# Topaz Video AI 6.0.2 + 6.0.3 **URL:** \ **Category:** Releases\ **Created:** [January 22, 2025, 11:50pm UTC](https://community.topazlabs.com/t/topaz-video-ai-6-0-2-6-0-3/86462 "2025-01-22T23:50:43Z")\ **Posts on this page:** 20\ **Page:** 5
**Author:** ![markwaltermd](https://community.topazlabs.com/letter_avatar_proxy/v4/letter/m/848f3c/32.png) [@markwaltermd](https://community.topazlabs.com/u/markwaltermd)\ **Post date:** [January 25, 2025, 5:38pm UTC](https://community.topazlabs.com/t/topaz-video-ai-6-0-2-6-0-3/86462/87 "2025-01-25T17:38:08Z")
That would be amazing! ---
**Author:** ![menditsa](https://community.topazlabs.com/user_avatar/community.topazlabs.com/menditsa/32/182463_2.png) [@menditsa](https://community.topazlabs.com/u/menditsa)\ **Post date:** [January 25, 2025, 5:45pm UTC](https://community.topazlabs.com/t/topaz-video-ai-6-0-2-6-0-3/86462/88 "2025-01-25T17:45:50Z")
Give it a test on files you know have been saved as vfr and let Tony know. My post includes a command to verify vfr or cfr cfr will always be 0.00000 šŸ‘ ---
**Author:** ![david.123](https://community.topazlabs.com/user_avatar/community.topazlabs.com/david.123/32/163156_2.png) [@david.123](https://community.topazlabs.com/u/david.123)\ **Post date:** [January 25, 2025, 6:14pm UTC](https://community.topazlabs.com/t/topaz-video-ai-6-0-2-6-0-3/86462/89 "2025-01-25T18:14:52Z")
That wonā€™t work as Topaz build their own customized FFMpeg to make it work with the TVAI filters. ---
**Author:** ![DannyEl](https://community.topazlabs.com/letter_avatar_proxy/v4/letter/d/ed8c4c/32.png) [@DannyEl](https://community.topazlabs.com/u/DannyEl)\ **Post date:** [January 25, 2025, 8:14pm UTC](https://community.topazlabs.com/t/topaz-video-ai-6-0-2-6-0-3/86462/90 "2025-01-25T20:14:32Z")
Confirmed. Copy/paste my reply from the beta section. ā€œI did a really short test using the GUI and MediaInfo shows a CFR (test done from 25fps to 60fps). I will test more thoroughly after I finish my queue since I donā€™t like to pause it.ā€ EDIT: A little bit more extended test (500 frames). Enhanced with Rhea, x264 codec, CFR 5Mb/s ā†’ OK. Interpolated same video from 30fps to 60 fps with Chronos, x265 codec (both Main and Main10), CFR 5Mb/s ā†’ OK. This might be the end of this headache. EDIT 2: It looks like the CFR is fixed someway in the enhancement models in the beta version but there are still issues with the interpolation. When you interpolate an entire video it creates some weird VFR. > Video > ID : 1 > Format : HEVC > Format/Info : High Efficiency Video Coding > Format profile : Main@L4.1@Main > Codec ID : hvc1 > Codec ID/Info : High Efficiency Video Coding > Duration : 3 min 4 s > Source duration : 3 min 4 s > Bit rate : 5 014 kb/s > Width : 1 920 pixels > Height : 1 080 pixels > Display aspect ratio : 16:9 > Original display aspect ratio : 16:9 > Frame rate mode : Variable > Frame rate : 60.000 FPS > Minimum frame rate : 30.000 FPS > Maximum frame rate : 15 360.000 FPS ---
**Author:** ![RDPA](https://community.topazlabs.com/letter_avatar_proxy/v4/letter/r/ce73a5/32.png) [@RDPA](https://community.topazlabs.com/u/RDPA)\ **Post date:** [January 26, 2025, 4:10pm UTC](https://community.topazlabs.com/t/topaz-video-ai-6-0-2-6-0-3/86462/91 "2025-01-26T16:10:41Z")
I just opened up the app this morning and got an update for 6.0.3. I let it update. I donā€™t see the release notes here on the community forum or the standalone link to download yet. Iā€™ll do a DVD to FHD 1st pass and Iris Low Q 2nd pass on the same DVD and see if itā€™s good. ---
**Author:** ![JakSpoon](https://community.topazlabs.com/user_avatar/community.topazlabs.com/jakspoon/32/40644_2.png) [@JakSpoon](https://community.topazlabs.com/u/JakSpoon)\ **Post date:** [January 26, 2025, 5:02pm UTC](https://community.topazlabs.com/t/topaz-video-ai-6-0-2-6-0-3/86462/92 "2025-01-26T17:02:14Z")
6.0.3 stable released with VFR fix??? GREATā€¦ thanks! @tony.topazlabs ---
**Author:** ![TPX](https://community.topazlabs.com/user_avatar/community.topazlabs.com/tpx/32/194727_2.png) [@TPX](https://community.topazlabs.com/u/TPX)\ **Post date:** [January 26, 2025, 8:10pm UTC](https://community.topazlabs.com/t/topaz-video-ai-6-0-2-6-0-3/86462/93 "2025-01-26T20:10:48Z")
> **[Getting NVIDIA's RTX 50 Series GPUs At MSRP Would Be A "Surprise"; Team Green...](https://wccftech.com/getting-nvidia-rtx-50-series-gpus-at-msrp-would-be-a-surprise/)** > > NVIDIA's GeForce RTX 50 series GPUs, particularly the higher-end variants, won't be available at MSRPs even after launch. ---
**Author:** ![menizzi](https://community.topazlabs.com/letter_avatar_proxy/v4/letter/m/cab0a1/32.png) [@menizzi](https://community.topazlabs.com/u/menizzi)\ **Post date:** [January 26, 2025, 8:23pm UTC](https://community.topazlabs.com/t/topaz-video-ai-6-0-2-6-0-3/86462/94 "2025-01-26T20:23:56Z")
Mine will come in my Falcon Northwest system and my 285k build ---
**Author:** ![david.123](https://community.topazlabs.com/user_avatar/community.topazlabs.com/david.123/32/163156_2.png) [@david.123](https://community.topazlabs.com/u/david.123)\ **Post date:** [January 26, 2025, 8:39pm UTC](https://community.topazlabs.com/t/topaz-video-ai-6-0-2-6-0-3/86462/95 "2025-01-26T20:39:27Z")
5090 = 4090Ti If you have a 4090, unless you like wasting money, I would wait for 60 series, which will be on new fab process, thinner wafer, more efficient! ---
**Author:** ![TPX](https://community.topazlabs.com/user_avatar/community.topazlabs.com/tpx/32/194727_2.png) [@TPX](https://community.topazlabs.com/u/TPX)\ **Post date:** [January 26, 2025, 8:39pm UTC](https://community.topazlabs.com/t/topaz-video-ai-6-0-2-6-0-3/86462/96 "2025-01-26T20:39:35Z")
Great, so you will betatest for me. šŸ˜ƒ ---
**Author:** ![TPX](https://community.topazlabs.com/user_avatar/community.topazlabs.com/tpx/32/194727_2.png) [@TPX](https://community.topazlabs.com/u/TPX)\ **Post date:** [January 26, 2025, 8:41pm UTC](https://community.topazlabs.com/t/topaz-video-ai-6-0-2-6-0-3/86462/97 "2025-01-26T20:41:36Z")
> [@david.123](#): > > If you have a 4090, unless you like wasting money, I would wait for 60 series, which will be on new fab process, thinner wafer, more efficient! That depends on what tensorRT does with Video AI and how the performance is in Gigapixel AI, Capture One and Resolve once the software has been adapted to the hardware. But knowing the internet, I have to test almost everything myself. ---
**Author:** ![david.123](https://community.topazlabs.com/user_avatar/community.topazlabs.com/david.123/32/163156_2.png) [@david.123](https://community.topazlabs.com/u/david.123)\ **Post date:** [January 26, 2025, 8:43pm UTC](https://community.topazlabs.com/t/topaz-video-ai-6-0-2-6-0-3/86462/98 "2025-01-26T20:43:04Z")
> [@TPX](#): > > once the software has been adapted to the hardware. Topaz never optimized anything specifically for the 4090 so donā€™t expect it on the 5090. All they did was release TensorRT models. ---
**Author:** ![david.123](https://community.topazlabs.com/user_avatar/community.topazlabs.com/david.123/32/163156_2.png) [@david.123](https://community.topazlabs.com/u/david.123)\ **Post date:** [January 26, 2025, 8:44pm UTC](https://community.topazlabs.com/t/topaz-video-ai-6-0-2-6-0-3/86462/99 "2025-01-26T20:44:47Z")
As we have learned the most important component is how much **L2** cache is on the 5090? Nvidia always skimping where it actually matters. Edit: found this: ![image](https://community-cdn.topazlabs.com/original/3X/5/4/54e2e3bc6ba87a704d9991b71216bf8a02aa7195.png) They should have doubled it to 144MB! $5090 ---
**Author:** ![david.123](https://community.topazlabs.com/user_avatar/community.topazlabs.com/david.123/32/163156_2.png) [@david.123](https://community.topazlabs.com/u/david.123)\ **Post date:** [January 26, 2025, 8:49pm UTC](https://community.topazlabs.com/t/topaz-video-ai-6-0-2-6-0-3/86462/100 "2025-01-26T20:49:35Z")
Anyway, if you do get a 5090 Iā€™ll be looking forward to your benchmarks! ---
**Author:** ![TPX](https://community.topazlabs.com/user_avatar/community.topazlabs.com/tpx/32/194727_2.png) [@TPX](https://community.topazlabs.com/u/TPX)\ **Post date:** [January 26, 2025, 8:53pm UTC](https://community.topazlabs.com/t/topaz-video-ai-6-0-2-6-0-3/86462/101 "2025-01-26T20:53:37Z")
> [@david.123](#): > > They should have doubled it to 144MB! 960 mb was the sweet spot for a architecture paper. [![](https://research.nvidia.com/sites/default/files/styles/wide/public/publications/Fig1.jpg?itok=Ubcdz5A7) ](https://research.nvidia.com/sites/default/files/styles/wide/public/publications/Fig1.jpg?itok=Ubcdz5A7) > **[GPU Domain Specialization via Composable On-Package Architecture | Research](https://research.nvidia.com/publication/2021-12_gpu-domain-specialization-composable-package-architecture)** > > As GPUs scale their low-precision matrix math throughput to boost deep learning (DL) performance, theyĀ upset the balance between math throughput and memory system capabilities. We demonstrate that a convergedĀ GPU design trying to address diverging... > **[GPU Domain Specialization via Composable On-Package Architecture | ACM...](https://dl.acm.org/doi/full/10.1145/3484505)** > > As GPUs scale their low-precision matrix math throughput to boost deep learning (DL) > performance, they upset the balance between math throughput and memory system capabilities. > We demonstrate that a converged GPU design trying to address diverging... ---
**Author:** ![menizzi](https://community.topazlabs.com/letter_avatar_proxy/v4/letter/m/cab0a1/32.png) [@menizzi](https://community.topazlabs.com/u/menizzi)\ **Post date:** [January 26, 2025, 9:42pm UTC](https://community.topazlabs.com/t/topaz-video-ai-6-0-2-6-0-3/86462/102 "2025-01-26T21:42:21Z")
I work offshore so i will be back in 3 months so i guess most of the testing will be done. But if not I will be doing some stuff ---
**Author:** ![menizzi](https://community.topazlabs.com/letter_avatar_proxy/v4/letter/m/cab0a1/32.png) [@menizzi](https://community.topazlabs.com/u/menizzi)\ **Post date:** [January 26, 2025, 9:44pm UTC](https://community.topazlabs.com/t/topaz-video-ai-6-0-2-6-0-3/86462/103 "2025-01-26T21:44:14Z")
lol as if this is the only thing to use a 5090 on ---
**Author:** ![david.123](https://community.topazlabs.com/user_avatar/community.topazlabs.com/david.123/32/163156_2.png) [@david.123](https://community.topazlabs.com/u/david.123)\ **Post date:** [January 26, 2025, 9:45pm UTC](https://community.topazlabs.com/t/topaz-video-ai-6-0-2-6-0-3/86462/104 "2025-01-26T21:45:54Z")
> [@menizzi](#): > > 5090 \*4090Ti ---
**Author:** ![jojje](https://community.topazlabs.com/letter_avatar_proxy/v4/letter/j/b19c9b/32.png) [@jojje](https://community.topazlabs.com/u/jojje)\ **Post date:** [January 26, 2025, 11:28pm UTC](https://community.topazlabs.com/t/topaz-video-ai-6-0-2-6-0-3/86462/105 "2025-01-26T23:28:32Z")
> [@tony.topazlabs](#): > > we have seen some reports that ffmpeg exports even without tvai\_up are exporting as VFR. Has anyone experienced this with a Video AI ffmpeg build or a different one? Any data on this bug helps Tony, as a followup post to my other one, one which I dug up from an old post-mortem issue with an old pipeline. The issue could stem from a timestamp quantization issue which affects non-integer frame rates (like 23.976), specifically when using containers with fixed time bases, such as milliseconds, for which integer frame rates (25, 30) donā€™t exhibit the problem. hereā€™s very simple example that shows how a CFR source can become subject to VFR interpretation. ```auto ffmpeg -y -f lavfi -i "color=size=2x2:rate=24000/1001:duration=5" -c:v libx264 -g 30 -pix_fmt yuv420p src.mkv ``` Now this clip is clearly CFR. Does ffmpeg agree? ```auto ffmpeg -i src.mkv -vf vfrdet -f null - 2>&1| grep vfrdet [Parsed_vfrdet_0 @ 0000028923737900] VFR:nan (0/0) [Parsed_vfrdet_0 @ 0000028924464bc0] VFR:0.579832 (69/50) min: 41 max: 42 avg: 41 ``` Nope. Insta-fail. Ok, does Mediainfo agree with ffmpeg? ```auto mediainfo src.mkv | grep "Frame rate" Frame rate mode : Constant Frame rate : 23.976 (24000/1001) FPS ``` Nope. But it was right in this case. So why are the two behaving differently? IIRC the doom9 discussion netted down to ffmpeg checking if the framerate is an even multiple of the time base or not, while mediainfo uses some other heuristic. So ```auto ffprobe src.mkv -show_streams 2>&1 | grep -E "(frame_rate|time_base)=" r_frame_rate=24000/1001 avg_frame_rate=24000/1001 time_base=1/1000 ``` A frameā€™s presentation time canā€™t of course be expressed accurately when the bases differ and the time base isnā€™t a function of the frame rate nominator itself. This problem doesnā€™t happen for even framerates such as for 25 and 30, but is a pain with the US film and ā€œntscā€ standards. So how about mp4 then? ```auto ffmpeg -y -f lavfi -i "color=size=2x2:rate=24000/1001:duration=5" -c:v libx264 -g 30 -pix_fmt yuv420p src.mkv ffmpeg -i src.mp4 -vf vfrdet -f null - 2>&1| grep vfrdet [Parsed_vfrdet_0 @ 0000024bb13fa200] VFR:nan (0/0) [Parsed_vfrdet_0 @ 0000024bb2397640] VFR:0.000000 (0/119) mediainfo src.mp4 | grep "Frame rate" Frame rate mode : Constant Frame rate : 23.976 (24000/1001) FPS ffprobe src.mp4 -show_streams 2>&1 | grep -E "(frame_rate|time_base)=" r_frame_rate=24000/1001 avg_frame_rate=24000/1001 time_base=1/24000 ``` Oh, a time base derived from the frame rate, how convenient šŸ™‚ Same issue happens with transport streams for instance due to the same reason. Now is ffmpeg wrong? Well Yes, in practice. But not from its own definition of what CFR means. As you know, one can easily see that a clip is CFR if there is a short repeating pts distance pattern between even and odd time deltas. And this is what many tools take into account (though apparently not this one). ```auto fprobe -v error -show_frames src.mp4 | grep pts= | head -5 pts=0 pts=1001 # 1001 timebases since previous pts=2002 # 1001 timebases since previous pts=3003 # 1001 pts=4004 # 1001 ... ffprobe -v error -show_frames src.mkv|grep pts=|head -5 pts=0 pts=42 # 42 timebases since previous pts=83 # 41 timebases since previous pts=125 # 42 pts=167 # 42 pts=209 # 42 pts=250 # 41 4 frames since last odd pts=292 # 42 pts=334 # 42 pts=375 # 41 3 frames since last odd pts=417 # 42 pts=459 # 42 pts=501 # 42 pts=542 # 41 4 frames since last odd pts=584 # 42 pts=626 # 42 pts=667 # 41 3 frames since last odd pts=709 # 42 pts=751 # 42 pts=792 # 41 3 frames since last odd pts=834 # 42 pts=876 # 42 pts=918 # 42 pts=959 # 41 4 frames since last odd ``` This looks super weird, so ffmpeg must be right? Well, it depends on oneā€™s definition of CFR. Since we know CFR clips can be played back smoothly with no observable time shift or judder due to screens not having a 1khz refresh rate (yet), one millisecond more or less doesnā€™t matter during playback. As such a CFR clip should exhibit a cycling pattern (due to rounding) when the rate and base isnā€™t evenly divisible, but the source is known to be constant rate. Here are the odd (41ms) frame distances for the sec.mkv file: 4343343434334343433434343343434334 We can clearly see itā€™s composed by a single repeating pattern: 3343434 as in ā€œ434 3343434 3343434 3343434 3343434 334ā€ So an improvement to the vfr\_det algorithm in ffmpeg would consider the presence of a repeating cycle throughout the clip before making a snap decision regarding VFR. This is what other tools do after all, something a bit more sophisticated than just checking the frame rate to time base ratio. So how does this apply to the VAI bug people keep mentioning? No idea. Perhaps format conversion is involved, such as sources not being mp4 but output is, or vice versa. Or it might be that you do have a bug in the software where time bases and frame rates arenā€™t handled properly. I can see how this would be an issue regarding your framerate manipulation models and the code that wraps the inference calls. For the upscale ones though it shouldnā€™t matter as long as the inputā€™s timebase and framerate is just carried over to the output stream since you just do 1:1 frame manipulation in those filters. Let me know if this led somewhere or if I wrote all this in vainā€¦ _PS. my mentioning of split and stitch was a bit of a red herring, since I did use different container formats in such a pipeline when I first ran into this issue. My fuzzy memory just mixed the two up initially. But the above should distil my current understanding of that root cause._ ---
**Author:** ![ujak](https://community.topazlabs.com/letter_avatar_proxy/v4/letter/u/ecd19e/32.png) [@ujak](https://community.topazlabs.com/u/ujak)\ **Post date:** [January 26, 2025, 11:56pm UTC](https://community.topazlabs.com/t/topaz-video-ai-6-0-2-6-0-3/86462/106 "2025-01-26T23:56:49Z")
Just wondering ,is this VFR issue everybody is talking about the same issue Iā€™ve been experiencing since at least 6.0.0 ā€“ basically if i upscale a longer TV show (50+ mins) or a movie, the upscaled video will have seemingly random playback freezes while watching the content. Iā€™ll have to skip past the ā€œbadā€ area by a few minutes, in order to get playback to continue working. Sometimes it happens 3-5 times per movie. But I never get the issue in shorter content upscales (like 10-15 minute cartoon episodes). Basically has made Video AI unusable for me until itā€™s fixed, which maybe it is now in 6.0.3? thread i made a while back with my issue: [Issues with video playback after upscaling movies/TV shows](https://community.topazlabs.com/t/issues-with-video-playback-after-upscaling-movies-tv-shows/85054) [Previous page](https://community.topazlabs.com/t/topaz-video-ai-6-0-2-6-0-3/86462.md?page=4) [Next page](https://community.topazlabs.com/t/topaz-video-ai-6-0-2-6-0-3/86462.md?page=6)