Topaz upgrade model is not related to versions, it’s up to the buyer to decide when will he/she is willing to upgrade. It’s not mandatory to upgrade when v3 will come out. Some will get this version for free anyway. If you’re not paying until, for example, Proteus 4 release it’s totally fine. Stabilization, Apollo models, and a new GUI can attract new consumers while you wait.
1 Like
the new version 3.x applies automatic values for each frame where the old version 2.x only made an automatic setting for the whole video!
Forgive me, but that seems hard to believe without processing taking way longer than v2, which it doesn’t. Is that written anywhere in the changelogs?
I have the same issue, so with SloMo 100% and using original FPS (because changing it wont work in this version) you end up with quality worse than the original because as you frame step you see every other frame is a duplicate frame.
2 Likes
I mostly agree, but in my case in would save me quite a bit of time if I had some basic colour correction in VEAI, it would save me the time of having to load tiny little clips into resolve, and colour correct each one. Ideally it would be great if VEAI could use AI to colour correct the footage. Imagine if you could say original source is daylight, but turn it into dusk lighting.
But manual gamma adjust and white balance would be enough for me.
Thanks. That is pretty impressive, albeit it depends how well it works - you could argue that you wouldn’t want all of the settings being changed frame by frame, though it could be programmed to only change on shot changes perhaps.
For me the GUI doesn’t look primetime, I’m not sure if that’s still being worked on?
1 Like
Thank you for sharing this.
Can you please upload your logs here?
Thank you for noting this issue. We have resolved this in the upcoming build
1 Like
Can you please share the input file with me as well as a screenshot comparing the degradation if possible?
Please upload the videos to this DropBox Folder*
Just uploaded the log. Made a fresh one, was able to reproduce the issue on first try.
I’ve found that with every update there are some new glitches, unfortunately with this version I had a problem with the download of the model, like it’s happened in the past, the download speed is so slow that it’s almost impossible to download. About 700Kbps, which is less than 10KB/s speed. Well the model is proteus
The old version was completely uninstalled before installation. Probably a problem that had occurred in version 2.0, but there was no such problem in the last 3.0 (0.3b). Then it’s back in 0.4b.
When set bitrate 16 Mb/s FFmpeg Error!
Reading option ‘-b:v’ … matched as option ‘b’ (video bitrate (please use -b:v)) with argument ‘16M’.
Reading option ‘-map_metadata’ … matched as option ‘map_metadata’ (set metadata information of outfile from infile) with argument ‘0’.
Reading option ‘-movflags’ … matched as AVOption ‘movflags’ with argument ‘frag_keyframe+empty_moov+delay_moov+use_metadata_tags’.
Reading option ‘-map_metadata:s:v’ … matched as option ‘map_metadata’ (set metadata information of outfile from infile) with argument ‘0:s:v’.
Reading option ‘-map_metadata:s:a’ … matched as option ‘map_metadata’ (set metadata information of outfile from infile) with argument ‘0:s:a’.
Reading option ‘-c:a’ … matched as option ‘c’ (codec name) with argument ‘copy’.
Reading option ‘-metadata’ … matched as option ‘metadata’ (add metadata) with argument ‘videoai=Enhanced using thd-3 with reduce noise at 40, sharpen at 5, and revert compression at 0. framerate changed from 30 to -1 using chf3’.
Reading option ‘C:/Users/user/Downloads/원본/video_287047481_thd3_chf3.mp4’ … matched as output url.
Finished splitting the commandline.
Parsing a group of options: global .
Applying option hide_banner (do not show program banner) with argument 1.
Applying option nostdin (enable or disable interaction on standard input) with argument 0.
Applying option y (overwrite output files) with argument 1.
Applying option loglevel (set logging level) with argument debug.
Applying option nostats (print progress report during encoding) with argument 0.
Applying option filter_complex (create a complex filtergraph) with argument veai_up=model=thd-3:scale=0:w=3840:h=2160:noise=0.4:blur=0.05:compression=0:device=0:vram=1:instances=1,veai_fi=chf-3:fps=30:device=0:vram=1:instances=1,scale=w=3840:h=2160:flags=lanczos:threads=0.
Successfully parsed a group of options.
Parsing a group of options: input url C:/Users/user/Downloads/원본/video_287047481_180207.mp4.
Applying option t (record or transcode “duration” seconds of audio/video) with argument nan.
Invalid duration specification for t: nan
To import the image sequence are you selecting one image file or all of them. If you are selecting all of them, can you try selecting just one and letting me know if it is still crashing?
Thank you for all the testing and reporting so many issues.
The early access release 3.0.0-2 has fixes to many of the issues reported here.
Here is a list of the issues that were fixed:
- Slowmo + change FPS does not work
- Processing failure due to t: nan
- Deinterlacing is too slow when also upscaling
- Video timeline shows zero duration when switching videos
- Duration reported incorrectly
- Trim resets when switching video
- Filters reset when switching from preview to export
- ETA does not report days
- Looping of preview while processing
- Export preview issues
- Some GUI crashes ( Still there are quite a few scenarios left to be reproduced and fixed)
- Auto audio does not work
- Transcode doesn’t allow extra options in the encoders file
- and some more
https://community.topazlabs.com/t/topaz-video-ai-early-access-v3-0-0-2/34392
4 Likes
I was selecting just one, the first one. And then importing process starts. It last few minutes, probably because there are +6K images. UI seems frozen like nothing is happening, but in the sysinfo I can see that HDD are doing massive data reading. When import is complete, imported sequence appears in the good place, but there is no preview. Few seconds after that, app crashes. If you wish, I can make screen recording of this and send it to DropBox.
Below is the previous reply in 2.6.4. Doesn’t seem to have seen it. So post again.
Hi, I have a question and idea. 2.6.4 includes a question about setting the resolution in the whole 2.0 version. After “Lock Output size to scale” is not selected and the specified resolution is set, the output result will be trimmed, that is, a white frame can be seen in the preview window of the modified resolution. The actual content of the output video is cropped by this framework. That’s not good, because it’s a cropping function, just like 3.0 is going to do, or 3.0 can take this modification and the resolution will be trimmed.
Of course, this is not important. What is important is the following suggestion, namely: add aspect ratio options to modify the resolution. There are three good options. ①Ignore the aspect ratio (ignore the aspect ratio of the original video, output the video according to the ratio of the given resolution, and do not crop)
②Maintain aspect ratio (Output the video with the aspect ratio of the original video and no cropping)
③Crop the screen (cut the screen according to the white frame given the specified resolution and output the video)
For example, 720x480 enlarged by 4 times is 2880x1920. If the output resolution is specified as 2560x1920, then ignoring the aspect ratio results in 2560x1920 and the image ratio is 2560:1920. Keeping the aspect ratio is 2560x1920 and the frame ratio is 720:480. The cropped image is 2560x1920, the scale is 2560x1920, but the image is cropped out. (Software can also be cropped and zoomed in to save resources.)
This option can appear after “Lock Output size to scale” is unchecked and appears as a check box for selection
Well, the above is what was sent before in 2.6.4. Now some additions are made below. Also sad to hear that your development was outsourced to someone else.
This is about resolution changes and cropping issues that are still unimproved in version 3.0.0.4b. (Now there is reason to doubt the skill level of the outsourcers) Some basic human common sense rather than anti-human settings and operations should be developed, this is the most basic. Stop doing things against humanity. ①Cutting is cutting, what function should it have, a human should know, no need to say much, because many people have put forward suggestions. But still no improvement ② 3.0.0.4b put the output customization settings at the top, which is good. But why stupidly change resolution (select resolution - select width and height) into cropping? ? ? ? ? Why is it so inefficient? ? ? You already have a separate option for cropping, the crop function and the option to add black bars should be grouped into the crop screen instead of repeating it in custom output settings. (This is a problem left over from 2.0, because 2.0’s cropping is based on this). But what does it mean to choose an output resolution under normal circumstances? It means that I just want my video to output the video according to the resolution and ratio of my output, and automatically stretch or zoom out（shrink） the picture according to the resolution. Don’t you want it to be cropped? ! Do I need to crop? No need to be okay! If I need to crop, I just need to set it on the crop button, and then the output resolution will naturally be enlarged according to the cropped picture. If you need to add black borders, you will naturally set whether to open it in the crop button options. It will never say that you have to be unique, unconventional, and different from the norm. Why is veai so awkward? Why do you have to be weird and maverick? I have a problem with a 3880x1920 video, after processing I want it to output at 2560x1920 without losing any frame, simply other software will stretch/scale it instead of cropping it. But veai was very “smart” and chopped it up for me, directly discarding part of the picture and giving me a disabled scene with only half the screen.
I get ffmpeg errors on every video I try with this build.
A big thank you for the ability to increase the preview time to 15 seconds. Really appreciate it.
Regarding the crop I prefer the older version where one could mouse click to the select the area of cropping compared to the current movement of the orange lines at beginning and end.
1 Like
I think they’re putting that feature into 3.0, but it isn’t working yet, at least on the Windows version.
I think @luciovids talks about trimming, not cropping 
I agree with him, did the same request too.
2 Likes
Does the latest “early access” version 3.0.0-2 incorporate everything that’s in this beta version 3.0.0.4.b? Should we be testing 3.0.0-2 instead now?
