Topaz Video AI v3.1.8

Do you only say don’t use Kaspersky because of the Russian connection ?

Wonder if you get this question deleted again.

2 Likes

I have had some strange temp files stuck next to the complete ones. Maybe it will be sorted out in the next update. Running 3.1.7

This is the 3.1.8 release thread, and in the release notes:

Fixes app not cleaning temp files when process fails (doesn’t apply to crashes)

Suggest you install 3.1.8 and see if it fixes your issue.

1 Like

Thanks!

1 Like

Just tested the newest version of Video AI with my RTX 4090 and I have to say, WOW! Looks like the power the 4090 is finally being tapped. I’ve gone from 6-8 hours processing a 1080P video with Artemis v13 to 2.25 hours (for a 90 minute movie). Its hitting roughly 14 frames per second. Additionally, with the NVIDIA AV1 - the encode is both fantastic quality and file size, coming extremely close to an x265 Slow for the same file size. Video AI + 4090 + NVENC AV1 = good.

2 Likes

The alpha version (not generally released yet) is showing even bigger performance gains, so stay tuned!

3 Likes

Slow is an understatement, I can’t even get it to load a 6Mb file.

topaz

Windows 11, 6Mb MP4, Nvidia 3080. Each successive rev. over the last several weeks have seemed to have slowed processing, now down to zero. I’m unimpressed.

3 Likes

This is the error I receive:

(codeformer) \CodeFormer>python inference_codeformer.py -w .7 -i ./inputs/whole_imgs/00.jpg
Face detection model: retinaface_resnet50
Background upsampling: False, Face upsampling: False
[1/1] Processing: 00.jpg
detect 4 faces
Failed inference for CodeFormer: nvrtc: error: invalid value for --gpu-architecture (-arch)

nvrtc compilation failed:

#define NAN __int_as_float(0x7fffffff)
#define POS_INFINITY __int_as_float(0x7f800000)
#define NEG_INFINITY __int_as_float(0xff800000)

template
device T maximum(T a, T b) {
return isnan(a) ? a : (a > b ? a : b);
}

template
device T minimum(T a, T b) {
return isnan(a) ? a : (a < b ? a : b);
}

extern “C” global
void fused_sigmoid_mul(float* tx_1, float* aten_mul) {
{
float tx_1_1 = __ldg(tx_1 + (long long)(threadIdx.x) + 512ll * (long long)(blockIdx.x));
aten_mul[(long long)(threadIdx.x) + 512ll * (long long)(blockIdx.x)] = tx_1_1 * (1.f / (1.f + (expf(0.f - tx_1_1))));
}
}

    Failed inference for CodeFormer: nvrtc: error: invalid value for --gpu-architecture (-arch)

nvrtc compilation failed:

#define NAN __int_as_float(0x7fffffff)
#define POS_INFINITY __int_as_float(0x7f800000)
#define NEG_INFINITY __int_as_float(0xff800000)

template
device T maximum(T a, T b) {
return isnan(a) ? a : (a > b ? a : b);
}

template
device T minimum(T a, T b) {
return isnan(a) ? a : (a < b ? a : b);
}

extern “C” global
void fused_sigmoid_mul(float* tx_1, float* aten_mul) {
{
float tx_1_1 = __ldg(tx_1 + (long long)(threadIdx.x) + 512ll * (long long)(blockIdx.x));
aten_mul[(long long)(threadIdx.x) + 512ll * (long long)(blockIdx.x)] = tx_1_1 * (1.f / (1.f + (expf(0.f - tx_1_1))));
}
}

    Failed inference for CodeFormer: nvrtc: error: invalid value for --gpu-architecture (-arch)

nvrtc compilation failed:

#define NAN __int_as_float(0x7fffffff)
#define POS_INFINITY __int_as_float(0x7f800000)
#define NEG_INFINITY __int_as_float(0xff800000)

template
device T maximum(T a, T b) {
return isnan(a) ? a : (a > b ? a : b);
}

template
device T minimum(T a, T b) {
return isnan(a) ? a : (a < b ? a : b);
}

extern “C” global
void fused_sigmoid_mul(float* tx_1, float* aten_mul) {
{
float tx_1_1 = __ldg(tx_1 + (long long)(threadIdx.x) + 512ll * (long long)(blockIdx.x));
aten_mul[(long long)(threadIdx.x) + 512ll * (long long)(blockIdx.x)] = tx_1_1 * (1.f / (1.f + (expf(0.f - tx_1_1))));
}
}

    Failed inference for CodeFormer: nvrtc: error: invalid value for --gpu-architecture (-arch)

nvrtc compilation failed:

#define NAN __int_as_float(0x7fffffff)
#define POS_INFINITY __int_as_float(0x7f800000)
#define NEG_INFINITY __int_as_float(0xff800000)

template
device T maximum(T a, T b) {
return isnan(a) ? a : (a > b ? a : b);
}

template
device T minimum(T a, T b) {
return isnan(a) ? a : (a < b ? a : b);
}

extern “C” global
void fused_sigmoid_mul(float* tx_1, float* aten_mul) {
{
float tx_1_1 = __ldg(tx_1 + (long long)(threadIdx.x) + 512ll * (long long)(blockIdx.x));
aten_mul[(long long)(threadIdx.x) + 512ll * (long long)(blockIdx.x)] = tx_1_1 * (1.f / (1.f + (expf(0.f - tx_1_1))));
}
}


I tried reinstalling and I get this error when installing the dependencies:
(codeformer) C:\CodeFormer>python basicsr/setup.py develop
No CUDA runtime is found, using CUDA_HOME=‘C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.5’

It does work now but it’s running on CPU, not on my video card. I’m not sure how to get this to run on my card without errors.

The error is: (‘Running on CPU now! Make sure your PyTorch version matches your CUDA.’

You should try to use the latest environment, uninstall the current cuda and install cuda 11.8, uninstall pytorch and install the beta version.

conda install pytorch torchvision torchaudio pytorch-cuda=11.8 -c pytorch-nightly -c nvidia

1 Like

Currently updating to v 3.18 from 3.17. I very much doubt that the issues I talk about have been changed. If they have I will be very grateful.

Lots of us have complained about the terrible TRIM interface. Huge waste of time.
In addition, once you view the TRIM area, it is difficult to further refine the selected area. Moving the markers with the mouse does not work well, the resolution is very poor. If you use the arrow keys then it is way too slow. There needs to be something like Ctr Arrow to move by 5 or 10 frames. This is not a new suggestion. The RESOLUTION for finding the correct starting and ending FRAME needs to be improved,

JUST GO BACK to the prior interface [ ] which was much, much better.
It’s really terrible that TOPAZ no longer pays much attention to any feedback about the interface.
The update from 2.xxx is overly complicated and possibly only of interest to a very small percentage of users. Stop frustrating the hell out of the rest of us.

Thanks for listening.

2 Likes

Complaint:

Why do we have to endure the waste of time when Video AI, automatically comes up in Trial Mode, and we have to wait a bunch of time, for the login authentication to finish?

Who the hell wants to waste so much time, when you have paid for the program and you have been authenticated so many times.

The other Topaz programs don’t do this, they come up normally. We want the same thing for Video Ai.

Thanks

1 Like

Issues:

Even when I turn the Autoplay Off, play engages automatically when I attempt to delete one of the Previews. This should not happen.
In fact, there should be a Checkbox next to the Previews, so one could mark one or more and then delete all the marked ones. This would be quicker than messing with one at a time.

How about improving the efficiency by having a better interface. Focusing on other issues should not mean that interface issue fall on completely deaf ears.

I would like to be able to edit more easily, more efficiently, and I’m sure no one is opposed to that idea. I am sure everyone at Topaz wants the same thing.

Now for the Praises:

The quality of the output has once more improved a lot. I also think that rendering is much faster than it used to be. So, Congrats on doing that job really well.

Thanks for listening.

In my case, I only have to log in once, after installing a new version. Every time after that, it opens automatically logged in.

Regarding trim, here is a secret: To mark the start point on the timeline, use the keyboard shortcut Shift [ and for the end point, Shift ].

1 Like

Or for the eternally shortcut-forgetful (like me) it’s

EDIT > Start trim position, then
EDIT > End trim position.

I hate having to remember shortcut keys, and even more, big ribbons (like MS ones) lol.

I’ve been trying v3.x since the beginning (successfully used v2.6.4), and was having no success with any release so far. I initially tried v3.1.8 on my system (Windows, ARC A770) with no luck - it kept failing giving “unknown error”. Saw a post about Intel’s OpenVino runtime version provided with v3.1.8 may be old, so replaced it. No luck.

So, I cleaned my system out of VEAI (both versions), uninstalled and deleted other directories/files left behind and reinstalled both from scratch. After installing the v3.1.8 release, I immediately replaced the OpenVino runtime files with the latest version and restarted the system. It works good now, and I’m able to process 2 files at the same time (I wasn’t able to get through a single file without failing before). I’m getting about 11.3fps on each process running simultaneously. Keeping my fingers crossed I’ve got it fixed now.

I was able to get it working after your help! Thank you so much! Legends do exist.

Newer version is available: Topaz Video AI v3.1.9

1 Like