Nodes/ComfyUI Qwen Rectified Flow Tools/Latent Hybrid Inverter (Qwen)
ComfyUI Node

Latent Hybrid Inverter (Qwen)

A ComfyUI node in Qwen/Sampling with 12 inputs and 1 output.

By ttulttul·Created 11 months ago·Updated 7 months ago· 7
Latent Hybrid Inverter (Qwen)
  • model
  • latent_image
  • positive
  • LATENT
steps20
strength0.50
blend_factor0.50
inverter_seed0
forward_diffusion_seed1
velocity_amplification1.00
velocity_perturb_strength0.20
internal_precisionbfloat16
normalize_outputenable
CategoryQwen/Sampling

Inputs (12)

NameTypeDefaultDescription
modelMODELUNet model used for both inversion and diffusion passes.
latent_imageLATENTLatent tensor to invert and re-noise.
positiveCONDITIONINGPositive prompt conditioning that guides both passes.
stepsINT201–1000Solver steps executed for inversion and forward diffusion.
strengthFLOAT0.500–1Strength for both inversion and forward diffusion. The target noise level.
blend_factorFLOAT0.500–10.0 = 100% Forward Diffusion (Stable), 1.0 = 100% Inverter (Creative)
inverter_seedINT00–18446744073709550000Seed for the creative, perturbed inversion process.
forward_diffusion_seedINT10–18446744073709550000Seed for the stable, 'perfect' noise anchor.
velocity_amplificationFLOAT1.00-0.9–2Deterministically scales the predicted velocity each step.
velocity_perturb_strengthFLOAT0.200–5Random velocity noise level for the inverter stage.
internal_precisionCOMBObfloat16Precision used for inverter UNet autocast.
normalize_outputCOMBOenableNormalizing the inverter output is highly recommended.

Outputs (1)

NameTypeDescription
LATENTLATENT