Oh, ok, many thanks
UPDATE: I tried again after lowering the “Max. Processes” preference back down from 4 to 3, and this time it completed without a crash.
1 Like
Hello everyone, from today I will only post my opinion on TVEAI, I will no longer comment on the team and the old version. Let’s get down to business with the latest version 3.0.4
1/ Why did they change the Trim bar? Scrolling the trim to be able to take part of a movie is not pleasant, in my opinion, because it is not easy. In the first versions 3, you could cut without scrolling the trim. Why not go back to the old Trim it would be easier for us.
2/ During video output encoding, the preview still has this display problem. Why doesn’t this actually work while the video is encoded? It still takes a while to appear. It is a little faster, of course, but that’s not it yet.
3/ I couldn’t find the AV1 video output, is it in this version? If so, I haven’t found it at all, or is it in the next version that will be released at the end of the year? Someone tell me?
4/ I don’t remember, in the beginnings of version 3, you could export a video in Prores 4444XQ, moreover the output was in 12bits (not 10bits if I remember correctly), do you know why this option is not is more ? Has anyone been able to put it back in the encoder for the outputs? If yes I would like to know what to put in the encoder.json file
5/ The audio encoding is not yet at the top but I have confidence in the final version at the end of the year, it will be reliable and top. For my part, I had encoded a video with the “copy” option in audio, and at the output of the export, the sound crackled, is this normal?
6 / The Artemis model, whether high quality or low quality, has a really horrible rendering, why? I hope by then, the team will work on the models for a better final quality. 
I don’t know what else to say, but this is my complete opinion on the latest version, I hope other people can answer some points above that I gave for an improvement of TVEAI which will be released at the end of In any case, I encourage the team behind this software to fine-tune this magnificent software for next Christmas
4 Likes
Yeah, I think I will reinstall the v2.6.4 for now, still having an eye on upcoming versions. 
I’ve read that the software always needs a connection to the internet to work? But I saw this in your advertising:
Does it run locally? Or does it need to communicate with the outside world?
I am quite confused.
1 Like
What side is supposed to look better? The right looks blurry, lacking detail.
the right one is better, joachim try i think to get ride of the interlaced effect (you can see it around all the edges of the face, we see a grid !!) which is burned on a progressive video (but i can be wrong)
the ads talk about the 2.6.4 version, which can run locally by downloading all the models via a model download manager included in the software.
this feature is not yet available on the 3.x version.
Ok - I see that yes the de-interlacing artifacts on the edges have been smoothed out, but in the process the image now lacks detail it once had. To me it appears now to be “out of focus.”
Look at the hair or eyebrows or the person’s facial blemishes. They are smoothed away.
I don’t get it. In the effort of removing artifacts, you blur out real detail.
4/
FFMPEG’s ProRes encoder supports formats up to 10bit (yuv422p10le yuv444p10le yuva444p10le), so 12bit output is meaningless.
This is the specification of the standard encoder in FFMPEG, so it can be solved if Topaz makes a ProRes encoder that supports more than 12 bits and incorporates it.
3 Likes
There was a change in development plans to make FFMPEG GUI in v3.0, but I feel that the advantages are not being utilized.
The advantages of FFMPEG, as I see it, are as follows.
*The input and output, color conversion, etc., can be left to FFMPEG, which reduces the developer’s cost.
*The total processing speed can be increased by parallel processing FFMPEG.
However, what is the actual situation?
*The development cost of GUI has increased, and the improvement of core parts such as AI models has not progressed much.
*Parallel processing is unstable, and the development cost of the GUI to compensate for this will further increase.
While it is useful for small segments of video, it is not useful when processing long videos.
It is a good point that batch processing can be done with the CLI, but there may be users who do not like the CLI.
I hope that improvements will be made so that users will not feel that the change in development plans was a waste of time and money.
The annoyance is when you know the exact model you require has been downloaded, for example when processing a series of videos all the same size etc., and with the same parameters. After starting processing on the first video, I would in 2.6.4 go offline, knowing that the model is already on the computer. That’s how I prefer to work and it’s how most companies set up their video software - all that I deal with, large and small and there are good reasons for doing that when software will be left running for long periods of time.
But now in 3.0.x including 3.0.4, the second and subsequent processed videos embed the Topaz watermark as used for unlicensed operation. You have to go online again to not download the model, because you already have it. It is pointless for users, wasteful of resources at both ends and is a retrograde step from 2.6.4.
There is absolutely no need for such software to phone home for licence verification more often than, say, once a month. For me, this will be a deal breaker when next asked to do a paid upgrade. It is one of the basics that, like the awkward preview, should be fixed BEFORE working on enhancements.
I have to click between “all outputs” and “preview” to make the bottom pane refresh and show me it is making the preview I requested.
Just curious what graphic card are you using ?
Do your graphic card have build in NVIDIA’s NVENC AV1 encoder?
AFAIK, only the latest 4090 and 4080 have NVIDIA’s NVENC AV1 encoder.
UPDATE 2: Uh Oh, it crashed again partway through another SD video using the same settings. Now I am trying Command-Line to see if it’s somehow more stable. It took me a while to figure out that I needed to replace “FFmpeg” in the copied Export Commands with the actual path on Mac to make it execute:
/Applications/Topaz\ Video\ AI.appffmpeg [export commands here]
This worked! However, the audio is buggy: The sound can be heard using Quick Look, but the sound won’t play in QuickTime, nor after processing the clip through Compressor.
Tell me about the Proteus auto-correction mode in 3.0…
Since we only want sharpness, even if all other parameters are set to -100, compression recovery and noise reduction will work and detail will be lost.
Not so in manual mode.
What kind of coefficient is the set value for auto correction mode?
I have an rtx 3090
I think you need RTX 40 series GPU to get AV1 encoding.
RTX3090’s CUDA Capability is 8.6.
2 Likes
Having to “phone home” poses a security risk to your internal network.
1 Like
What models did they add? What models have been updated? Where does one find this information? And are new or updated models only available for buggy v3?
Correct me if I’m wrong but stabilization does not rely on a trained model.
