First I wasn’t sure about de V3 interface, but with the lastest version of V3 its already quiet good. Of course it’s a different behaviour, but the possibility to run multiple task at the same time is a massive plus I think. Even I haven’t found a way to start transcoding multiple imported video at the same time, but maybe this issue lies on my site 
Comparing the v2 production release to the v3 alpha is pointless as they are at completely different development stages.
Alpha means very early version and the v3 alpha is not meant for production use, it’s being built as a preview of the v3 series, which means complete architectural changes to the way processing is being done. I’m sure the GUI will change and improve as it reaches beta and release stages. v2 will never be capable of the features that v3 will bring.
You can prefer v2 if you want but your assertions about v2 being developed further are misguided, Suraj detailed in his ‘back to the future’ post exactly why the v2 series does not have a future. Go read it again if you want to know why v3 is being developed.
Personally I don’t see why invest so much time and resources on a new GUI. As a new I/O is written we should just have more export options, that’s all. All the rest is fine.
The preview thing in V3 is just a regression to the current one since it requires to first render a temporary video file where previously all was done in real time and stored in the memory
VEAI should stayed focused on its main purpose : upscaling videos and not trying to be some media encoder.
It’s now been months since we haven’t have any model improvements and that’s I think the majority of users want rather than some fancy GUI or new functions that have nothing to do with the upscaling process.
Gigapixel AI has now a face recovery AI and I’d rather prefer a V3 having a similar feature than some new codecs to export and a new GUI.
10 Likes
tootaly agree with you.
In terms of development direction, I think the decision to use the VEAI engine as a filter for ffmpeg is a good one. By leaving the reading and writing of video to ffmpeg, stable operation can be expected.
However, I would feel uncomfortable if it were said that the filter was used to increase GPU utilization in parallel processing.
Multiple processing of short videos is fine, but for long videos, it is time-consuming to split videos in advance and merge them later.
I think multiple processing is important, but is there any way to improve the efficiency of single processing?
For example, it seems that the conversion to veai_up color space (BGR48) is done on CPU (swscale) and transferred to GPU, but if it is done on GPU, it can be transferred as YUV420, which may increase the efficiency of CPU to GPU transfer.
(I don’t know if this is true or not, but I got that impression from ffmpeg error messages, etc.)
Also, there are 1x, 2x, and 4x processing models, and if you want to expand 720p (half HD) to 2160p (4K), you have to do 4x and then bicubic reduction
This is disadvantageous in terms of processing time and image quality.
I look forward to the veai_up filter update, but I feel like all the GUI updates are taking up GUI development resources, which is why v3.0 is criticized.
5 Likes
I don’t mind the new GUI I kind of like the direction it’s taking (I don’t care much about the presets, I go straight into settings and pick what I want and go). At least the selections on the right hand side panel.
Now, it’s still buggy (no working progress bar, preview doesn’t happen until complete for me) and the rendering times are much longer (I see much less CPU/GPU usage than with v2). And the models always need to be improved.
It’s an Alpha release, I will continue to test, but I agree with comments made in this forum, other aspects need to improve after we have provided sufficient feedback on the GUI. I keep v2 for my regular use.
Thanks!
1 Like
I like the new gui. Especially that all previews are saved now and i can switch between them and not loosing the settings. great job!
One thing I would love is if they could figure out how to not have scrolling within scrolling for the filters side bar. Just look at what Lightroom does for its right side panel. Each section can be expanded or collapsed – and when expanded it takes up its full size. There is no scrolling within the section itself.
Today with the alpha I have to carefully position my mouse within a section or within the overall panel to get the right thing to scroll. It’s finicky and annoying 
1 Like
I agree that would be awesome if the new V3 had the face recovery feature.
3 Likes
I was giving it grace because its in alpha, I had their first alpha release and then jumped to the newest and the newest one looks way better than it did (3.0.0.31a). But, I have mixed feelings still with the latest alpha vs v2 appearance wise. The newest alpha feels more clean than V2 and simplified in ways, I will say that, but… it lacks some character V2 had. V2 still seems more feature rich and customizable.
I kind of like V3’s sidebar showing me everything at once (organization) and minimized until I click it and then within it I can change its settings. But it’s lacking features like:
- Crop
- Constant Rate Factor
- Keep Audio
- Grain
- Model downloading and a wider selection
- Comparison Preview
- Doesn’t have recommendations
- Doesn’t have a view button
- The side bar gets busy when I enable a setting. I would prefer it minimize and still be enabled when I am done with it.
Overall, I wouldn’t use V3 as it stands over V2. I like its organization better but it lacks many features, maybe the grey scale UI makes it less welcoming, but at the same time it appears modern. Time will tell, maybe Topaz with create a UI that we can tweak to our liking. Who knows.
I completely agree with the original post. I love VEAI and consider its results magic, but I find the new 3.0 GUI virtually unusable. My biggest complaint is the complete lack of visual feedback while processing. It’s very difficult to understand what the program is doing until after it finishes rendering. The 2.0 GUI shows each frame as it’s rendered which is very helpful in detecting problems in certain sections of a video. I work at large raster sizes and with the 3.0 GUI I can’t tell what’s going on until processing is finished. That can take hours! I understand the technical reasons for the new GUI but I really think it needs to be rethought to provide better immediate feedback and be easier to use.
4 Likes
I thought it was just me. I have cancelled the processing so many times thanks to a working preview while processing.
3 Likes
The option to download all the models for the current processing unit is also gone, right?
An option to remove the consecutive repeating frames while retaining audio synchronization - would that kind of processing find its way here, too?
It’s applies to videos recorded with OBS for example - sometimes the game freezes for a fraction of seconds, and it’s seen like a stutter on otherwise smooth game video later, that’s why I’m suggesting it.
main dev suraj already answered many times about this. To be able to display a real time preview, it needs a frame per frame preview saving, actually the preview is saved as video. this is just temporary, but i’m pretty sure the real time preview / frame is very high in the list. I agree it’s very frustrating, but it will be there for sure before the release, and i hope, even before.
about the overall Gui, i would have prefered one identical to what topaz do, like the one on Photo Ai. it’s better looking for me but it’s pure personnal opinion.
