I've uploaded a separate unedited video showing the results of using Video Enhance on regular SD 480i video sourced from a MiniDV tape: ua-cam.com/video/vIOqrMz0KW0/v-deo.html
I appreciate you doing this; very few people put in the time and effort to make review videos of this detail. And - excellent work moving an actual camera around and pointing it at a screen vs just screencapping with OBS or something! I'd have constantly over/undershot my hand motions!
I wa thrilled to see your Canon GL1. I bought mine in October of 2000 and still have it. I took lots of great family videos, along with professional video as well. A great camera and I have hundreds of Mini-DV tapes I would like to enhance, which is the reason I'm watching this video. I paused right after the GL1 appeared to write this to you, but now, back to your video! Thanks!
Those examples you show do look quite good. I had a go with Video Enhance AI a while back and wasn't too impressed. It certainly didn't seem any better than QTGMC, but took a lot longer and is much more expensive, so I think I'll stick with Avisynth based deinterlacing and scaling.
I always thought QTGMC was slow, but Video Enhance makes it look fast by comparison. If the price was lower, it didn't take so long to process video, and it didn't botch the audio, I'd actually give some consideration to using it full-time. For now, I'm going to stick with QTGMC and possibly experiment with other AVISynth plugins to sharpen and enhance the video.
How good would this be for upscaling and denoising hi8 footage where the majority of shots are fast motion, handheld footage. I assume it will struggle with fast motion scenes?
Hi bro I am planning to buy a laptop with an RTX 4070. Can it render 1080p to 4k fast using Topaz Video AI? What do you think about this one? Here are the specifications of the laptop ASUS TUF Gaming F15 FX507ZI Gaming Laptop, 15.6" FHD 144Hz Display, Intel Core i7-12700H, 16GB DDR4 RAM, 1TB SSD, 8GB NVIDIA RTX 4070, Backlit ENG K/B, Win 11 Home, Gray Thank you!
thanks for the video - the dropped audio is a deal breaker for me. Q - Is there software that can upscale a 2 hour 720p video to 4k in 6 hours (Nvidia RTX3060 or slightly better if nec. - and if really nec. pehaps I'd break the bank for RTX3090)? How long does QTGMC take? TIA
The rtx 3060 is an awesome card for AI/video editing because of the 12gb vram and it's price, it can run some AI programs at higher settings than the faster 3070 that only have 8gb of vram.
After seeing all the rave reviews praising Video Enhance for breathing new life into old SD video, I wanted to give it a go. Going forward, I'm going to experiment using a HDMI capture device similar to that which Technology Connections highlighted in his video titled "Best Easy Way To Capture Analog Video."
It's actually more useful upscaling from already decent resolution to higher resolutions, the Gaia upscale model does a good job intelligently sharpening things like hair, problem is if your source resolution is is too low then the upscale looks more like a "guess" and faces start to look "painted".
I've been using this for a while (a couple of months) as a denoiser to get the same quality out of cameras that I've seen other UA-camrs get out of the same models (we don't need to discuss how I got my hands on a "full" version of the software). I've noticed, when using it with my 480i and 1080i cameras, that the deinterlacing leaves a bit to be desired (I've defaulted back to using QTGMC since I can tune that with Avisynth+ to use my GPU and get damn near realtime deinterlacing in HD on my RTX 3080). Also, MPEG-2 SD video (from stuff like DVD and flash memory cameras) always seems to be TFF, but anything recorded to analog or digital tape seems to be BFF (haven't looked at micromv footage, but that's hardly ever relevant). As for the render times, the AI models are VERY intense as far as processing power. As mentioned, I'm using a RTX 3080 so I have plenty of very fast video memory as well as a fairly capable GPU and while I'm able to do 4K video, it takes something like 5 hours to do 2 minutes of 4K video with the Proteus model (for denoising and detail enhancing). Hell, the render times and the lackluster performance of the Chronos models are what ultimately pushed me to buy a Canon GX10 to replace my Canon HF G50. I've noticed that all models seems to have an issue with audio sounding like it's done with low bitrate (even when exporting with the ProRes option) if they output audio at all. This seems to be a universal thing so I end up just taking the outputs and making extensive use of the "map" command in ffmpeg to strip the audio out of the original files (video codec as copy from source 0, audio codec as whatever will work in a MP4 container from source 1--apparently 24-bit PCM audio is not supported by the MP4 container so I have to use something like AAC). As for upscaling, the results with SD to HD are more impressive than 4K to 8K out of my Canon GX10, but as for removing compression, denoising, and making the details a little bit more crispy, this does seem to be one of the best available options (beats the tar out of my "not entirely legitimate" copy of the NeatVideo filter in Sony/Magix Vegas before mopping the floor with it). I've had bad luck deinterlacing with this program and have defaulted to to using QTGMC since that's a pretty reliable standby (see above), but with already deinterlaced video, upscaling 1080p is a bit meh, upscaling SD to HD is damn impressive, working at the native resolution for denoising and removing compression artifacts is also damn impressive. As an upscaler (with a starting resolution at already HD), deinterlacer, or frame doubler, I cannot recommend this program, though they have been making visible progress in those areas. As an enhancer (at native resolution) and an upscaler for SD video, I can *mostly* recommend this program (if the end user is familiar enough with avisynth to deinterlace their videos beforehand and ffmpeg to deal with the audio problems). The biggest problem that keeps me from fully recommending this program is that you need a really beefy computer to actually use it to its fullest potential (I have an 11th gen i7 and RTX 3080, my RTX 3080 is exponentially faster with the renders than my CPU and it's still not quite fast enough to be comfortably usable even with the video memory allotment slider cranked all the way up to 16GB).
I've been wanting to try out the new version myself but haven't had any compelling reason to do so. It seems that the refreshed version has been oversimplified, though I'm going off of screenshots I've seen online.
I just tried it on an old vhs home movie and it did help a little.... it seems that if you have a fairly good video, it will make it better but I'm interested in making a not so good video a lot better... which I think it is showing it can do but in my opinion not really. I think the company who gets their stuff together will get rich in a hurry. @@stereophonicstuff
This wasn’t recorded at 720p. It was recorded in 480i MPEG2 using a standard def camcorder that I deinterlaced and upscaled using QTGMC, NOT Video Enhance AI because it’s what I felt like using. DUH!
I've uploaded a separate unedited video showing the results of using Video Enhance on regular SD 480i video sourced from a MiniDV tape: ua-cam.com/video/vIOqrMz0KW0/v-deo.html
I appreciate you doing this; very few people put in the time and effort to make review videos of this detail. And - excellent work moving an actual camera around and pointing it at a screen vs just screencapping with OBS or something! I'd have constantly over/undershot my hand motions!
I wa thrilled to see your Canon GL1. I bought mine in October of 2000 and still have it. I took lots of great family videos, along with professional video as well. A great camera and I have hundreds of Mini-DV tapes I would like to enhance, which is the reason I'm watching this video. I paused right after the GL1 appeared to write this to you, but now, back to your video! Thanks!
Those examples you show do look quite good. I had a go with Video Enhance AI a while back and wasn't too impressed. It certainly didn't seem any better than QTGMC, but took a lot longer and is much more expensive, so I think I'll stick with Avisynth based deinterlacing and scaling.
I always thought QTGMC was slow, but Video Enhance makes it look fast by comparison. If the price was lower, it didn't take so long to process video, and it didn't botch the audio, I'd actually give some consideration to using it full-time. For now, I'm going to stick with QTGMC and possibly experiment with other AVISynth plugins to sharpen and enhance the video.
Nice quality, but the lack of audio passthrough and only offering choices of MP4 or ProRes would make it unusable for me.
For all the people that have told and will continue to tell me I need a new graphics card to speed things up: which (affordable) one should I get?
How good would this be for upscaling and denoising hi8 footage where the majority of shots are fast motion, handheld footage. I assume it will struggle with fast motion scenes?
Hi bro
I am planning to buy a laptop with an RTX 4070. Can it render 1080p to 4k fast using Topaz Video AI? What do you think about this one?
Here are the specifications of the laptop
ASUS TUF Gaming F15 FX507ZI Gaming Laptop, 15.6" FHD 144Hz Display, Intel Core i7-12700H, 16GB DDR4 RAM, 1TB SSD, 8GB NVIDIA RTX 4070, Backlit ENG K/B, Win 11 Home, Gray
Thank you!
A gtx 3060 is an awesome card for this program, it can upscale about a second of video every 3 seconds, about 3x's faster than a gtx 1070.
When you doubled frame rate in Virtualdub, is that the same as interpolation of frames, giving a smooth look?
Yes
this software is extremely slow, wont use more than about 20% of my gpu or cpu. Im upscaling dvds 2x and upping to 60 fps 4k, any suggestions
thanks for the video - the dropped audio is a deal breaker for me. Q - Is there software that can upscale a 2 hour 720p video to 4k in 6 hours (Nvidia RTX3060 or slightly better if nec. - and if really nec. pehaps I'd break the bank for RTX3090)? How long does QTGMC take? TIA
The rtx 3060 is an awesome card for AI/video editing because of the 12gb vram and it's price, it can run some AI programs at higher settings than the faster 3070 that only have 8gb of vram.
Does topaz upload video on some server?
Why not just use a HDMI to dv converter for your GPU issue? Is it a desire to keep DV native?
After seeing all the rave reviews praising Video Enhance for breathing new life into old SD video, I wanted to give it a go. Going forward, I'm going to experiment using a HDMI capture device similar to that which Technology Connections highlighted in his video titled "Best Easy Way To Capture Analog Video."
I'd be curious to see if the "upscaling" is actually useful or not.
You can do one jump pretty good, but not so much two
@@aman-mn5kc From what to what?
It's actually more useful upscaling from already decent resolution to higher resolutions, the Gaia upscale model does a good job intelligently sharpening things like hair, problem is if your source resolution is is too low then the upscale looks more like a "guess" and faces start to look "painted".
@@budthecyborg4575 I've found that it does best with straight lines and text, at least judging by the results of the Early Edition upscale.
its slow because youre supported to use the gpu, not cpu
I've been using this for a while (a couple of months) as a denoiser to get the same quality out of cameras that I've seen other UA-camrs get out of the same models (we don't need to discuss how I got my hands on a "full" version of the software).
I've noticed, when using it with my 480i and 1080i cameras, that the deinterlacing leaves a bit to be desired (I've defaulted back to using QTGMC since I can tune that with Avisynth+ to use my GPU and get damn near realtime deinterlacing in HD on my RTX 3080). Also, MPEG-2 SD video (from stuff like DVD and flash memory cameras) always seems to be TFF, but anything recorded to analog or digital tape seems to be BFF (haven't looked at micromv footage, but that's hardly ever relevant).
As for the render times, the AI models are VERY intense as far as processing power. As mentioned, I'm using a RTX 3080 so I have plenty of very fast video memory as well as a fairly capable GPU and while I'm able to do 4K video, it takes something like 5 hours to do 2 minutes of 4K video with the Proteus model (for denoising and detail enhancing). Hell, the render times and the lackluster performance of the Chronos models are what ultimately pushed me to buy a Canon GX10 to replace my Canon HF G50.
I've noticed that all models seems to have an issue with audio sounding like it's done with low bitrate (even when exporting with the ProRes option) if they output audio at all. This seems to be a universal thing so I end up just taking the outputs and making extensive use of the "map" command in ffmpeg to strip the audio out of the original files (video codec as copy from source 0, audio codec as whatever will work in a MP4 container from source 1--apparently 24-bit PCM audio is not supported by the MP4 container so I have to use something like AAC).
As for upscaling, the results with SD to HD are more impressive than 4K to 8K out of my Canon GX10, but as for removing compression, denoising, and making the details a little bit more crispy, this does seem to be one of the best available options (beats the tar out of my "not entirely legitimate" copy of the NeatVideo filter in Sony/Magix Vegas before mopping the floor with it). I've had bad luck deinterlacing with this program and have defaulted to to using QTGMC since that's a pretty reliable standby (see above), but with already deinterlaced video, upscaling 1080p is a bit meh, upscaling SD to HD is damn impressive, working at the native resolution for denoising and removing compression artifacts is also damn impressive.
As an upscaler (with a starting resolution at already HD), deinterlacer, or frame doubler, I cannot recommend this program, though they have been making visible progress in those areas. As an enhancer (at native resolution) and an upscaler for SD video, I can *mostly* recommend this program (if the end user is familiar enough with avisynth to deinterlace their videos beforehand and ffmpeg to deal with the audio problems). The biggest problem that keeps me from fully recommending this program is that you need a really beefy computer to actually use it to its fullest potential (I have an 11th gen i7 and RTX 3080, my RTX 3080 is exponentially faster with the renders than my CPU and it's still not quite fast enough to be comfortably usable even with the video memory allotment slider cranked all the way up to 16GB).
I bought it and used it..... it appears to need a long way to go before it is good to go. I can't really tell a lot of difference from the original.
I've been wanting to try out the new version myself but haven't had any compelling reason to do so. It seems that the refreshed version has been oversimplified, though I'm going off of screenshots I've seen online.
I just tried it on an old vhs home movie and it did help a little.... it seems that if you have a fairly good video, it will make it better but I'm interested in making a not so good video a lot better... which I think it is showing it can do but in my opinion not really. I think the company who gets their stuff together will get rich in a hurry.
@@stereophonicstuff
Yeah you need a modern GPU to use AI software efficiently.
Any suggestions for a good middle of the road option that won’t break the bank?
@@stereophonicstuff PC Part Picker
UA-cam automatically deletes all my responses.
Decides to do a video about UPSCALING and then proceeds to record it in 720P. What was your starting resolution 240P? DOH!
This wasn’t recorded at 720p. It was recorded in 480i MPEG2 using a standard def camcorder that I deinterlaced and upscaled using QTGMC, NOT Video Enhance AI because it’s what I felt like using. DUH!
lame review
Lame comment
@@stereophonicstuff I don't think you could have done a worse job.
anybody want my build g1.assassin/16gb ram/d9dx cooler/this cpu/3060 for 570 or without the gpu for 300