Nodes/BSAI-Qwen-Prompt-Enhancer/BSAI Qwen image 2.1 Sampler
ComfyUI Node

BSAI Qwen image 2.1 Sampler

BSAI Qwen image 2.1 sampler: all KSampler parameters (seed/steps/CFG/sampler/scheduler/denoise) are kept as widgets, and each can be right-clicked 'Convert widget to input' to wire an upstream node (e.g. the enhancer's RECOMMENDED_STEPS/CFG). Adds optional guider / sigmas input ports to wire distilled-sampling upstreams such as BasicGuider / ViggleTurboSigmas directly. Steps rule: steps always wins — external sigmas only apply when their step count matches steps; when they mismatch, external sigmas are ignored and regenerated from steps (for a distilled LoRA keep steps equal to the LoRA/sigma sequence length).

By xm6018924·Created about a month ago·Updated 3 days ago· 14
BSAI Qwen image 2.1 Sampler
  • model
  • positive
  • negative
  • latent_image
  • guider
  • sigmas
  • LATENT
  • denoised_output
◄seed0►
◄steps20►
◄cfg1.0►
◄sampler_nameeuler►
◄schedulersimple►
◄denoise1.00►
Categorysampling/custom_sampling

Inputs (12)

NameTypeDefaultDescription
modelMODEL—
positiveCONDITIONING—
negativeCONDITIONING—
latent_imageLATENT—
seedoptINT00–18446744073709550000—
stepsoptINT201–10000—
cfgoptFLOAT1.00–100—
sampler_nameoptCOMBOeuler44 options: euler, euler_cfg_pp, euler_ancestral, euler_ancestral_cfg_pp, heun, heunpp2, +38
scheduleroptCOMBOsimple9 options: simple, sgm_uniform, karras, exponential, ddim_uniform, beta, +3
denoiseoptFLOAT1.000–1—
guideroptGUIDER—
sigmasoptSIGMAS—

Outputs (2)

NameTypeDescription
LATENTLATENT—
denoised_outputLATENT—