Haven't really tested this a lot, (I would probably need multiple faster gpus with lots of vram to really have fun with it) I'm wondering:
a. are there significant differences between the different SeedVR2 models?
b. what experiences do you have in regard to what prefiltering should be done before feeding a source to SeedVR2?
c. is it just me or does the 3b model somehow create somewhat strange/creepy skin?
d. is there
Cu Selur
----
Dev versions are in the 'experimental'-folder of my GoogleDrive, which is linked on the download page.
I tried the workflow SeedVR2 1080p-template available in ComfyUI (see attachment).
On my RTX5070Ti it was able to resize a small clip of size 640x504 to 1370x1080 at a speed of about 2.0 fps, using only the 16GB of VRAM (shared VRAM was not used).
Quote:I was unable to get "x4 ESRGAN" working on my side
Strange works fine here (in Hybrid).
NNEDI3:
ESRGAN-x2plus:
ESRGAN-x4:
Waifu2x:
RealESRGAN-OpenProteus
BasicVSR++ Vimeo-90-K-BD:
RealSR:
SRMD:
SwinIR Lx4:
SwinIR Sx4:
SeedVSR2:
(note: SeedVSR2 has slightliy different resolution, since the PAR wasn't adjusted)
all files: https://www.mediafire.com/folder/sj4adnb...omparision
I think 'sharper' is the wrong term.
To me SeedVR2 does really sharpen more, but does add (invent) more details.
Cu Selur
----
Dev versions are in the 'experimental'-folder of my GoogleDrive, which is linked on the download page.
02.02.2026, 17:27 (This post was last modified: 02.02.2026, 17:45 by Selur.)
Any suggestion how to limit normal RAM (not VRAM) usage?
Wanted to filter some 1080p content, not really upscale, but I get: DefaultCPUAllocator: not enough memory: you tried to allocate 59719680000 bytes.
Cu Selur
Ps.: just noticed my ComfyUI is 0.6 and current is 0.10 => updating now, maybe that fixes the issue
----
Dev versions are in the 'experimental'-folder of my GoogleDrive, which is linked on the download page.