225% uses the 2x upscaling model, so no need for multiple passes. I do experiment with different passes, sometimes it’s better, sometimes not, it all depends on the source video and which AI model I select.
2 Likes
Thanks Johnny
I’m glad it works for you, but keep in mind, one man’s successfully upscaled video is another man’s disaster. Software should work across the board for everyone, not for a select few. Sorry, but I don’t agree with you.
i still don’t get it. i see no change in the workflow of this application big enough to change the skill required to use the application from “easy to use” to “requiring some sort of degree”. can you go into further detail on what exactly you’re trying to do that became so much harder to do in 3.x?
just for the record, i wouldn’t call myself an advanced user of this app either. i don’t even see what’s the big deal about the version 3.x, other than the refreshed UI. i see no further improvements in functionality, besides added H265 support (which would be a lot better if it presumably didn’t force you into GPU encoding, i’d rather do CPU encoding)
Just wanted to thank the developers for their hard work improving VEAI I was getting 11 FPS when upscaling a 480p to 1080p in version 3.0.6 and with 3.1.0 I’m getting a solid 26 FPS so the speed is really improved all around. Thank you again.
4 Likes
But the mess is still there.
ADD NOISE 0 and 10 give exactly the same result.
oh yeah, i missed that. seems like a minor inconvenience though, nothing a tiny non-mathematician-degree math cannot solve. should absolutely be fixed anyway
The servers simply download the latest rendering instructions. They all live in a common folder. The application, TVAI, checks for the latest ones. If the server has nothing new to download, the application just uses the ones from the folder. If it’s new, it gets downloaded.
Formerly, there was a control to turn this behavior on and off in preferences.
I disagree. Every time I use the same model - it is always downloading it. Yet nothing has changed in my image quality. They remain identical.
I would use a different enhancement model, maybe Proteus or Gaia HQ.
same result with proteus ;-( if i enable motion debleur themis
Your reasoning is flawed. That’s not how it works thank gawd.
Currently, only 2x4x models are available, but I would like a 3x model.
During the transition to FHD, there were many 720p HalfHD due to hardware limitations.
Considering upscaling to 4K (2160p), I think that 3x from 720p is more efficient than 3x from 4x with a reduction.
1 Like
How do you get those speeds? I use Mac M1 Pro and and set my processing to that (as opposed to CPU) but wonder what I have to do to get the speeds you have. Mine are usually between 6 and 7 fps.
The new 3.1 version still loads my RTX 3060 GPU just between 20-40% !? I dont believe this is a good resources utilization . My OS is Win10_64.
Hi… I am trying the new 3.1.0 with some files, but I have problems with Proteus that are not present in version 3.0.11. It’s like dehalo is 100 and you can’t change it
These are the drive links with the two files for comparison
What happens if you use just Themis, without other enhancement added?
1 Like
Actually, the 2.x versions worked exactly the same way as version 3.x, for scaling.
2 Likes
It could be my setup. I use a 13900k cpu and a 4090 gpu. If I run a single process I could get anywhere between 33-35 FPS. If I run two instances at the same time (ex: upscaling two episodes at the same time) I’m getting a good 25-26 FPS each. I have the memory at 90%. It used to take a about an 2:15 hours to upscale a 24 min episode from 480p to 1080p. Now I can convert the FPS and upscale it same resolution and it takes me about 40 minutes tops if I process two at the same time. Less if I’m only upscaling one.
1 Like
