Time to address 'the elephant in the room' with Video AI - the janky playback of the instant preview

This has been the most frustrating element of the v3 series for me - the lagging, unpredictable, unresponsive and virtually unusable playback of the active export preview! The only time it’s responsive is when the export is completed, or when you initially press play.

If we can’t have a live frame by frame preview as in v2, then the playback experience of the active export preview needs to be fixed in the v3 series so it’s at least usable.

Current issues:

Play/Pause button (yellow circled on screenshot) is extremely unresponsive during an active export, sometimes it will play, sometimes it won’t, then you press it again and maybe 10 seconds later it will pause, or play again, it’s impossible to predict what will happen. The playback and pause button needs to actually do something when you click it, not 10 seconds later!

I understand the exported video is written in chunks and only the written data can be accessed, but the playback marker (red circled on screenshot) does not even play to the end of the progress line, it should do this, at a minimum.

And as soon as the next chunk of video is written as indicated by the progress bar, you should be able to play that part back immediately.

I love this software but this needs to be addressed!

Thanks!

4 Likes

Yeah the preview is broken since the first 3.X version, they didn’t fixed it yet.
BTW why in yours the encoder shows like “NVENC HEVC QP 22”, I don’t have that encoding preset:
image

1 Like

Great feedback!

We are continuing to optimize the preview experience in the v3.x series. You are correct in stating that the way that files are processed and therefore previewed when processing is quite different than with the v2.x series. We are working on a smooth instant preview experience and this part of the app will continue to improve :slight_smile:

2 Likes

Hi, that’s a custom encoder entry that I set up (also the only one I ever use as the quality is visually lossless to me and the size is reasonable). You can use it as well if you add it to your encoders.json file in the models directory for Video AI:

  {
    "text": "NVENC HEVC QP 22",
    "encoder": "-c:v hevc_nvenc -rc constqp -qp 22 -pix_fmt yuv420p",
    "ext": [
      "mov",
      "mkv",
      "mp4"
    ],
    "transcode": "aac",
    "os": "windows",
    "device": "nvidia"
  },
1 Like

@david.123 can you not use CRF in NVenc? qp is generally perceived as being inferior to crf.

I don’t have that .JSON file…

I tried creating it and isn’t shown on VAI.

it’s located at : C:\ProgramData\Topaz Labs LLC\Topaz Video AI\models
(encoders.json)

if you prefer [lossless] in TVA instead, then HandBrake it to whatever you like.
You can set this Lossless Method in TVAI.
this is my method I use, Lossless encoding in TVAI -then-> Handbrake to x265 (libx265) crf 23.
In general libx265 (aka CPU / Software encoding) is much smaller in file size compared to NVEnc (GPU / Hardware encoding) x265 encoding at the same crf level/settings, some claim also it is better image quality, even though the size is smaller.

In general; CPU/libx265 generates smaller file size at the same image quality or even better image quality, then GPU/NVEnc encoding if same settings is used (crf).

I chose this as my output format based on the information in the following post:

https://superuser.com/questions/1236275/how-can-i-use-crf-encoding-with-nvenc-in-ffmpeg/1236387

Specifically from the poster who wrote about encoding blurays with no perceivable quality loss using qp.

The resulting output is high quality, small in size and the encoding performance is super fast.

I don’t feel like I need to try anything else.

based on this post, crf is the prefered way of going if you want to maintain same quality throughout the entire movie even when scene are changing, in general crf will have smaller file size.

CRF versus Constant QP

This one is also interesting. it saying to use “cq” instead of “qp” with NVenc if you not using crf (i think since 2022 crf became available for NVenc)

1 Like

The answer is no. It is not part of any hardware implementation.

1 Like

which part?

Y’all hijacked my post, lol

ohhh no we were talking while you were away :slight_smile:

1 Like

On most my video files the preview gets stuck shortly after beginning of the processing and for the rest of the processing until finish. I once converted the file to ProRes before processing and this might prevent the application to get stuck, but I need more tests to confirm this. Unfortunately converting a file to ProRes with a tool of your choice before might take an additional half an hour to an hour itself.

I added what I was replying to.

Sorry about that.
I think that was supposed to be in the name of a work-around?
Personally, I have not seen this topic’s main issue much. When I do, it doesn’t bother me because I don’t want to see the preview playing until it’s done processing anyway.
I’ll even take it one step more and render about 4 previews before comparing them.

For actual exporting, I only use the CLI. So, no preview at all for me. I’ve had to redo about 5 videos because I got logged out before it started and it added the watermark.

@nipun.nath @suraj @jacob.muchow

I’d like to make a suggestion for the devs regarding this issue:

To make the instant preview more responsive, because processing an export is so resource intensive, perhaps active exports can be automatically paused when trying to view and navigate the instant preview.

Currently the system resources are occupied processing the export job which is why I believe the instant preview is so unresponsive/unpredictable/unusable.

Of course this would mean that pausing functionality for an active export job is implemented, the most voted for feature…

I think this would make the experience even worse for me. :eyes:

If you have a better idea then let’s hear it.