Nodes/RocM Ninodes/ROCm VAE Decode
ComfyUI Node

ROCm VAE Decode

ROCM-optimized VAE Decode for AMD GPUs

By iGavroche·Created 10 months ago·Updated about a month ago· 37
ROCm VAE Decode
  • samples
  • vae
  • IMAGE
tile_size768
overlap96
use_rocm_optimizationstrue
precision_modeauto
batch_optimizationtrue
compatibility_modefalse
enable_temporal_tilingauto
temporal_chunk_size16
temporal_overlap2
last_frame_fixfalse
CategoryROCm Ninodes/VAE

Inputs (12)

NameTypeDefaultDescription
samplesLATENTThe latent to be decoded.
vaeVAEThe VAE model used for decoding the latent.
tile_sizeINT768256–2048Tile size. Larger values use more VRAM but are faster.
overlapINT9632–512Overlap between tiles. Higher values reduce artifacts but use more VRAM.
use_rocm_optimizationsBOOLEANtrueEnable ROCm-specific optimizations for AMD GPUs
precision_modeCOMBOautoPrecision mode. 'auto' selects optimal for your GPU.
batch_optimizationBOOLEANtrueEnable batch processing optimizations
compatibility_modeoptBOOLEANfalseEnable stock ComfyUI compatibility mode (disables all ROCm optimizations)
enable_temporal_tilingoptCOMBOautoFor LTX/WAN videos: 'auto' enables temporal tiling for large outputs, 'enable' forces it on, 'disable' turns it off.
temporal_chunk_sizeoptINT164–256Temporal tile size in LATENT frames. 16 latent = ~121 output frames for LTX.
temporal_overlapoptINT21–8Overlap in LATENT frames between temporal tiles. Higher = smoother but more overhead.
last_frame_fixoptBOOLEANfalseRepeat last latent frame before decode, then discard extra output frames. Fixes end artifacts.

Outputs (1)

NameTypeDescription
IMAGEIMAGE