Nodes/Halo-TeaCache/Halo-TeaCache v1.0.0 (AMD LTX2)
ComfyUI Node

Halo-TeaCache v1.0.0 (AMD LTX2)

A ComfyUI node in Halo-TeaCache with 4 inputs and 1 output.

By bkpaine1·Created 7 months ago·Updated 7 months ago· 4
Halo-TeaCache v1.0.0 (AMD LTX2)
  • model
  • model
rel_l1_thresh0.20
start_percent0.15
end_percent1.00
CategoryHalo-TeaCache

Inputs (4)

NameTypeDefaultDescription
modelMODELThe LTX2 diffusion model to accelerate.
rel_l1_threshFLOAT0.200–2Cache threshold - higher = more skipping (faster but lower quality). 0 = disabled. Try 0.15-0.30.
start_percentFLOAT0.150–1Start caching after this % of steps (early steps need full compute).
end_percentFLOAT1.000–1Stop caching after this % of steps.

Outputs (1)

NameTypeDescription
modelMODEL