But it ‘looks’ better when the list of issues is short!
2 Likes
All frame interpolation still creates large artifacts in any high speed scenes.
Original
24 to 50 FPS apollo v8
When a preview loops, it has a pause of about 3-5 seconds before playing again.
6 Likes
Unique to 4K and above resolutions. It’s does not do that on lower resolutions.
Ok but…WHY???
Thants Strange…
If i want to “slow down” a video, thant’s can’t be depending from resolution of that video!!
2 Likes
Thanks to topaz staff for the email, address, ⇒ help@topazlabs.com ⇐ it will prove to be of incredible help to us!
1 Like
Tried a few tests. Nothing has changed.
Yes, but if you try Nyx2 with original pixel, it oversharpen a lot more.
Yeah windowdressing by sales intended for a target group that is not typically Topaz consumers , it actually pretty much insulting , has Topaz marketing still not understood you can not fool Topaz consumers. Topaz sales please start selling oranges
That may be true, but it’s still too sharpened—making it useless for the stack of DVDs I was hoping run it on.
The issue topic I created detailing it got changed to deferred. I hope that means the AI model team knows about it and is working on version 3, or something.
2 Likes
Protos ‘Relative to Automatic’ does not work properly.
If I render with relative to automatic with all settings set to 0, the result differs significantly from automatic.
This should not be the case, because this way you cannot influence the settings in a predictable way.
And what is also extremely annoying:
View 1 is set to Original and View 2 has the focus. If I now set a in/out marker, View 1, which is set to Original, changes to a model by itself (although it has no focus) and displays ‘Not Processed’.
Also I don’t understand why there are so many problems with frame accurate scrolling. AviDemux also uses ffmpeg but has no problems as well as many other programs.
Maybe an ‘optional’ indexing of the video file could help.
4.0.0 Disaster – three versions further… Window is now dimmed… good, but the rest of the faulty design has not changed.
The idea of the 2 window and 2 model preview is not bad, but it has been implemented incorrectly.
Even simple things like a 100% zoom or saving the last settings in the Original tab so that they are not lost when the last preview is deleted are not fixed.
4 Likes
the 1 and 2 view switch is necessary for the people who like to have a single screen to compare the results, because we can’t switch between 1 and 2 unless we are viewing the side by side screens…
This is a little more than alarming.
Are you sure though? I just tried some tests, and they not only came out the exact same, the commands generated were the exact same.
Most annoyingly even in version 4.0.3 there is no option to pause the processing to continue later. I have more things to do than to just wait endless hours for Video AI to finish work and I can not use the GPU for more than one application at a time because if i try to either Video AI drops a red cross error soon or the other application just crashes! 
6 Likes
Confirmed that the dupe-bug is still present for videos that are not 25/30 fps, anything less, (23fps for example) produces dupes while using chronos, as with the last general release, so the bug still exists!
2 Likes
@wdsongama Fair enough. Then the two 1/2 buttons (active one blue) are a good for that. It can even have a CHANGE button next to it instead of the stretch/fit to screen symbol on the right.
Right now either the highlighted rectangle with digit or the “View 1” text look like a button.
agree, fix the viewer it is a pain in the xss was better in older ver
So anyway, 3 weeks on from 4.0.0, three new releases…
- Currently not all the models are supported on Intel ARC (model support will be added week to week).
So which three models are now working on Arc?
Wait, let me guess … SFA1, SFA2, and SFA3.
Hello,
I just did interpolation tests with 4.0.3 from an HD clip at 59,940 fps.
Once with Apollo 8 from 59,940 to 120 fps
Général
Nom complet : F:\temp\26 aout Test 59_940 ips 1080p.mp4
Format : MPEG-4
Profil du format : Base Media
Identifiant du codec : isom (isom/iso2/avc1/mp41)
Taille du fichier : 111 Mio
Durée : 2 min 39s
Débit global moyen : 5 811 kb/s
Images par seconde : 59,940 Im/s
Application utilisée : Lavf59.16.100
Vidéo
ID : 1
Format : AVC
Format/Info : Advanced Video Codec
Profil du format : High@L4.2
Paramètres du format : CABAC / 3 Ref Frames
Paramètres du format, CABAC : Oui
Paramètres du format, RefFrames : 3 images
Identifiant du codec : avc1
Identifiant du codec/Info : Advanced Video Coding
Durée : 2 min 39s
Débit : 5 670 kb/s
Largeur : 1 920 pixels
Hauteur : 1 080 pixels
Format à l'écran : 16/9
Type d'images/s : Variable
Images par seconde : 59,940 Im/s
Images/s mini : 59,920 Im/s
Images/s maxi : 59,960 Im/s
Espace de couleurs : YUV
Sous-échantillonnage de la chrominance : 4:2:0
Profondeur des couleurs : 8 bits
Type de balayage : Progressif
Bits/(Pixel*Image) : 0.046
Taille du flux : 108 Mio (98%)
Gamme de couleurs : Limited
Coordonnées de chromaticité : BT.709
Caractéristiques du transfert : BT.709
Coefficients de la matrice : BT.709
mdhd_Duration : 159593
Codec configuration box : avcC
Audio
ID : 2
Format : AAC LC
Format/Info : Advanced Audio Codec Low Complexity
Identifiant du codec : mp4a-40-2
Durée : 2 min 39s
Type de débit : Constant
Débit : 128 kb/s
Canaux : 2 canaux
Channel layout : L R
Echantillonnage : 44,1 kHz
Images par seconde : 43,066 Im/s (1024 SPF)
Mode de compression : Avec perte
Taille du flux : 2,44 Mio (2%)
Titre : ISO Media file produced by Google Inc.
Default : Oui
Alternate group : 1target :
Général
Nom complet : H:\TEMP\test_1\26 aout Test 59_940 ips 1080p_apo8.mov
Format : MPEG-4
Profil du format : QuickTime
Identifiant du codec : qt 0000.02 (qt )
Taille du fichier : 22,3 Gio
Durée : 2 min 39s
Type de débit global : Variable
Débit global moyen : 1 200 Mb/s
Images par seconde : 120,000 Im/s
major_brand : qt
minor_version : 512
compatible_brands : qt
videoai : Slowmo 100% and framerate changed to 120 using apo-8 ignoring duplicate frames
encoder : Lavf60.14.101
Vidéo
ID : 1
Format : ProRes
Version du format : Version 0
Profil du format : 422 HQ
Identifiant du codec : apch
Durée : 2 min 39s
Type de débit : Variable
Débit : 1 200 Mb/s
Largeur : 1 920 pixels
Hauteur : 1 080 pixels
Format à l'écran : 16/9
Type d'images/s : Constant
Images par seconde : 120,000 Im/s
Espace de couleurs : YUV
Sous-échantillonnage de la chrominance : 4:2:2
Type de balayage : Progressif
Bits/(Pixel*Image) : 4.823
Taille du flux : 22,3 Gio (100%)
Bibliothèque utilisée : Apple
Langue : AnglaisThen the same source file I used Chronos for interpolation of 59,940 to 25 fps
Général
Nom complet : H:\TEMP\26 aout Test 59_940 ips 1080p_chr2.mov
Format : MPEG-4
Profil du format : QuickTime
Identifiant du codec : qt 0000.02 (qt )
Taille du fichier : 4,89 Gio
Durée : 2 min 39s
Type de débit global : Variable
Débit global moyen : 263 Mb/s
Images par seconde : 25,000 Im/s
major_brand : isom
minor_version : 512
compatible_brands : isomiso2avc1mp41
videoai : Slowmo 100% and framerate changed to 25 using chr-2 ignoring duplicate frames
encoder : Lavf60.14.101
Vidéo
ID : 1
Format : ProRes
Version du format : Version 0
Profil du format : 422 HQ
Identifiant du codec : apch
Durée : 2 min 39s
Type de débit : Variable
Débit : 261 Mb/s
Largeur : 1 920 pixels
Hauteur : 1 080 pixels
Format à l'écran : 16/9
Type d'images/s : Constant
Images par seconde : 25,000 Im/s
Espace de couleurs : YUV
Sous-échantillonnage de la chrominance : 4:2:2
Type de balayage : Progressif
Bits/(Pixel*Image) : 5.034
Taille du flux : 4,85 Gio (99%)
Bibliothèque utilisée : Apple
Coordonnées de chromaticité : BT.709
Caractéristiques du transfert : BT.709
Coefficients de la matrice : BT.709
Audio
ID : 2
Format : PCM
Paramètres du format : Little / Signed
Identifiant du codec : in24
Durée : 2 min 39s
Type de débit : Constant
Débit : 2 117 kb/s
Canaux : 2 canaux
Channel layout : L R
Echantillonnage : 44,1 kHz
Profondeur des couleurs : 24 bits
Taille du flux : 40,3 Mio (1%)
Titre : ISO Media file produced by Google Inc.
Default : Oui
Alternate group : 1And to finish I created a script in avisynth+ with advice to test if there were exact dupes using the exactdup plugin for each of the interpolated files (here for the last)
Film1 = "H:\Temp\August 26 Test 59_940 IPS 1080p_chr2.mov"
Ffmpegsource2 (film1, vtrack = -1, atrack = -2, cache = true, cachefile = "", fpsnum = -1, fpsden = 1, \
Threads = -1, timecodes = "", Seekmode = 1, Overwrite = False, Width = -1, Height = -1, \
Résizer = "bicubic", colorspace = "yv12")
Exactdup (FirstPass = False, Dupinfo = "Dupinfo.txt", Times = "Times.txt", MaxdupCount = 0, Keeplastframe = False, Show = True)Exactdup works in 2 passes (firstPass = True and FirstPass = False) with all others same parameters
If he finds a real dupe bit for bit, he will dismiss it.
Each time I send the scripts one by one in virtualduB2 and generates an output file for each pass. As at the end, the outputs pass 1 and pass 2 have the same number of images, there is no dupe.
Do I have something wrong?
Do you have Windows Display Scaling enabled?
