Well, that’s not really gonna work for me since I use Rhea quite a bit lately and that model wasn’t made available until v5.2.1.
I have installed Topaz Video AI 6 Beta.
I can’t judge what is still wrong, but what is new and improved in Topaz Video AI 6 ?
It is missing the shutdown of the operating system after the export, and that you can pause the export so that you notice when you shut down the operating system and the next time you open Topaz Video AI again that it continues where you left off.
Has Topaz Video AI 6 become even faster? And have all AI models hopefully been improved in picture quality?
Why isn’t there Live Render in Topaz AI 6 Beta, it still needs to be fixed and LiveRender finalised?
I have Intel I7 12700k with MSI RTX 3060 Ti VENTUS 2X OCV1 LHR not always fast enough. Which graphics card do you need to make it really fast?
With RTX4070 RTX4070TI, RTX4080?
With MSI RTX 3060 Ti VENTUS 2X OCV1 LHR it can take 10-20 hours if the source video is better and depending on the AI model and settings.
How fast would they be with RTX4070 RTX4070TI, RTX4080? Would you need 3-5 hours with them, which takes an extremely long time with RTX3060TI with export?
What I notice in Taskmanger is that CPU and GPU are utilised when exporting TopazVideoAI. What is calculated with CPU and what with GPU during export?
And please enable Natvie’s audio export in next updates if you have multi-channel audio, Topaz should also be able to keep multi-channel audio and not only export stereo audio.
This is my TopazVideoAI 6 beta benchmark.
Topaz VideoAI 6 Benchmark
Topaz Video AI Beta v6.0.0.1.b
System Information
OS: Windows v11.24
CPU: 12th Gen Intel(R) Core(TM) i7-12700K 31.686 GB
GPU: NVIDIA GeForce RTX 3060 Ti 7.8359 GB
GPU: Intel(R) UHD Graphics 770 0.125 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 12.31 fps 2X: 09.17 fps 4X: 03.10 fps
Iris 1X: 12.26 fps 2X: 07.49 fps 4X: 02.32 fps
Proteus 1X: 12.00 fps 2X: 08.94 fps 4X: 03.19 fps
Gaia 1X: 04.07 fps 2X: 02.80 fps 4X: 01.90 fps
Nyx 1X: 04.93 fps 2X: 04.24 fps
Nyx Fast 1X: 09.73 fps
Rhea 4X: 01.63 fps
RXL 4X: 01.56 fps
Hyperion HDR 1X: 29.93 fps
4X Slowmo Apollo: 18.55 fps APFast: 48.08 fps Chronos: 09.58 fps CHFast: 15.73 fps
16X Slowmo Aion: 21.48 fps
To everyone that has asked for this: Running two 60 watt incandescent light bulbs will use more power, than an idle computer that is capable of running TVAI for the same amount of time.
Just want to put that into perspective, in hopes of inspiring some reevaluation about what really uses power around your house.
There is. Add an extra tab at the top, chose different settings for each tab. Press play.
Also, post these questions in the dedicated 6.0.0.1.b thread perhaps?
Topaz Video AI Beta v6.0.0.1.b - Topaz Video AI / Video AI Beta - Topaz Community
Regarding your questions about GPU’s. My 3080 10GB renders almost twice as fast as your 3060 + cpu/memory combo. A RTX 4090 seems to render almost 3x as fast as your 3060 Ti judging from the benchmark thread.
Have a look here:
Video AI 5.5.X - User Benchmarking Results - Topaz Video AI / User Benchmarks - Topaz Community
To use other encoders, you only have to pipe the TVAI ffmpeg output to another ffmpeg built with the encoder you need. (Using CLI obviously)
How?! I’ve never seen ‘-raw’ option of any kind. But maybe I can just remove the “-c:v” “ffv1” part and such, and have it output in a classical way.
TVAI uses a simple Lanczos resize, which is not only a lot better than those nvidia specific encodes, but very light-weight, CPU-wise. Lanczos is broadly considered the best function for upscaling.
LOL, didn’t even know that.
I was looking for their source on github,
I’ve got the same problem …
Unfortunately no release with Rhea/ XL / Hyp is as stable as 5.0.4
this app is a drama
You can use a command line that
ffmpegTVAI “TVAI options” -c:v rawvideo -c:a copy -f nut - | ffmpeg(your own built) -i - “encoder option”
Thank you kindly!
So you’re saying that I can use Rhea in 5.0.4 even though Rhea was not officially part of the “package” until 5.2.1?
I haven’t investigate such way, but “should be” possible if using using correct compiled ffmpeg and command.
What I mainly want to say is that nothing has been truly stable since 5.0.4, and given the number of versions since then, it’s a bit dramatic.
To be honest, I don’t even test updates anymore… that’s to say
As soon as I have something less unstable than usual, I’m happy with it.
I gave my 5.3.4 Pro license some Chronos jobs. This is the best I can get from it with at most 70% of total GPU use.
I don’t know if this is the place to say it but at least I know the thread is monitored. The obvious next step for TVAI - a very good product in terms of features - is a generative model as in giga that can essentially do vid to vid, style transfer or change, etc. Video that keeps generated objects, style and faces consistent. Maybe the face part is a big ask so that would be a recover but vid to vid would be a game-changing piece of software.
I like this product but I’m noticing a growing annoyance in the audio delay when converting from format A to format B. The audio appears to be ever so slightly off with the lips.
I was wondering if you could provide some guidance on ensuring the audio syncs better or some way we can (globally)) adjust the audio or create sections (local adjustments) where we can locally adjust the audio by some time offset. I have a powerful computer (i9 @ 3+Ghz, 64GB RAM, 3090 Nvidia, and 4TB SSD’s all over, so it’s not my computer.
Thank you.
yes you’re right, power consumption in idle state is low, but a shutdown implementation is super easy and update robust, they can use builtin windows shutdown command and that’s it.
My RTX 4090 is pulling max 500W (I limited it that way) + about 200W CPU for a Rhea1 process. So, at 700W, for about 8 hours, I couldn’t care less about my computer running idle for the rest of the night.
Also, I like to get up in the morning, and see what possibly went wrong: lost frames on the command line, crashed driver maybe? A shutdown wipes all of that info.
Oh, the solution is as annoyingly simple as the cause: Topaz needs to finally fix the VFR bug.
That’s likely a PSU problem. Your video card draws too much power and therefore driver crashes.
Reconnect your cables first, clean your power connectors then check again.