Nodes/ComfyUI-Grok-SmartVAE/🎬 Universal VAE Decode (v11.3 + Ignore Warnings)
ComfyUI Node

🎬 Universal VAE Decode (v11.3 + Ignore Warnings)

A ComfyUI node in latent/video with 10 inputs and 1 output.

By uczensokratesaΒ·Created 7 months agoΒ·Updated 3 months agoΒ· 17
🎬 Universal VAE Decode (v11.3 + Ignore Warnings)
  • samples
  • vae
  • IMAGE
β—„frames_per_batch8β–Ί
β—„overlap_frames2β–Ί
β—„force_time_scale0β–Ί
β—„enable_tilingfalseβ–Ί
β—„tile_size512β–Ί
β—„verbosefalseβ–Ί
β—„max_ram_frames500β–Ί
β—„ignore_warningsnoneβ–Ί
Categorylatent/video

Inputs (10)

NameTypeDefaultDescription
samplesLATENTβ€”
vaeVAEβ€”
frames_per_batchINT81–128Frames to decode per batch. Auto-reduces on OOM.
overlap_framesoptINT20–16Temporal overlap for seamless stitching.
force_time_scaleoptINT00–16Manual time scale override (0=auto). E.g., 8 for LTX-Video.
enable_tilingoptBOOLEANfalseForce spatial tiling. Auto-enables on OOM.
tile_sizeoptINT512256–2048Tile size in pixels for spatial tiling.
verboseoptBOOLEANfalseShow detailed progress logs.
max_ram_framesoptINT500100–10000Threshold for disk offload. Tune to system RAM: 16GB=300, 32GB=600, 64GB=1000.
ignore_warningsoptCOMBOnoneCorrupted latent handling: β€’ none = Stop on NaN/corruption (safest) β€’ minor = Try if <10% corrupt (may produce black frames) β€’ all = Force decode anyway (high crash/black frames risk)

Outputs (1)

NameTypeDescription
IMAGEIMAGEβ€”