I’m having consistent issues where the exports are hanging for hours.
My configuration: Intel I7-12700KF, 128GB DDR5 Memory, Four 2TB NVME drives
I tried switched from 2 to 4 processes, at a certain point the exports hang. Leave it sit for 4+ hours, zero additional progress
The clip I’m exporting is quite small - source is about 30 seconds. I’m trying 2x, 4x, 8x, 12x, 16x. It’s usually hanging when it gets to the 8x. I’m using the Apollo Model, Stabilization with 2 passes. Encode to ProRes 422LT
I am going to use this, as I need to get rid of my frame-inaccurate issues with my AVFS setup. Naturally, I need to preview the job in question first. So, I will continue to use AVFS to do the preview, then use the CLI for real processing.
I think y4mp belongs to avs2pipemod64, btw. Normally (towards x265) I do:
VSPipe -c y4m "f:\jobs\job.vpy" - | x265 ...
Would be interesting to see whether TVAI’s ffmpeg can be moved to output to a pipe too. Then I could do the entire job in one fell swoop, and forego on a ~1.5T ProRes file.
Thank you for this! I was unaware of process lasso and gave the free version a try. The More-suspend feature WORKS WONDERFULLY, I was able to suspend the app, which paused it, then reenable it and it picked up right where it was. +1 bro is a hero until Topaz gives us a pause.
Brightness changes on a previously upscaled .mov when I run it through Enhance using Proteus fine tune again after I ran it through Stabilization. Suggestions? Topaz V AI v3.1.9, 2020 Retina iMac, OS 12.6.3, 128Gb ram, 10 Core 10th generation i9 (comet lake), Radeon Pro 5700 XT, 16 Gb vram. Any thoughts?Thank you! You have to tab quickly between the two jpegs to see the difference.
Quality I don’t know as I’m not a pixel peeper and I don’t use Proteus, but Artemis High is exactly twice as fast for me in 3.1.9 as it is in 2.6.4 on 1080p 2x upscales. Win10, Ryzen 7 5800x, 32 GB RAM, Nvidia 3060 ti.
There are other some issues as discussed in the thread but at least on speed, it’s now measurably much faster. PS there’s nothing wrong with being pixel peepers but on TVAI I don’t really do it, I focus on the bigger picture at the moment.
What are you iMac specs year, gpu, cpu, ram, vram,? What type of drive are you exporting to? Read/Right speeds? You should also know that both SSDs and platters read/write speeds slow down with Ventura for some people. This has been documented.
All things being equal (equal input format to output format), you should always increase bitrate. That is because you always lose on the deal with each new compression step, like H264/H265 (which is why it’s called lossy).
With an output format which differs from the input format, and assuming you are using a better compression on output, going from, say, MPEG2 → H265, things are different, of course, and you can suffice with an equal, or lower bitrate even.
Since you’re doing up upscale to 4K, from say 1080p, you are increasing your surface 4-fold. So, it stands to reason you need a much higher bitrate. But not necessarily much more (see 2nd rule). And A.i may smoothen out your image too (from all the corrective work alone), which will lower the bitrate requirements too.
In principle, you don’t even want a fixed bitrate, but rather something like CRF (Constant Rate Factor), as you should ideally let the encoder itself decide what bitrate is needed for what scene.
Yeah… so extract the video from the original file then add the audio back after the fact. Here is how:
extract just video:
ffmpeg -i originalvideowithaudio -c copy -an newfilename
do your video processing on the new file in video ai, when completed put it together
Add audio back:
ffmpeg -i videoaiexportfile -i originalfilewithaudio -c copy -map 0:v -map 1:a -shortest finalfilename
Essentially that command will use video from the 1st file, then audio from the second combined into the finished file using a straight copy, so no audio processing or reencode or anything.