I’ve been having fun experimenting with this new app, but like others, I think I’ll go back for a little while until some of these issues get ironed out. The deal breaker for me is that it seems to have lost the ability to accept AVISynth scripts as input, which means that I now have to create huge intermediary files. Was that deliberate, or is it a bug?
I keep running into errors processing a Blu-ray compatible h.264 file:
This one is related to an ffmpeg command
-vsync is deprecated. Use -fps_mode
Passing a number to -vsync is deprecated, use a string argument as described in the manual.
Could not find codec parameters for stream 1 (Attachment: none): unknown codec
Could not find codec parameters for stream 2 (Attachment: none): unknown codec
Could not find codec parameters for stream 3 (Attachment: none): unknown codec
Processing just stops and the error message can be revealed by clicking on the red “x” next to the clip.
This is on an RX6900XTX with latest AMD GPU drivers, Windows 10. Proteus relative to auto, 100% (denoise/deblock). ProRes output.
Some times, the computer just hard resets instead.
2022-10-20-12-41-51-Main.tzlog (188.0 KB)
2022-10-20-12-37-35-Main.tzlog (89.1 KB)
I have disabled “allow recovery” and am now getting these messages:
2022-10-20-13-15-15-Main.tzlog (2.6 MB)
Last message repeated 3 times
unknown SEI type 7
Finished processing frame prob-3 1 21.856000 21.856000
nal_unit_type: 6(SEI), nal_ref_idc: 0
nal_unit_type: 1(Coded slice of a non-IDR picture), nal_ref_idc: 0
Finished processing frame prob-3 1 21.898000 21.898000
nal_unit_type: 6(SEI), nal_ref_idc: 0
Last message repeated 1 times
nal_unit_type: 1(Coded slice of a non-IDR picture), nal_ref_idc: 2
Last message repeated 3 times
unknown SEI type 7
Finished processing frame prob-3 1 21.939000 21.939000
nal_unit_type: 6(SEI), nal_ref_idc: 0
nal_unit_type: 1(Coded slice of a non-IDR picture), nal_ref_idc: 0
Finished processing frame prob-3 1 21.981000 21.981000
nal_unit_type: 7(SPS), nal_ref_idc: 3
nal_unit_type: 8(PPS), nal_ref_idc: 3
nal_unit_type: 6(SEI), nal_ref_idc: 0
Last message repeated 3 times
nal_unit_type: 1(Coded slice of a non-IDR picture), nal_ref_idc: 2
Finished processing frame prob-3 1 27.277000 27.277000
unknown SEI type 7
nal_unit_type: 6(SEI), nal_ref_idc: 0
nal_unit_type: 1(Coded slice of a non-IDR picture), nal_ref_idc: 0
Finished processing frame prob-3 1 27.319000 27.319000
nal_unit_type: 6(SEI), nal_ref_idc: 0
nal_unit_type: 1(Coded slice of a non-IDR picture), nal_ref_idc: 0
Finished processing frame prob-3 1 27.361000 27.361000
nal_unit_type: 6(SEI), nal_ref_idc: 0
Last message repeated 1 times
nal_unit_type: 1(Coded slice of a non-IDR picture), nal_ref_idc: 2
Last message repeated 3 times
unknown SEI type 7
Finished processing frame prob-3 1 27.402000 27.402000
nal_unit_type: 6(SEI), nal_ref_idc: 0
nal_unit_type: 1(Coded slice of a non-IDR picture), nal_ref_idc: 1
Finished processing frame prob-3 1 27.444000 27.444000
unknown SEI type 7
nal_unit_type: 6(SEI), nal_ref_idc: 0
nal_unit_type: 1(Coded slice of a non-IDR picture), nal_ref_idc: 0
nal_unit_type: 1(Coded slice of a non-IDR picture), nal_ref_idc: 2
unknown SEI type 7
Finished processing frame prob-3 1 115.658000 115.658000
nal_unit_type: 6(SEI), nal_ref_idc: 0
nal_unit_type: 1(Coded slice of a non-IDR picture), nal_ref_idc: 0
Finished processing frame prob-3 1 115.699000 115.699000
nal_unit_type: 7(SPS), nal_ref_idc: 3
nal_unit_type: 8(PPS), nal_ref_idc: 3
nal_unit_type: 6(SEI), nal_ref_idc: 0
Last message repeated 3 times
nal_unit_type: 1(Coded slice of a non-IDR picture), nal_ref_idc: 2
Finished processing frame prob-3 1 115.741000 115.741000
nal_unit_type: 6(SEI), nal_ref_idc: 0
nal_unit_type: 1(Coded slice of a non-IDR picture), nal_ref_idc: 2
Finished processing frame prob-3 1 115.783000 115.783000
nal_unit_type: 6(SEI), nal_ref_idc: 0
nal_unit_type: 1(Coded slice of a non-IDR picture), nal_ref_idc: 1
Finished processing frame prob-3 1 115.825000 115.825000
nal_unit_type: 6(SEI), nal_ref_idc: 0
nal_unit_type: 1(Coded slice of a non-IDR picture), nal_ref_idc: 0
Just installed the new Video AI. It is a big upgrade. Congrats and it looks promising.
I was disappointed, though, that apparently I am not able to output as mp4 from the h264 and h265 encoders. It mentions a command error. The previous version was able to do this without any issue (at least to H264).
This is the message that I get when I click on the command error:
- Some parameter(s) may not be supported by the codec. Check FFmpeg command.
- Current container (file extension) is not be supported by the codec. Try changing the file extension.
This is a huge step backwards, I sure hope I can uninstall this and get the old Video Enhance back. First and foremost, mp4 files produced by Video AI CANNOT be used in Premiere Pro, the most widely used video editor in the world!!! This is a show stopper and makes Video AI totally USELESS to me!!! In addition, anything to do with cropping is plain old wrong. It says it is producing 16:9, but the result is NOT 16:9!!! As someone else said, what’s amazing is that Topaz is releasing it with so many bugs.
I’ve been having lots of fun with this release.
Trying out new models and commands. This is my kind of tool!
I’ve only been using the GUI to get starting points on commands for new models, and previews.
With time, this can become a really great tool for everyone, and not just people like me who love creating batch files and scripts.
Thank you everyone at Topaz Labs! Keep on keeping on!
It’s finally here!
Some notes/suggestions.
- Would like to have punctuation on preset titles. I tend to do version numbers as I’m working on tuning upscales (v1.0 etc). Right now I’m doing v10, v11 etc.
- Preview while ‘exporting’ doesn’t seem to really work terribly well, I’m assuming this is being worked on.
- I wish it wouldn’t play the upscaled video over and over when it finishes exporting.
- It would be nice in the GUI to be able to set the CQ when encoding x264/x265. All we get is bitrate auto? If we can’t set the CQ, at least set it on your end at 17 or less. (Really should be around 15 since for most of us the VEAI/VAI output is intermediate.)
- This was a bug in VEAI as well, even if you ‘close all inputs’ the application still holds handles open for exported and imported videos preventing renaming, moving, etc.
- “Grain” doesn’t look quite as good as it did with 2.6.4. (And there we had separate controls for amount and size. This just has one slider.)
- I’m not sure if you can do this and I haven’t found it yet but it would be cool if the GUI could let us open/close the ffmpeg output window.
What container were you using? Just curious.
(I selected mkv for my tests tonight and it worked fine to H265. I haven’t tried H264 yet.)
Notes on Topaz Video AI’s special ffmpeg:
- ffmpeg -version exposes one of your developer’s Windows user name in the include list.
- I’m surprised your version is leaving out so much: no lib-x264 or lib-x265? Those would be useful to provide options if users don’t want to be required to use the nvenc variants.
This is probably because VEAI is not an application for encoding.
If you want to encode freely, why not output with Prores and then use a specialized application?
I think it’s a licencing issue and not that it wouldn’t be the best solution to have it integrated into one software to optimize the workflow
Thanks for VEAI 3.0.0 which is great, BUT after upgrading from Early Access 3.0.8 which was working well, now it’s impossible to do anything ![]()
- The selected input video won’t show above (in the video window, which remains black)
- Impossible to select presets (the drop-down menu is gray)
- The blue “Export” button does not have any effect. And “Saving as…” won’t help.
- The zoom % shows “nan%”
- Preview button in rightclick menu is greyed… No blue preview button…
→ Impossible to do anything.
I’m blocked
CORRECTION: After selecting “None” for the AI model loading at first in “Preferences”, now all works well ![]()
And I can even re-select the custom AI model in the drop menu, that seemed to “block” the software.
I hope this helps if someone gets across the same problem.
Btw., this soft is really wonderful and very useful.
Version 2.6. was already fantastic, but now the version 3 is really a new era.
Thanks!
I have reported lack of frameserver input direct to TVEAI a couple of times in Beta, but have not had a response as to whether thats another feature being worked on to be added, or has been removed as I too usually import Avisynth scripts when I can direct into Topaz to save time.
On current process, exporting first then importing into TVEAI adds between 30 minutes and 6x hours to the process depending on what pre-script I am using, so somethings its not a large issues, sometimes its quite extensive. The biggest bugbear for me is that when testing changes to scripts, you can can save, import and preview in 2.3 in a few seconds, and if it looks terrible revert quickly.
Exporting first then re-importing means adding in time and processing and temporary input files so it just makes testing new scripts more painful - though so far most testing of 3 Beta has resulted in everything being harder than it was in version 2.
I think the best comparison is it feels a little like Apple and Windows - Apple likes to hold the hand and try to do everything for the user, Windows tends to allow you to do anything you want (even if it breaks it). The new GUI feels a lot like Apple - simplified for users to try it make it more straightforward or easier to use, but at the (current) sacrifice that many advanced functions have vanished or don’t work properly yet.
Hopefully these features get added back in.
I am still curious on trying this new app on a workflow - I am just not sure I feel up to the work involved in adjusting Avisynth scripts for lots of testing when they don’t import directly. It feels like I might be in for a tedious time.
ok, thanks for this test!
I will test with beta version 3.0.0.11b which I still have, but for now my videos look better with version 3.
OK, first attempt at using this new version, already a fairly large problem - output processing fails immediately, can’t export.
This line is a little baffling to me:
2022-10-20 15-16-32 Thread: 49516 Critical Unable to open file “D:/Video/Voyager/Test/S04E08%06d.png”
That’s the name I tried to name the export (as for some reason it wants a name and won’t auto create one) - it can’t open its own output files? That’s not an input file its trying to open.
Is there a trick for the name of the output in this version?
2022-10-20 15-16-32 Thread: 57168 Debug passProcDone 0 3 3 1 0
2022-10-20 15-16-32 Thread: 57168 Debug 8
2022-10-20 15-16-32 Thread: 57168 Debug 9
2022-10-20 15-16-32 Thread: 57168 Debug Updating video info {"sar":1,"framerate":25,"startNumber":13300,"frames":3000,"progress":0,"status":0,"frame":0,"procStatus":{"status":0,"eta":0,"fps":0,"message":"","pass":1,"error":"Reading option '-metadata' ... matched as option 'metadata' (add metadata) with argument 'videoai=Enhanced using alq-13. Changed resolution to 1440x1080'.\nReading option '-start_number' ... matched as AVOption 'start_number' with argument '13300'.\nReading option 'D:/Video/Voyager/Test/S04E08%06d.png' ... matched as output url.\nFinished splitting the commandline.\nParsing a group of options: global .\nApplying option hide_banner (do not show program banner) with argument 1.\nApplying option nostdin (enable or disable interaction on standard input) with argument 0.\nApplying option y (overwrite output files) with argument 1.\nApplying option loglevel (set logging level) with argument debug.\nApplying option nostats (print progress report during encoding) with argument 0.\nApplying option vsync (set video sync method globally; deprecated, use -fps_mode) with argument 0.\n-vsync is deprecated. Use -fps_mode\nPassing a number to -vsync is deprecated, use a string argument as described in the manual.\nApplying option filter_complex (create a complex filtergraph) with argument veai_up=model=alq-13:scale=0:w=1440:h=1080:device=0:vram=1:instances=1,noise=c0s=6:allf=t:all_seed=0,scale=w=1440:h=1080:flags=lanczos:threads=0.\nSuccessfully parsed a group of options.\nParsing a group of options: input url D:/Video/Voyager/V3 test/tiff%d.tiff.\nSuccessfully parsed a group of options.\nOpening an input file: D:/Video/Voyager/V3 test/tiff%d.tiff.\nCould find no file with path 'D:/Video/Voyager/V3 test/tiff%d.tiff' and index in the range 0-4\nD:/Video/Voyager/V3 test/tiff%d.tiff: No such file or directory","progress":0,"frame":0,"priority":3,"requestPos":0,"processorIndex":-1}}
2022-10-20 15-16-32 Thread: 49516 Debug Open Called for file "D:/Video/Voyager/Test/S04E08%06d.png" 13300
2022-10-20 15-16-32 Thread: 49516 Critical Unable to open file "D:/Video/Voyager/Test/S04E08%06d.png"
2022-10-20 15-16-32 Thread: 57168 Info EXITED: 3 1 0
2022-10-20 15-16-32 Thread: 49516 Critical Unable to open file at "D:/Video/Voyager/Test/S04E08%06d.png"
2022-10-20 15-16-32 Thread: 57168 Info ~TProcess(): destroyed
2022-10-20 15-16-32 Thread: 57168 Warning input video opened FAILED: ""
2022-10-20 15-16-32 Thread: 57168 Warning qrc:/videoplayer/TPreviewPane.qml:213:21: Unable to assign [undefined] to double
2022-10-20 15-16-52 Thread: 57168 Info Opening log folder: C:\Users\Vaengence\AppData\Roaming\Topaz Labs LLC\Topaz Video AI\logs\
My thoughts exactly. I was extremely disappointed to see that this was deemed fit for release.
There was a very roundabout discussion on this with VEAI devs saying they can’t do it because of licensing, and multiple people saying that they, in fact, can. It appears that AVISynth input will not be coming back though.
I saw a discussion on licencing for other matters, but Avisynth acts as a frameserver and has a fully open public licence - there would not be a licence issue importing these scripts so this statement is confusing. Are you sure they were referring to licencing on this specific issue?
I even just did a search for Avisynth licence on this forum and it didn’t come up with anything in case I missed it.
EDIT: Found it, they were using different spelling of licence. But yes, reading that I am still confused - there should not be any licencing required for using the Avisynth script import.
I used an mp4 container, as I always used in the previous version.
I’m finding it really slow even with a 4090 card and having trouble importing any files from my NAS.
If i copy the files to my local drive, they all import fine.
Imports used to work fine from my NAS on the older Video Enhance AI.
I’m normally not active on the forums but I felt I needed to respond to this release with my opinion. Feel free to disagree ofc.
Just like with Photo AI this is a huge letdown. After several hours of testing my conclusion is the same:
- Very buggy, it will take Topaz a year to get back to the same level of stability.
- The new UI is not better, just different (subjective) and one of many UI changes over the years.
- Most importantly: the end result does not look better. Nor is the process faster on my hardware.
Some users may be happy with some of the features I don’t personally use, like interpolation (I use flowframes for that) or command line, but for me personally this all seems pretty pointless and the time and effort would be better spent at trying to further improve the output quality.