V3 critiques

I know there’s a couple of threads on V3 but I didn’t want to hijack since my critiques are a bit different. I did not like V3 in beta so I dropped it for V2 until about a week ago. Still a bit buggy but it fixed a couple of longstanding issues and last week (before V3.03) it actually seemed faster than V2.

With version 2.anything there has always been a problem pretty much any models with particular interlaced DVDs. They show up as 60FPS from the source and no matter what Topaz as well as most other deinterlacers cause duplicate frames, but not every other frame…like every 3rd or 4th. This is a huge PAI because no matter what the end framerate they end up choppy. and since it’s not every other frame I can’t just re-encode them out

I tried V3 just prior to 3.03, it completely resolved this issue, even though it would be great if ddv and dtv could work without duplicating frames. It’s great that it interpolates 30fps interlaced into 60 flawlessly, but I don’t need 60fps source to wind up 120fps. Dione Robust does not look the same as DDV so I don’t like using it. I can re-encode 120fps back to 60fps after the fact, I’m just happy the duplicate issue is resolved in v3. The fact that DDV/DTV keeps audio is a huge step up. Didn’t make any sense why V2 removed the audio from the video. The original audio always synced back up, it was just a PIA to have to extract the audio and re-add it afterwards.

I did notice after the update that 3.03 is quite a bit slower than 3.02, I was actually surprised how fast 3.02 was so I assume they can fix it. I can kind of tell someone else has developed some or most of V3 because they seem to have missed a couple of things that are important:

It’s great that you can run multiple processes in 1 instance. With V2 I ran multiple instances. However, they messed up by not allowing previews to be a separate process from the exports. The export thing as it’s laid out is more confusing than the original setup, so as soon as you get the settings going for that clip you want to immediately hit export to get that clip out of the way. It is serviceable like that but the problem is you can no longer run a preview at all because your processes are used up as exports that are running. You have to allow a single preview to generate regardless of how many export processes are running or the workflow becomes a mess because of how the input bin is setup. It’s tolerable if you can set a clip up and not have to go back to it, but still be able to check previews on following videos. I stick to 2 processes because 3 slows stuff down. If I could be running 2 exports and a single preview it would be fine because I could keep moving to different clips. Selecting 3 processes does not help because it will just run 3 exports slower than I’d like and I still can’t preview the 4th.

I also don’t understand why V3.03 holds onto videos after being processed. You can’t delete them because VEAI is holding them open. I immediately transfer videos to a NAS after completion. I can’t delete them until after I close VEAI.

I have 2 systems, 1 with 2 1080TI’s and 1 with a 6900xt. They both seemed pretty snappy last week on V3.02. I updated to 3.03 and the 2x 1080TI system took a slight hit but the 6900xt took a huge hit, had to go back to V2 on the 6900xt.

Also the naming is just odd. V2’s naming was much better because I could know exactly what was done to the clip. I guess V3 is like a timestamp? Not a fan of this.

All in all I think V3 is getting better. Starting to prefer it over V2, devs should definitely target these issues for the next release though.


BTW, allowing framerate interpolation to run with the AI model is a huge improvement over V2 as well. Chronos on V2 was so slow I never used it anyway. I generally used Flowframes after VEAI on 30FPS non interlaced stuff. V3.02 seemed to be able to run Apollo + GAIA without a huge time increase and I think that’s why I decided to use V3. V3.03 is slow for some reason in general but since V3.02 was fine I assume it can be fixed.