The preview is displaying correctly for PNG output. At least for Proteus. I cannot say the same for interpolation models. They look like the preview is just the original copied.
Just an observation. It does not hinder me in any way.
Hi, I have a bug report for 4.08/Windows 11/RTX GPU.
Is this the correct place to put them?
The timeline looks to have a wider âminimum widthâ than the edit area.
Reproduce by stretching main app window to minimum window width. You will see the timeline is overlapping with the settings column.
Windows 11, GTX 3060, Video AI 4.0.8 â
Iâm trying to download all of the Models but (Frame Interpolation) Apollo & Apollo fast, and (Enhancement) Artemis - Aliasing or Moire⌠all show that theyâve successfully downloaded, but going back into the Model Manager theyâre unchecked. Iâve tried restarting and updating, but it still shows âModel (22/25)â on my device. When I try to re-download them, they all download fine but fail to show as downloadedâŚ
Rolled back to 3.5 after the first major 4.0 update was released and have been holding onto it until Iâve seen some of the major issues with v4 are resolved. From what Iâve seen, its mostly been UI updates to improve the shit storm of broken features 4.0 brought. Glad to see live previews back. Can anyone tell me if its worth updating yet?
On a side note, I think the main reason most of us are using TVAI, are the models. Iâd rather see more emphasis on improving models than workflow/UI. Of course once UI features are broken, they need to be fixed, so understandable thatâs a priority right now, but Topaz, please stop trying to make this thing a shiny, pro-looking, eye catcher and work on the AI engine. Sometimes less is more. Back to basics, please, and the fundamental reason we use this product.
Speaking of models downloaded, installing v 4.0.8 and answering âyesâ to the âRemove Old Models?â prompt resulted an empty models folder. All gone. The lot. None of them âold modelsâ. Unhelpful if youâve previously downloaded all the models to allow offline working.
The prompt wording needs to be improved if itâs removing all models, to make that clear.
Many other tools out there for processing video, most of which are superior to Topaz. Trying to make this a one-size fits all application, wasnât such a good idea after all. Putting all their energy and resources into AI/models should be their primary goal, while leaving the other processing tasks to other specialized tools. It started out that way in 2020, but it seems the application has evolved into a behemoth, with a lot of unnecessary bells and whistles.
Since 4.0.7 I have not reverted back to 3.5.4 despite that Iris and Protheus can look like Craquele when enhancing old noisy videos that have been digitized to mp4
having said that Artemis and Gaya work and on Good Source Material from your mobile Iris is fine
Not sure what Protheus 4.0 is intended perfect source material or lousy noisy stuff
100%. I started using it with the 2.0 release and the simplicity and purpose was singular. It was an almost idiot proof way to restore and enhance video with AI processing. It seems somewhere along the way, the vison and focus was diverted away from improving the AI modelling, into what basically equates to UI gift wrap. The selling point of this product is AI enhancement. We donât need bells and whistles, we just need a simple way to enhance with better AI modeling. Leave the video editor workstation improvements to the big boys that excel at it and stick to what this was intiially intended to be, a simple, task free way to recover, restore and enhance video through AI technology.
Has the upscaling processing time improved? That was one of the biggest drawbacks of v4+ vs 3.5.4. Processing time all but doubled when upscaling the same video at the same frame rate., which made it unusuable for me without the ability to pause/resume a process. What was taking an hour or less to upscale 480p to 1080p for an hour long video, was now taking 2 to 3 times that or more depending on the model. I was getting about 20-25FPS average with Iris v2, and it dropped to like 6-8FPS with v4.
