Just started working with Topaz.
Started as 720 X 480 AVI file 29.7 FPS
File from S VHS video. Using Topaz I did this in two steps. First, I deinterlaced the video. Then I ran it a second time using IRIS LQ upscaling to 1280 X 720. (HD). I tried full HD but too fake looking. Too many artifacts and even some distorted faces in background.
Any suggestions for using other models and settings that would improve the video.
Thanks
still from the Topaz upscale!
Video_sample1|478x850
Check out this video tutorial the product team put together, might help you out.
Try Proteus 4 Auto 0% or 3% Recover Details (square pixels mode). The attached screenshot looks awfull. If the source is very bad, 720p maybe too much and 125% upscale/cleaning up would look better (576p). If the video is not sharp enough after Proteus 4 you can make 100% pass with Artemis 10 High quality with Recover details 100%
can you upload a orig sample? Just zip the file then you can upload it.
Here is a link to the original AVI file of the video that I am working with. Perhaps this is a better representation than a still
touristcottage2 001.avi
[
## touristcottage2 001.avi
](touristcottage2 001.avi - Google Drive)
I had my post removed from the forum that says it was flagged for spamming. I didn’t spam at all. I simply mentioned that I was planning to use topaz to restore videos to create youtube shorts. I never promoted anything, mentioned a channel. What is going on here?
After looking at this AVI file I wonder if you think I can at least upscale to 720
If you can’t really upscale vhs video what is the point of Topaz for anything except making high quality video even higher. I’ve read that handbrake can deinterlace better. So what is it that Topaz will do to improve the video I’m working with?
Interlaced can get really nice with topaz. Upscale direct in one pass. 239% interlaced progressive robust + some grain. This was a pretty low detailed source though and they guy runs out of the picture.
There are sources that are poorly suited to improving them, this seems to me to be one of those videos. It contains not enough pixel informations, smeared look, very low contrast, unclearly defined objects, blured faces often lead to creepy heads when using a reconstruction model. But you can’t draw the conclusion about TVAI based on only videos like this.
Source video deinterlaced (with Hybrid), black bars croped away + aspect ratio corrected (4:3). Use this one as base, for you’re upscale tests 
Upscale direct in one pass. 239% interlaced progressive robust + some grain.
Do you mean deinterlace and upscale in one pass? and which model. And do you mean upscale 239% do you use a custom setting. Are you talking about what the new resolution would be set to? And how much grain would you suggest?
thanks
| ArtndFun Topaz Video AI Owner
December 6 |
- | - |
Interlaced can get really nice with topaz. Upscale direct in one pass. 239% interlaced progressive robust + some grain. This was a pretty low detailed source though and they guy runs out of the picture.
Transfer and send large files for free with Sprend. Sensitive files can be encrypted and sent with Sprend. Transfer your files easily, quickly and safely!
QTGMC Deinterlaceing what Hybrid does (Avisynth/Vapoursynth) is way better, best you can get.
So, if Topaz isn’t great for deinterlacing for that kind of video, and you can’t really upscale older VHS video without weird faces in background… what is the point of using Topaz at all?
I’m getting so much conflicting information. I know everyone has their opinions and it’s subjective to a point. But technically there must be one process that is better than the other.
from a post on a forum
"My experience that TVAI produce better, sharper and detailed output HD when converting Interlaced SD / Video8 footage (Sony TRV-45E), when using the built-in Interlaced → DV/TV option (DV gives slightly more details and sharpness in the expense of a bit more noise, but very subtle).
When I de-interlaced the original file 1st (losless) to QTGMC / bob / “Slower” preset, using hybrid and then used the Progressive option in TVAI (tried: Artemis, Proteus, Theia & Gaia in all sort of combinations/options - I spent hours testing all type of combinations), it yielded less quality video then using Direct TVAI → Interlaced → DV/TV, with the original interlaced file.
Even though my QTGMC output file I fed TVAI with was really good quality compared to the original interlaced video.
So my conclusion with my testing, the best is to go Direct with TVAI with your interlaced video and upscale/deinterlace in one shot, you would get better quality output, more sharp and more details then upscaling your QTGMC progressive file. with the Progressive upscaling, the output HD file was a bit muddy, while the Interlaced Upscale was much cleaner and sharp/more details.
For me personally it’s a clear winner, Direct TVAI De-Interlace + Upscale. the best results:
My experience that TVAI produce better, sharper and detailed output HD when converting Interlaced SD / Video8 footage (Sony TRV-45E), when using the built-in Interlaced → DV/TV option (DV gives slightly more details and sharpness in the expense of a bit more noise, but very subtle).
When I de-interlaced the original file 1st (losless) to QTGMC / bob / “Slower” preset, using hybrid and then used the Progressive option in TVAI (tried: Artemis, Proteus, Theia & Gaia in all sort of combinations/options - I spent hours testing all type of combinations), it yielded less quality video then using Direct TVAI → Interlaced → DV/TV, with the original interlaced file.
Even though my QTGMC output file I fed TVAI with was really good quality compared to the original interlaced video.
So my conclusion with my testing, the best is to go Direct with TVAI with your interlaced video and upscale/deinterlace in one shot, you would get better quality output, more sharp and more details then upscaling your QTGMC progressive file. with the Progressive upscaling, the output HD file was a bit muddy, while the Interlaced Upscale was much cleaner and sharp/more details.
For me personally it’s a clear winner, Direct TVAI De-Interlace + Upscale. the best results"
You find 100 opinions. It just as I can see QTGMC deinterlacing produces best results, you can googel for it. Take care to correct settings;
Hybrid
set “very slow”. Deinterlaceing must happen slow to get best results, this takes time.
Most I don’t set the “BOB” flag which is doubleing framerate. I wachted the result frames by frame and the addional generates ones didn’t look good to me. But you can check this out pending on source it can differ. Then import deinterlaced video into TVAI and do the upscale.
When for you TVAI is clear winner doing all in one, then do it, everyone should do that when own eyes says it’s better
Thanks for the input.
This is the tutorial for working with VHS from Topaz Laps itself. It seems really really confusing.
I have to check this tutorial when I have time. I did some test with different models of your video sample. As I tought, this sort of video can be hardly upscaled. Best of the bad results I get was with Rhea1, here it is: touristcottage2 001_960x720_rhea1.zip (41.0 MB)
Don’t forget to compair the two pass upscale with the deinterlace upscale 
Thanks for all the advice. I’ll try everything. I’ve already done a lot of testing.
As of now Topaz Laps support team is running tests on some of my footage. I’ll post the results so everyone can see what they recommend.
I don’t set qtgmc to very slow, it removes a lot of details in the movements, I set it on draft, it deinterlaces badly, staircase effects are visible, but proteus v4 is powerful enough to correct them, and by using draft, all the details will be present. but after all it must depend on the sources, in my case it is hi8 digital8 and 8 mm tapes which are digitized in firewire from a trv 120e camcorder
This is response from Topaz Labs. I have to say it looks as good as I’ve seen it. They say “don’t do it one pass” I didn’t know there was an auto 2nd enhancement feature. Below were their instructions. I’ll post a link to their rendered image later.
I ran some tests and found that the settings in the screenshots below had the best results. I used Dione TV to handle the deinterlacing and then added a second pass enhancement (to enable go to File > Preferences > Advanced > Features > Show second enhancement feature) with Artemis with an input condition of aliasing or moire.
It’s not perfect, but since the input video quality is rather low quality to begin with, the AI models struggle to have enough pixel data information. These settings kept more of a “true” look as some of the models like Iris, Rhea, and Proteus can introduce some odd looks in the faces when it is trying to interpret them.
Here is the before and after
These are the instructions from Topaz Labs support. They have tested my original footage and suggest this is the best way to resurrect it.
They do not say do it all in one pass as many suggest. So far I would say these are the best results.
Full HD and no distorted faces in background.
Here is the process they suggested.
"I ran some tests and found that the settings in the screenshots below had the best results. I used Dione TV to handle the deinterlacing and then added a second pass enhancement (to enable go to File > Preferences > Advanced > Features > Show second enhancement feature) with Artemis with an input condition of aliasing or moire.
Here are links to the before and after (new footage)
Here is the original AVI file
Here is the video after Topaz deinterlace and upscale
It’s not perfect, but since the input video quality is rather low quality to begin with, the AI models struggle to have enough pixel data information. These settings kept more of a “true” look as some of the models like Iris, Rhea, and Proteus can introduce some odd looks in the faces when it is trying to interpret them."
