Looping can be enabled by setting an In/Out point on the timeline where you’d like playback to repeat
1 Like
The developers seemed to ask for better models, well, Flowframes processes faster than Aion, and in some places the quality is much better.
1 Like
Soon the free models will be left idle, you haven’t come up with anything new after buying Topaza, except for the torment of the intrerface. Сорян за английский))
How do I turn off real-time rendering?
We released four new models for desktop processing in 2024, and we expect to continue delivering progress on local models
1 Like
Real-time rendering will only run when a video is playing and has filter settings applied
I don’t want it to render until I press the “render” button. How do I do that? I jumped ahead a bunch of versions, but the one I was on could do that.
I’m trying to look through a video. I would like to watch a few second, find a spot to test, then press “render” and have it render. Real-time rendering needlessly slows that process down.
Play the “Original” tab in single view mode, and then switch to a tab with your settings to view any rendered section.
That’s workable. Thanks. Feels like extra clicks though? I test a spot, click to the “project”, render, click back to “original” find another spot, repeat…
Yes with the 4.0.9 release we get
- Proteus v4
- Iris LQ v2
- Nyx Fast v1
that was really good! …but since then nothing happens except RheaXL, was already criticized in the beta because it produced so many artifacts and almost nobody uses it (here I am almost an exception)
So the benefit since 4.0.9 is RheaXL, Hyperion, the new UI (exist someone who likes it?) and mplayer (as the problems startet…)
I don’t just want to criticize, free access to starlight testing is a nice move, but unfortunately that doesn’t change the fact many could ask why they should upgrade, certainly not because of the new UI and not because of RheaXL either. So personally I think I a real incentive would be a fixed RheaXL or a Iris MQ V2 or merge both into one new model that can replace them.
4 Likes
The last stable version 5.3.6, then buggy slag, there is no point in installing new versions.
3 Likes
2 things now that I’m trying something other than 5.3.6… This new interface. Keep it Simple.
Man, I can’t believe 5.3.6 had more information available without having to open any new interface. I could see the job and when it was done I could see the little dots moving as it moved the file. Yesterday I decided to give the new version a try since I read a studddddder bug had been fixed… Last night I did a movie that took 4 hours with 2 enhancements. I caught the las 30 seconds of the file being processed… the countdown and then it was done. There was nothing telling me (that i could find) that it was moving a file… Since the que was moved to it’s own little sub panel nothing was evident. I brought up the panel and it still didn’t seem to show the stage in the process… I clicked around trying to find something that would clarify it’s status and it froze. HOURS of processing and it froze. After a little whle the Windows popup saying the app was unresponsive and if I wanted to stop it came up… I ignored it hoping tvai was just really busy. I locked the compute and walked away frustrated. I came back before I closed out for the evening and saw the file had been processed and save although the cache file it generates in the original file location was still there and the whole app was still frozen (I had to kill the app off). FRUSTRATING. Please make it go back to how it was. You added more to the interface… more complicated just for the sake of making seem more… Advanced? Techy? It’s just more bloated for no reason. My pc isn’t dedicated to just processing video so it’s much easier the old way having all the info there in the default screen. Why hide it behind unecessary tabs and popus?
2. Why does text get mangled so? It appears that text ends up with like a negative dark halo the more detail you bring up in Proteus… You can find the sweet spot for the video file where it produces awesome results… but text… wow… It gets the jackhammer treatment every time.
Love the app otherwise… Good work!
1 Like
The text problem goes all the way back to the very beginning, with version 1. From what was discussed a very long time ago, the models can’t be trained on copyrighted material, where much of the text comes from, like show titles and credits. There really needs to be a model, which is trained specifically on text, but without violating copyrights.
Just tried this version. Still the same problem with the constant FPS rate still not respected.
SOURCE mediainfo Metadata:
Général
Complete name : F:\_SUPER_8\_1_Convert_Cineform\02_SIM_P2\V2\Movie00004.mov
Format : MPEG-4
Format profile : QuickTime
Codec ID : qt 0000.02 (qt )
File size : 2,86 Gio
Duration : 6 min 43 s
Overall bit rate mode : Variable
Overall bit rate : 60,8 Mb/s
Frame rate : 18,000 im/s
Writing application : Lavf57.79.100
Vidéo
ID : 1
Format : CineForm
Codec ID : CFHD
Codec ID/Info : CineForm High-Definition (HD) wavelet codec
Duration : 6 min 43 s
Bit rate mode : Variable
Bit rate : 60,8 Mb/s
Width : 1 728 pixels
Height : 1 296 pixels
Display aspect ratio : 4/3
Frame rate mode : Débit im/s constant
Frame rate : 18,000 im/s
Color space : YUV
Chroma subsampling : 4:2:2
Bit depth : 10 bits
Scan type : Progressif
Bits/(Pixel*Frame) : 1.509
Stream size : 2,86 Gio (100%)
Language : EnglishTARGET mediainfo Metadata:
Général
Complete name : H:\__Capture_KODAK\_2_Convert_Cineform_TPZ\02_SIM_P2\V2\Movie00004_2_stab_chf3_iris3.mov
Format : MPEG-4
Format profile : QuickTime
Codec ID : qt 0000.02 (qt )
File size : 11,7 Gio
Duration : 6 min 43 s
Overall bit rate mode : Variable
Overall bit rate : 249 Mb/s
Frame rate : 25,000 im/s
minor_version : 512
major_brand : qt
compatible_brands : qt
videoai : Stabilized auto-crop with smoothness 8. Framerate changed to 25 using chf-3 replacing duplicate frames. Enhanced using iris-3; mode: relative to auto; revert compression at 35; recover details at 0; sharpen at 0; reduce noise at 35; dehalo at 0; anti-alias/deblur at 0; focus fix Off; and recover original detail at 20. Changed resolution to 1728x1296
encoder : Lavf61.7.100
Vidéo
ID : 1
Format : ProRes
Format version : Version 0
Format profile : 422 HQ
Codec ID : apch
Duration : 6 min 43 s
Bit rate mode : Variable
Bit rate : 249 Mb/s
Width : 1 728 pixels
Height : 1 296 pixels
Display aspect ratio : 4/3
Frame rate mode : Débit im/s **variable** <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
Frame rate : 25,000 im/s
Minimum frame rate : 12,500 im/s <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
Maximum frame rate : 12 800,000 im/sv<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
Color space : YUV
Chroma subsampling : 4:2:2
Scan type : Progressif
Bits/(Pixel*Frame) : 4.447
Stream size : 11,7 Gio (100%)
Writing library : Apple
Language : EnglishI went back to the latest version that works well: 5.3.6.
My subscription to Video (since 2021) and Photo AI will not be renewed until this problem is resolved.
Distressing!
7 Likes
REALLY? You mean they couldn’t drive down a street in las vegas and record some video and run it through with ai?
Or go on Fiver and have someone generate some video of signs… lots of text on signs or some with just a couple of signs… mix it up
Hell even going up into the mountains somewhere and putting up posters with text amongst the trees. Spend the weekend making said videos and train the hell out of that generative a.i on it.
I don’t get it lol…
3 Likes
One question, do the templates always have to be downloaded for each new version, I generally use the ones downloaded from the previous version in other location on the HDD.
You really need to ask Topaz that question, because I’m just telling you what was already discussed years ago, when people complained about the text quality.
It was rhetorical lol 
Thanks for the test. Then my eyes didn’t deceive me when my encode suddenly had vfr (was V6.03), they simply didn’t fix it! I also I can get over the bad ui design and some bugs can be worked around, but not the VFR thing, that also kills every new release, tragic they can’t (won’t?) solve that finally.
1 Like
