Nodes/velvet-vice-ltx/VELVET VICE LTX — FP16 Chunked Image Scale By
ComfyUI Node

VELVET VICE LTX — FP16 Chunked Image Scale By

Scales complete frames in small batches. Each active chunk is calculated with ComfyUI's native common_upscale routine in FP32 and written directly into one FP16 output tensor. It never spatially tiles or temporally interpolates the clip.

By Velvet-Vice·Created 5 days ago·Updated 5 days ago· 1
VELVET VICE LTX — FP16 Chunked Image Scale By
  • image
  • image
upscale_methodbicubic
scale_by1.00
frame_chunk_size4
CategoryVELVET VICE/LTX

Inputs (4)

NameTypeDefaultDescription
imageIMAGE
upscale_methodCOMBObicubic5 options: nearest-exact, bilinear, area, bicubic, lanczos
scale_byFLOAT1.000.01–8
frame_chunk_sizeINT41–64

Outputs (1)

NameTypeDescription
imageIMAGE