
| degridopt | FLOAT | 0.00–10 | Only with a VAE connected: removes the faint 2-pixel line grid of Krea 2's VAE from the image shown and returned here (the latent is untouched). Max correction in grey levels: 2 = recommended, 0 = off |
Outputs (5)
| Name | Type | Description |
|---|---|---|
| latent | LATENT | — |
| sigma_plot | IMAGE | — |
| report | STRING | — |
| used_seed | INT | — |
| image | IMAGE | — |