- Same here.
I had a similar case with VEAI. During the rendering process, my PC was giving me bluescreens almost every time. And it was very really annoying. After decrypting the logs of all the BSODs and averaging them, I deduced that it was the BIOS/UEFI. I updated it and never had another BSOD.
interesting, i havenât tried that yet.
thanks for letting me know. iâll try that when iâm back home next week
Load previous session please. When I have many videos in queue and it crashes, having the ability to continue where it left off would be great. after every video is completed the save should be updated to not include the finished videos but only the ones not completed. maybe a future version.
hi , iâm not sure of this but
- starting again en encoding in a middle of a video is not possible with H264 (donât know about ProRes) only the export as picture is possible.
- here (that does not mean it did for you), if it crash after processing many videos, when i relaunch the autosave backup, the processed videos are not removed, but if you look at the thumb, there is an arrow in a white box, it means the video has been processed. you can put your mouse on the thumb as well, it will tell you that the video has been processed and give you the time it took. unless itâs not 2000 videos imported, you can remove them manually until a feature exist to remove them automatically.
Proteus is a wonderful model. But, like all other models, face upscaling is bad. This is especially noticeable in Gaia models. Artemis is better, but often it just adds âsoapâ, without adding any details or image improvements.
Proteus also sometimes does strange things with faces, especially with eyes. When the face is close-up, it works well. But the faces in the distance either do not improve, or turn into an even bigger mess.
Perhaps it makes sense to process human faces with a separate neural network?
great update as always, thank you topaz team
but in all honesty, i think lack of hevc encoding and very poor selection of encoder settings (just one sliderâŠ) is a huge disservice to the entire program. yes, i can export as png and encode with any program i like, but this requires user interaction - and for a program of this price, itâs honestly unacceptableâŠ
guess iâll keep waiting
Iâm afraid we donât have the same view of the word âunacceptableâ.
First there will be no HEVC encoding so iâm afraid you can wait forever. Fault is not totally on Topaz Side. issues are on the Mpeg side too.
But If you find a better software who does the same thing for this price, more or less, please tell us, because there is just not a single software able to do as best or better, no matter the price and available for the normal user for now. Today the only and good option is here.
I find with Proteus and the eyes problem that dropping back on the Revert Compression, Recover Details and Sharpen helps a lot. While it may not be as sharp most of the sharpness can be helped with a little post processing with something like StaxRip or any other package that you prefer.
issues? what issues? what causes an issue for an advanced program with an asking price of $300, and why is it not a problem to countless free and open source programs that employ HEVC encoding without any issue?
also the argument about no better alternatives - really? ok, i understand a program that doesnât really have much other competition doesnât need to improve or listen to people all that much, because what else are people going to use? i guess i should have set my bar lower than i initially had it set to, thanks for the reality check.
- to use HEVC in a COMMERCIAL software you need to pay a licence to the Mpeg fondation. Thatâs why some company didnât implanted Mp3 as an encoder during a long time except as a separated purchase / feature. it was the case of Steinberg Cubase during a long time, same with Sony (Vegas / Soundforge) etc⊠Thatâs one of the issue.
The second one is a technical problem which will make VEAI working Worst. I can assure you from what i know (donât know if iâm allowed to tell this publically thatâs why i donât tell it clearly sorry), that if they implement HEVC as saving, youâll go back immediatly to use H264 instead (or ProRes) or picture export (PNG / Tiff).
- there is absolutly at the moment no other GOOD alternatives ! There are some, but if you want to use them, please do it. Even one of the best have been removed and is not sold anymore (a 4K upscaler by Red Giant).
youâll see that youâll come back to VEAI pretty fast, because what you talk about VEAI is just stuff that are easy to deal with. I donât know in which area you work or are (video professionnal or amateur), but I can assure you that for what VEAI does, 300$ is very very cheap !
what stops them from going the âaudacityâ route? add support for hevc but require people to download and setup ffmpeg themselves? they will not be distributing the x265 encoder this way so no licensing fees apply, right?
i know what i need hevc for (basically: VR videos, which playback can be GPU accelerated beyond 4096x4096, unlike avc) so no, i would not go back to ProRes at any point. âmaking VEAI working worstâ shouldnât be a problem if itâs done rightâŠ
i already expressed what i think about this silly âno other good alternativesâ argument.
ps: and donât they have to pay licensing fees for h.264/avc as well? also iâd love someone whoâs well versed in codec licensing to explain to me how much fees would topaz labs potentially have to pay, because if what googling is telling me, below 100k âdevicesâ sold this codec is loyalty free (did VEAI really sell more than that?), and past that itâs something like $2 per device - thatâs like below 1% of the programâs value⊠but iâm most likely just a dumbarse and i donât know the full story, so please tell me
Reducing the values of âRecover Detailsâ and âSharpenâ partially solves the problem with face artifacts, yes. But at the same time, the overall image quality deteriorates (it becomes less sharp and detailed). Therefore, I assume that processing parts of the image with faces should be a separate neural network. A similar approach, as far as I know, is used in Gigapixel.
x265 encoding is extremely slow and would be a huge bottleneck for VEAI during a job. You shouldnât be producing a finished product encode from VEAI anyway, but rather something nearly-lossless that you can toss into another specialized program like Handbrake to do the final encode with more control over the quality of that encode.
Whenever Iâm making something (like a long 4K video) that will end up being an x265 at the end of my workflow, I do a VEAI job using 264 at a virtually lossless CRF like 12, and then toss that into Handbrake and do a 265 encode at Slow speed at 22 CRF and let it bake for a day. Even if I was going to use 264 for the final encode, Iâd still do the same process in VEAI first and then toss it into Handbrake for a Very Slow encode to get a much more efficient final compression.
Read what thespiffyitalian wrote, the first line, if youâre âsmartâ, youâll understand why.
The licence is not the main issue.
Like everybody you want a all in one product who does everything. The best example is the remastering / Upscaling of the Star Wars original trilogy that 2 guys are doing (easy to find) and are very well known. they did a better job of what Disney / Lucasfilm did for the Bluray. but until now, the result was only 720p.
this is years of work for making the upscalling acceptable, restauring 35mm original films, LAserdisc source, Dvd , Bluray etc⊠but recently one of them told that he went to the VEAI route to deliver a 1080p upscale and cancelled the work he did so far. that would give you an idea about the lack of any other âeasyâ alternative available, and why.
of course x265 is slower⊠would it be a bottleneck? depends on settings, only if we had access to those⊠yesterdayâs VEAI job took me about 1.5h to complete (png output), and converting the image sequence to h.265 on slower preset and 80% quality in aftercodecs took me about 30 minutes - so no, it wouldnât necessarily be a bottleneck, itâs up to your system configuration. but it does waste me plenty of time on bigger projects, which would eg. take 5-6h to get done in VEAI and almost 2 hours to encode in h.265 - if this codec was supported by VEAI, i would be able to just click âprocessâ and let it do everything while iâm away, and come back to a ready file. instead iâm wasting extra time for encoding.
if adding an encoder is such a huge problem, then at least add a very simple feature that makes VEAI execute a file (like a batch file) after processing is done - then weâll be able to use ffmpeg for whatever codec we want. bonus points if VEAI will provide some variables that could be used as arguments for the command (like output directory, resolution, framerate). at this point iâd be very content with that, it would open up a lot of opportunities for automation and integration
ps: if the process of simultaneous upscaling and encoding would be properly separated/threaded, then the concept of bottlenecking in this part of the program wouldnât even be an issue, if the encoder is slower than the upscaler then just let the upscaler continue the job while the encoder is working at its own pace⊠at least some part of encoding would be done, so even in that case it would help save some time. itâs already been done, a free tool called flowframes does that, except instead of upscaling it interpolates to higher framerates.
If an h265 encode for a 4K video on Slower is taking you 30 minutes then youâre likely using a hardware encoder (which is for live streaming and optimized for speed) rather than a software encoder (which is optimized for quality and file-size). In 99% of cases youâre going to want to use a software encoder, and at Slow speed thatâs going to mean a few hours per minute of video. h265 is really intended for cases where youâre willing to spend a lot of extra time and computing power to compress a 4K video. Iâd much rather have VEAI focus on processing a video ASAP in a relatively lossless format so that I can make sure Iâm happy with the result before spending the extra time running it through a real-deal h265 encode.
Does it use software (CPU) encoding or hardware (GPU) encoding ?
VEAI use CPU for encoding & decoding.
When using CPU, H.265 is far more difficult to encode due to its complexity, and can require up to 10 times the compute power to encode over H.264.
no, i am not using GPU encoding, iâm using CPU. that clip was encoded at 6000x3000 resolution, but was only 20 seconds long.