In my opinion, the number of rescaling options are all potentially useful. It may require the users to try a experiment with a few of them to become familiar with what they actually do.
3 Likes
Hi,
V3.0.0 is a nice improvement in general.
However, I want to point out a few things that I think requires additional effortā¦
- There is no real help available. Sure, generic info on things that are kind of self explanatory is available. What I call āsupportā-help, to solve very generic queries.
But in-depth info on for example the output encoder types and what using respective encoder actually means, such information is currently not found in SW or on the webpage support pages. - Iām on M1-pro, and by choosing āApple M1-Proā in the AI Processor list, Iām assuming the relevant processors will be used for relevant operations, which includes using the HW encoder for up-scaling and other video encoding operations.
- Are the HW-encoders at all in use on M1-Pro and M1-Max Macs for encoding?
- If not, why not?
- If not, will it be and when in that case?
- If available, is it required of me to choose a specific Output Encoder for that to happen?
(To me, this kind of information is important, so that I can make the correct settings. Yet, no information is found at all about it, nor does it seem important to talk about in reviews and suchā¦ comically)
Thanks for the word.
The new version seemed like itās a rushed BETA
Itās not faster than Version 2.6.4
I am using a 32core Threadripper and only 5-10% of the power is ever used.
and under 1% of my GPU. Why canāt more CPU/GPU power be used?
This program should be 10X faster on a faster computer.
ALSO why canāt the EXPORT remember the directory to save in like version 2.6.4
I want the program saved in the same folder as the source, but V3 doesnāt do that.
4 Likes
It appears that you will now have to buy two copies of the software as they are advertising purchase includes two seats. The days of unlimited are gone.
Version 3 speed is Horrible, its 4X SLOWER than 3.6.4
Here is my example:
Same program: one minute 480p mp4, Artemis filter output to 1080p
V2.6.4 Total time 5 minutes to complete
v3.0 Total time 20 minutes to complete
Why make a new version 4X slower?
Again I am using a 32core Threadripper, 32gb memory, nVidia gpu
Looks like I need to stick to V2.6.4
(If you need more Beta testers just let me know)
3 Likes
Weāll see. I really never expected to use all three computers running VEAI all the time. I will probably use the 8 core/16 thread machine for a dedicated pre and post processing computer. It has the weakest GPU of the 3 computers and I prefer software encoding. Not the end of the world.
1 Like
Well - I am seeing changes on their website for previous versions that now state 2 licenses (for v 2.6.4) and the updated page was October 20, 2022.
In addition a lot of information regarding previous versions has been removed from their website.
It seems you always need to authenticate your copy of the software with their servers, even if you have pre-downloaded all of the models.
Thank gawd for the WayBack Machine.
Topaz needs to clear up this licensing thing. There are many of us that use this software on multiple machines at the same time.
1 Like
I just ran what I think is similar to your video, 856x480 mp4 1 minute in length.
My computer is a Ryzen 9 3900x, 64Gb Ram and a RTX 3060Ti GPU. My 5600x, 32Gb ram with a RTX 3060(12Gb) is showing the same times as the 3900x.
2.6.4 Artemis V13 MedQual video (856x480 1 minute long) 0.11 sec./frame 2 minutes and 35 seconds
3.0.0 Artemis MedQual video (856x480 1 minute long) 0.11 sec./frame 2 minutes and 29 seconds
1 Like
I am not a regular poster but I had to come here to comment on this release. Itās that bad that I have had to revert to v2.6.4. My system is an Intel i9, 64 gb ram, Samsung SSD drives, Nvidea GeForce 2080 Super graphics and I have never had any issues with previous versions. It has been fantastic to use and I have learnt a lot over the last year to get the best out of it. Using 2.6.4 now, it is flowing smoothly as I watch it doing its work.
3.0.0 is a downgrade. It runs like a beta version, and I am probably being generous in saying that. After cursing your latest version for hours, I just redownloaded and installed 2.6.4. What a breath of fresh air it is after version 3.
I am currently encoding a video. 20 minutes on 2.6.4, yet it took 1 hour and 10 minutes on version 3 and the result was very average. The new interface, while looking more modern, is a PITA. Why not just keep the same options in it so you donāt confuse everyone? And why is it so slow ? And why are previews so jerky and out of synch ?
I donāt even want to start talking about the nightmare preview boxes that fill the screen whenever you want to look at at something.
I wish I could say something positive, but it is an absolute shocker and you need to really look at what you are putting out. If this was a Beta that people could download to test, thatās one thing. I cant believe something this slow and buggy was a general release.
I seriously hope there is a prompt upgrade to 3.0.1 as no doubt there are many like me that are coming up on there first year and will be having a good think about whether its even worthwhile continuing to pay money for upgrades. As it stands, I wont beā¦2.6.4 is a cut above this and it is hard to believe this 2 different versions came from the same company.
Its a bit like the photo software, I own all of your products. Now I get the semi dud Photo AI that is worse than the products it is supposed to replace.
Anyway, thats my feedback. Sorry it wasnt better.
9 Likes
apart from the technical problems people mention: Iām NOT convinced by the āimprovedā UI.It seems that simply that all the handholdy assistants and advisers and whatnot are now forced on me. So selecting the very same features has become more cumbersome due to unnecessary splitting into seperate menus. It feefls bloated but not better.
The Trim function doesnāt do any more than the older one but has now become needlessly cumbersome to use. Instead (thatās not even an option anymore!) of simply timeline click, āiā, click āoā) , you now have to click a button, move markers with the mouse, and then another pointles click to āapplyāā¦ And to change it, all of that again.Itās literally the same but with more fluff.
Yet basic real improvements it are still missing. It saves my last session as autosave. So there is a save function. Why canāt i still not save a session manually with a name then? If you have two instances running and one crashes, you can start from scratch because the backup is uusally from the wong one. Or multiple trim segments (more features would be a real reason for a seperate menu).
Or a pause function.
I guess the idea is to make things easier to understand and find for new users. And they probably do. But many a workflow has been sacrificed for that.The improved workflow claim seems almost sarcastic. It would be great to have the few real advantages (stabilisation, offline mode(!!), new ai modelsā¦ in the more streamlined old ui. As it stands, with these and the other issues i will not refresh my update period for now. The older version is simple better.
Btw: no file load via network: i had that issue with older versions too. The reason after a long search was it running as Administrator. As soon as i removed that (compatibility mode), i could drag files from the network in again.
1 Like
The w,h options are calculated factoring in the aspect ratio as well, so it is a black box whether none, 2x upscaling, or 4x upscaling was applied.
I would like the GUI to have the ability to display the scaling factor applied to AI scaling.
1 Like
Sorry, I thought it was a scaling factor issue since there was also a lanczos comparison.
I mainly use Proteus and I could hardly notice any difference between 2.6.4 and 3.0 with the same values set.
However, in gradation, I felt a difference between 2.6.4 (8bit) and 3.0 (16bit), so I have decided that 3.0 is more favored.
I agree, it would make SO much more sense to have a CQ quality setting instead - I have the opposite problem: I would like āfatterā higher quality output (for later processing) and be able to choose -cq1 (even add on -qmin 1 -qmax 1 and get control over the -pix_fmt settings as well). It would also be nice to start with a ā-preset p7ā instead of ā-preset normalā ā¦nobody is to live stream the output itself anyway and the āmediumā and ā-b:vā settings simply do not cut it quality-wide for me. I am forced to use ProRes 422 HQ OR utilize a nvenc commandline to get decent results - But it would be so much easier to patch encode had the GUI just presented better codec options.
Still unable to process files. New ffmpeg error:
Finished processing frame prob-3 0 67.600867 67.600867
1920x1080, 4:4:4 10 bits, MBAFF=0 ACT=1
Finished processing frame prob-3 0 67.634233 67.634233
1920x1080, 4:4:4 10 bits, MBAFF=0 ACT=1
Finished processing frame prob-3 0 67.667600 67.667600
1920x1080, 4:4:4 10 bits, MBAFF=0 ACT=1
Finished processing frame prob-3 0 67.700967 67.700967
1920x1080, 4:4:4 10 bits, MBAFF=0 ACT=1
Finished processing frame prob-3 0 67.734333 67.734333
1920x1080, 4:4:4 10 bits, MBAFF=0 ACT=1
Finished processing frame prob-3 0 67.767700 67.767700
1920x1080, 4:4:4 10 bits, MBAFF=0 ACT=1
Finished processing frame prob-3 0 67.801067 67.801067
1920x1080, 4:4:4 10 bits, MBAFF=0 ACT=1
Finished processing frame prob-3 0 67.834433 67.834433
1920x1080, 4:4:4 10 bits, MBAFF=0 ACT=1
Finished processing frame prob-3 0 67.867800 67.867800
1920x1080, 4:4:4 10 bits, MBAFF=0 ACT=1
Finished processing frame prob-3 0 67.901167 67.901167
1920x1080, 4:4:4 10 bits, MBAFF=0 ACT=1
2022-10-22-22-20-46-Main.tzlog (946.1 KB)
Strong ghosting artefacts on moving objects when using stabilisation with āreduce jittery motionsā 3rd, 4th or 5th pass, together with enhancements models used. Video example had strong noise and strong jitter. Created on gopro 4 silver in low light situation. Any help regarding the use of āreduce jittery motionsā?
Which Threadripper and which nvidia gpu?
Motherboard would also be nice.
See also release notes.
1 Like
There are also options that allow you to specify the width and the height, thus changing the aspect ratio. Iāve used these with cropped images and then embed the cropped video into a letterbox 16:9 that will fit on a modern screen. (Unless theyāve removed this functionality for the 3.0 release version, Iāve found it to be useful when it is needed.
I am unable to preview or export a single clip with any enhancements at all. I can crop, trim, export. But the moment I try to use any of the AI models, and I do mean ANY, the preview fails, and export does nothing.
Running on 14inch M1 Pro MacBook.
Anyone else having this issue? Or know how to resolve? Ive been using v2.6.4 instead for the mean time.
to your Point 2) I would like to add that programs like Handbreak are offering both settings. So you either can dial in a quality Level OR use Bitrate. Bitrate is usefull in two scenarios. Either you need to know exactly how big your file is going to be (since you can calculate that with Bitrate and clip lenght) or if you need to hit a specific Bitrate (because of technical, hardware or customer demand reason).
So I would hope that they donĀ“t get rid of Bitrate in favor or Quality Level but implement both. Either in the way that Handbreak did it, or by a global setting inside of Preferences
1 Like
