Major internal changes have been made in this build. There may be more issues than usual. Please include logs and problematic videos when reporting issues to the thread.
Another beta for Topaz Video AI 3.0 is available. Please test and give us your feedback. Thank you.
Please upload videos that are not working or having any type of conflicts to this DropBox Folder .
suggestion : gray color font on blue is hard to read, for preview and export, better would be white on blue .
the green mark with the blue/purple background is as well impossible to see.
suggestion : having the possibility to have a display by default (like fit to screen.)
there is stop and process again but didnât see a stop and resume, so i assume itâs stop and process again, and pause and resuming is not there yet ;).
the list has been reversed again from last file rendered at the end of the list instead of top, I personnally prefer the most recent file at the top and disagree with the change, i know there is one who disagree, (maybe more ?), but i still think itâs much better to use.
thanks for the playhead cursor staying in position, the crop issue fixed (the way it does it is quite funny), the crop staying on the source display/preview.
The thin progress bar is difficult to see now that the entire job bar is being highlighted in the same colourâŠone of the colours needs to be changed. As per Marty the text is also hard to see.
Confirmation to close application no longer pops up, even when an export job is running, this is a big problem and will create some angry users when their export jobs get closed accidentally - please re-enable prompt on application exit/close!
Why are the files in the preview and export window going newest at the bottom again. FIFO is the way it should be, just like you had changed it to a couple of releases ago.
Most recent on top please, just like in Outlook and every email program, you donât scroll the list to get to the bottom to see your emails. It makes sense to see readily what you just did.
The Chronos Fast model struggles to deal with peopleâs handwaving (as in waving goodbye). When it tries to increase the frame rate, there is flickering. I will upload an example to Dropbox for the dev. team to review.
I am getting âModel Not foundâ or âModel Failedâ for Gaia and Theia. I donât think this is new for me to 3.0.0.7.b because I have been mostly testing Proteus and I think Iâve seen the same errors recently on Gaia and Theia.
I just tried uninstalling VEIA and renaming the existing âC:\ProgramData\Topaz Labs LLC\Topaz Video AI BETAâ before installing again. Installation was ok with new models folder created. Preview says downloading and running model but then fails for both Gaia and Theia consistently.
This is Theia Model Not Found Applying option map_metadata:s:a (set metadata information of outfile from infile) with argument 0:s:a. Applying option map (set input stream mapping) with argument 0:a. Applying option c:a (codec name) with argument copy. Applying option metadata (add metadata) with argument videoai=Enhanced using thf-4 with reduce noise at 0, sharpen at 0, and revert compression at 0. Successfully parsed a group of options. Opening an output file: C:/Users/john/AppData/Local/Temp/268545064/Dark Shadows 0223_406955519_thf4.mp4. Setting default whitelist âfile,crypto,dataâ Successfully opened the file. detected 16 logical cores cur_dts is invalid st:0 (0) [init:0 i_done:0 finish:0] (this is harmless if it occurs once at the start per stream) Output size set to: 2560 1548 Here Config props model with params: up thf-4 4 1 1 1 640 387 0.000023 25.000000 2022-09-17 13-10-07 Thread: 21280 Info Model: âthf-4â Device: 1 VRAM: 0 threads: 1 Failed to configure output pad on Parsed_veai_up_1 Uninit called for thf-4 1 Error reinitializing filters! Failed to inject frame into filter network: Invalid argument Error while processing the decoded data for stream #0:1 Statistics: 0 bytes written, 0 seeks, 0 writeouts Statistics: 854695 bytes read, 3 seeks
This is Model Failed on Gaia Output size set to: 2560 1548 Here Config props model with params: up ghq-5 4 1 1 1 640 387 0.000023 25.000000 2022-09-17 13-08-19 Thread: 44704 Info Model: âghq-5â Device: 1 VRAM: 0 threads: 1 Failed to configure output pad on Parsed_veai_up_1 Uninit called for ghq-5 1 Error reinitializing filters! Failed to inject frame into filter network: Invalid argument Error while processing the decoded data for stream #0:1 Statistics: 0 bytes written, 0 seeks, 0 writeouts Statistics: 854695 bytes read, 3 seeks
@ida.topazlabs@suraj One thing which would speed up things a lot for my usecase, would be the possibility of scaling straight to 1.5x with the AI filter itself. Would that be possible?
(I have a folder with about a hundred 4096x4096/3840x3840/3840x2560px stereoscopic 360 degree-videos, and upscale them to double the size, but unfortunately after that I have to shrink them under the max of 35 million pixels (with the scale_qsv filter) which is the maximum the HW H265 decoder of the Meta Quest 2 can handle. It would speed things up a lot to just upscale 1.5x.
Exporting to h264 would be possible but would either be a bit blurry or stutter because of the high bitrate demandâŠ)
You open the video file, use the cut tool to select start and end, choose your processing settings and begin processing. You can close this (not the app, only the watching processing) and then choose another start and end via cut on the same video with new settings of your choice and then start processing these. Now you find two running exports on the lower right which you can observe both by clicking onto one or the other.
Thank you! Iâm not sure if we mean the same thing⊠I mean the ffmpeg-filter-parameter, e.g.
ffmpeg -i abc.mp4 -movflags use_metadata_tags -vf veai_up=model=alqs-2:scale=2:instances=0,veai_fi=model=chf-3:fps=59.94:device=0:vram=1:instances=0,scale=w=7200:h=4861:flags=lanczos:threads=0 -c:v hevc_nvenc -profile main -preset p7 -cq:v 17 -maxrate:v 100M -c:a copy
