You are not logged in. Please login or register.
SmoothVideo Project → Using SVP → New RIFE filter - 3x faster AI interpolation possible in SVP!!!
How can I test R70 vapoursynth instead of R64 provided for svp, and you have plan to updated ? I dunno if there a real benefit but a think so many things fix/changed on it , that a like to try it. So if someone get me pointers , it will be very appreciated.
> I dunno if there a real benefit
nope
> How can I test R70 vapoursynth
put portable release into SVP 4/mpv64
> I dunno if there a real benefit
nope
> How can I test R70 vapoursynth
put portable release into SVP 4/mpv64
Thanks for respond, which file do a put into mpv64 folder, if copy the entire folder i get unable to initialize vapoursynth api 4.0 error
all files
dunno which Python version it wants, probably you have to replace portable Python too
all files
dunno which Python version it wants, probably you have to replace portable Python too
Thanks, I think better leave like it is, wait until you decided update on the future.
For those interesting Model 4.26_heavy its out , for me at least don't work as well as 4.25, maybe for those with 4090 will.
4.26_heavy its out
These files should allow SVP to find files ending with "_large.onnx".
Does changing these as suggested in vsmlrt.py make sense for RTX 40XX gpus?
If yes... which numbers are recommended for 'Streams' and 'Workspace'?
any other recommendations for the other parameters?
Does changing these as suggested in vsmlrt.py make sense for RTX 40XX gpus?
If yes... which numbers are recommended for 'Streams' and 'Workspace'?
any other recommendations for the other parameters?
What changes?
narkohol wrote:Does changing these as suggested in vsmlrt.py make sense for RTX 40XX gpus?
If yes... which numbers are recommended for 'Streams' and 'Workspace'?
any other recommendations for the other parameters?
What changes?
Those specified in that header as 'basic performance tuning':
set fp16 = True (on RTX GPUs)
increase num_streams
increase workspace
set use_cuda_graph = True
That by default are:
fp16: bool = False
num_streams: int = 1
workspace: typing.Optional[int] = None
use_cuda_graph: bool = False
num_streams comes from the RIFE profile
force_fp16 = True, tf32 = True, use_cuda_graph = True
come from helpers.py
num_streams comes from the RIFE profile
force_fp16 = True, tf32 = True, use_cuda_graph = True
come from helpers.py
What about 'Workspace'?
vsmlrt.py 'performance tuning' header says "increase workspace", but in helpers.py it is set to backend.workspace = None
Chainik wrote:num_streams comes from the RIFE profile
force_fp16 = True, tf32 = True, use_cuda_graph = True
come from helpers.pyWhat about 'Workspace'?
vsmlrt.py 'performance tuning' header says "increase workspace", but in helpers.py it is set to backend.workspace = None
"The maximum workspace is set to None for the total memory size of the GPU."
none is the best
HELLLLLOOOOO
Good evening I'll stop you for two seconds, Excuse me
I dropped my post here
For those curious about image interpolation models rife is not the only one that exists for video
All that remains is to try to make them work in real time
A PLUS DANS LE BUS
https://www.svp-team.com/forum/viewtopic.php?id=7466
hello, when I use the transcode feature, *2 gives me 24fps for a 4k video (for the speed of generation, so it needs twice the length of the video), but anything above (like 2.5*, 3*, or 4*)will give me only 14fps.
I'm using a rtx 2080
What speed do you get and do you also get lower speed at anything higher than 2x ?
x2.5 and x3 are two times slower than x2
thanks, for those with rtx 4090, what fps do you get when transcoding 4k video ?
thanks, for those with rtx 4090, what fps do you get when transcoding 4k video ?
The RIFE model dependant!
Older the model - better performance, newer backwards!
My RTX 4070 can handle 4K source, transcoding 24>48 sequence at 48FPS speed - 4.4v2 performance!
i'm with 4.25v2, you're using 4.4v2 ?
i'm with 4.25v2, you're using 4.4v2 ?
The difference between 4.4 and 4.25 twise as much by performance, but without artefacts! Thats was example!
When the system accurately detects a scene cut, my understanding is the final frame is duplicated. On a fast panning shot, that comes across as a stutter as the motion is clearly slower on the final frame. Is there any combination of settings that prevents/minimises that from happening? Currently using 4.15 as my benchmark
oriento wrote:i'm with 4.25v2, you're using 4.4v2 ?
The difference between 4.4 and 4.25 twise as much by performance, but without artefacts! Thats was example!
Where can I find the 4.4v2 model?
I would like to try with him, with the 4.25vs model my speed is 20fps and I don't see that model to select, I only see 4.4
Thank you so much.
Where can I find the 4.4v2 model?
I would like to try with him, with the 4.25vs model my speed is 20fps and I don't see that model to select, I only see 4.4Thank you so much.
There is an external link at SVP Wiki page in RIFE chapter
Download there
raider10 wrote:Where can I find the 4.4v2 model?
I would like to try with him, with the 4.25vs model my speed is 20fps and I don't see that model to select, I only see 4.4Thank you so much.
There is an external link at SVP Wiki page in RIFE chapter
Download there
Thanks, but 4.4v2 version is missing 
Honza wrote:raider10 wrote:Where can I find the 4.4v2 model?
I would like to try with him, with the 4.25vs model my speed is 20fps and I don't see that model to select, I only see 4.4Thank you so much.
There is an external link at SVP Wiki page in RIFE chapter
Download thereThanks, but 4.4v2 version is missing
SmoothVideo Project → Using SVP → New RIFE filter - 3x faster AI interpolation possible in SVP!!!
Powered by PunBB, supported by Informer Technologies, Inc.