New tool - VapourBox

[tvd051955]

I found the fix!!!!

So simple:

# IVTC: Inverse Telecine (VFM field matching + VDecimate)
#clip = core.std.AssumeFPS(clip, fpsnum=30000, fpsden=1001)
clip = core.std.AssumeFPS(clip, fpsnum=24000, fpsden=1001)
#clip = core.vivtc.VFM(clip, order=1, cthresh=10)
#clip = core.vivtc.VDecimate(clip)

The template:

{{#DEINT_IVTC}}
# IVTC: Inverse Telecine (VFM field matching + VDecimate)
clip = core.std.AssumeFPS(clip, fpsnum=24000, fpsden=1001)
{{/DEINT_IVTC}}

Vern

[tvd051955]

Just processed 8 episodes from a series using the above IVTC settings. I used the first episode to compare NVenc and SW encoding. The NVEnc encode actually created a larger output MKV file and looked great.

For the first time I viewed some of them in my HT (where the audio delay settings on the receiver were calibrated using an oscilloscope and a light sensing probe and an audio sync calibration disc).

I don't know if anyone else is seeing audio sync issues, so here are my observations:

While audio delay for the deinterlace processing looks to be correct (at least within 20 ms) the IVTC process delay processing looks pretty far off.

To correct for proper audio delay on the IVTC episodes I had to reduce the delay from 240ms (the calibrated value in my HT for 24 FPS sources) by 220 ms (20ms on the receiver) when using nvenc, and using the software encoder, by -160ms (80ms on the receiver) to get acceptable audio sync.

[StuartCameronCode]

Oh wow, you've been down a rabbit hole, I hadn't been getting any notifications you'd been posting, not that I'd have been able to do much about it as I've had a busy week. I'll have a look into all this probably a bit later today. If you've got a short non-copyrighted mkv you could upload that I can use as a test case that'd be a huge help, I'm in Australia and I don't have many NTSC DVDs lying around or unadulterated rips.

[tvd051955]

I uploaded a test test clip from a NTSC DVD.

[StuartCameronCode]

Thanks! I'm working on a new "soft telecine" input filter as a third "deinterlacing" option on top of QTGMC & IVTC that uses the source frame rate correction you flagged. I'm still not sure about the source of the audio desync, but having a tinker and now I've got a good source to test against, thanks again! The trickiest bit is going to be making all this make sense to end users not completely familiar with everything.

[tvd051955]

I've got a suggestion on the handling of the presets:

When a preset currently exists, could you please add an option to update it?

Since the description shown on the list comes from an internal field in the preset file, rather than the actual physical file name, when I update the settings and then try to update the current preset by reentering the existing name, I wind up with multiple presets being saved with the same name, but containing different settings.

BTW, can you please concentrate of the existing stuff first?
Thanks

[tvd051955]

BTW, your % and ETA calculations are incorrect when processing IVTC data - 100% and -- with the frame #'s still incrementing during the last few minutes of the run.

[StuartCameronCode]

There's a new 0.7.0 release out, HOPEFULLY it addresses everyone's issues. It now supports soft telecine, hard telecine, improved auto-detection of input video format with an auto-switch to the best guess "deinterlacing" solution. There's improved preset support, and a bunch of smaller quality of life fixes.

https://github.com/StuartCameronCode/VapourBox

[StuartCameronCode]

Whilst I'm on a roll, version 0.8.0 is now out! This fixes some aspect ratio issues, improves scan type auto-detection and automatic selection of (hopefully) an appropriate deinterlacing method for the source video.

Oh, and one more thing - automatic subtitle generation! You can now tick the "subtitles" option to auto-generate subtitles for your video, either in .srt format or embedding them directly into the source video. Use can use this as a standalone tool by unticking all video filters and it'll generate subtitles for any existing video without any additional processing. You'll be prompted for permission to download the Whisper speech to text engine, but otherwise it'll "just work". And don't worry, it all runs on-device, NOTHING is sent to the cloud.

https://github.com/StuartCameronCode/VapourBox

Enjoy, and tell out if you encounter any issues.

[aramkolt]

0.8 didn't fix the semi-opaque "no preview" on the left of the slider and no thumbnails on MacOS 26 Tahoe. Could possibly be some conflict from having tried to install this via terminal or other installs of *synth stuff maybe. I suppose only way to really know would be to take a fresh install of MacOS or on a completely different machine and see if that has the same problems or not.

Anyone else tried this on their MacOS 26 machine and seen the same issue, or is it just me?

[StuartCameronCode]

0.8 didn't fix the semi-opaque "no preview" on the left of the slider and no thumbnails on MacOS 26 Tahoe. Could possibly be some conflict from having tried to install this via terminal or other installs of *synth stuff maybe. I suppose only way to really know would be to take a fresh install of MacOS or on a completely different machine and see if that has the same problems or not.

Anyone else tried this on their MacOS 26 machine and seen the same issue, or is it just me?

As far as I can tell the problem is specific to your machine, but I'd love to know why, it's doing my head in. I tried a complete new install of MacOS 26 into a VM specifically to test this and did nothing but install VapourBox on top of the clean install and it worked just fine. The tool explicitly tries to use the versions of all the bundled tools rather than looking in your path, all I can think of is that some dependent library used by one of those tools might be causing breakage or something.

[StuartCameronCode]

There's a new 0.9.0 release out! There's some fairly major changes this time, here's a few:

- Major processing pipeline switch from BestSource to ffmpeg piping, hopefully resolves some A/V desync issues, also no more waiting for indexing on import
- Direct DVD import, either from physical media or VIDEO_TS folder
- MANY bug fixes
- Lots of QTGMC (and other filter) parameter fixes and cleanups

Check out the release notes for more details.

[tvd051955]

Is there a reason you are not saving the resize kernel in the presets? It seems to always defaults to bicubic.

[StuartCameronCode]

Is there a reason you are not saving the resize kernel in the presets? It seems to always defaults to bicubic.

It looks like a bug, I've just fixed a bunch like this in the QTGMC filter. I haven't actually used the resize filter much myself just yet, I've been mostly focused on home VHS restorations. Thanks for letting me know.

[StuartCameronCode]

For anyone still following this thread, I released 0.9.1 a few days ago with a fix for the above issue amongst others.

[Barrythecrab]

I am really liking where this is going. It's an easy UI like Avidemux, but stays in 4:2:2 (I hope), and not as scary as my go-to Hybrid. What is up next for tweaks and filters?

[StuartCameronCode]

I am really liking where this is going. It's an easy UI like Avidemux, but stays in 4:2:2 (I hope), and not as scary as my go-to Hybrid. What is up next for tweaks and filters?

Honestly, I'm very open to suggestions. If you've got any favourite VapourSynth plugins (or anything else you'd like to see integrated) I'm VERY open to suggestions.

It's just a hobby project for me, something I built to make my own family VHS restorations a bit easier. Although Hybrid is great, it's daunting - it's not intuitive for people wanting something they can throw a video into, pick a few filters and hit "go". Plus the lack of a simple preview means iterating on settings is a bit of an exercise. So I figured I'd add something new into the mix.

[_Al_]

Honestly, I'm very open to suggestions.

AI ?
Focusing on VHS or camcorder DV stuff. Real life footages. There is plenty of anime restoration ai and vapoursynth users had found their way to handle it anyway. But real folk stuff, recordings, no movies. If there are models for those if requested, because those files are larger. If you had downloaded models, users just see footage before and after.
HolyWu is the guy to follow: https://github.com/HolyWu . His RIFE plugin for double frame rate from single frame rate (like 30p to 60p) you might start with to test things, because you already use QTGMC. It would test your app for sure, because responses for drawings would be longer etc.

Major challenge would by pytorch and dependencies, what versions work with what vapoursynth and python version, and how to make it portable, but Selur figured it all out if I am not mistaken.

[tvd051955]

I probably wouldn't concentrate on AI. Every AI tool that I've tried takes forever to run with results that just flat out look unnatural.

As far as movies go, in addition to 16mm / 8mm / S8mm home movies, there are a lot of titles out there that only exist today in interlaced or IVTC SD and will likely never be released in HD.

[Barrythecrab]

No no no AI!

[StuartCameronCode]

I don't have an issue adding support for AI filters if people want to use them - the models would be optional additional downloads anyway and would run locally. But each to their own, no-one's going to force you to download or use something you object to. I think there's plenty of scope for AI tools where they've been trained on open datasets in an ethical manner with the consent of the original authors, and used in a way that supports human creators in their tasks rather than trying to replace them, but this isn't the place to be starting holy wars. I'm not building an AI tool here or doing any model training, it's just a UI to simplify workflows.

I've had a play with a couple of open source video enhancement models outside of VapourBox just out of curiosity and have also found them to be impossibly slow to run compared to tools like Topaz Video etc, which are also slow but don't take a full month to process. If there's a specific model or two you want incorporated because they're models you actively use then I can add them to the TODO list, but I don't want to add a bunch of model filters that aren't useful to anyone if they're just going to be too slow or impractical.

Incorporating Whisper for the sake of adding automatic subtitle generation was pretty straightforward. I added that because I wanted my large collection of home videos to be easily searchable for old memories. Again, it's an optional download if people want to use it. It's an individual choice.

AI aside, if there's specific other filters anyone wants incorporated point me to them and I'll have a go.

[Tom Saurus]

I am rather interesting in this program. We still buy the occasional DVD Box Set and for example I bought the Cannon The Complete Series and Barnaby Jones The Complete Series and I used MakeMKV and then used Vidcoder for deinterlacing and doubling the frame rate so that the videos would play smooth without that window blind and jarring video stuff happening. Would this tool be a better way of doing things? I usually recode the audio along with the video as I have run across lip syn issues if I don't do that. Thank you StuartCameronCOD for this tool and I want to give it a try soon.

[aramkolt]

Main filters I think would be useful to a lot of people would be a "live preview" way to change the audio delay and a way to adjust Y/C delay, meaning a way to move chroma horizontally by a few pixels, but also in some cases a way to move chroma vertically with a live "zoomed way in" preview to know the correct value is being chosen before the whole file gets processed.

[Barrythecrab]

The ability to realign the chroma would be terrific, totally agree. I am so impressed that "Mister Cod" just sat himself down and put this program together, then responded to issues and now here we all are with a new tool. Furthermore I best not complain very much because Cods eat crabs...

[StuartCameronCode]

I am rather interesting in this program. We still buy the occasional DVD Box Set and for example I bought the Cannon The Complete Series and Barnaby Jones The Complete Series and I used MakeMKV and then used Vidcoder for deinterlacing and doubling the frame rate so that the videos would play smooth without that window blind and jarring video stuff happening. Would this tool be a better way of doing things? I usually recode the audio along with the video as I have run across lip syn issues if I don't do that. Thank you StuartCameronCOD for this tool and I want to give it a try soon.

If you're ripping interlaced video then the built-in QTGMC deinterlacer will let you double the frame rate natively in the tool. Give it a go, and yell out if you run into sync issues, but from my end it's been working well. I've been deinterlacing VHS tapes to 50fps PAL and the audio sync's been working fine.

[StuartCameronCode]

Main filters I think would be useful to a lot of people would be a "live preview" way to change the audio delay and a way to adjust Y/C delay, meaning a way to move chroma horizontally by a few pixels, but also in some cases a way to move chroma vertically with a live "zoomed way in" preview to know the correct value is being chosen before the whole file gets processed.

The live previews are zoomable, so that's already there. The previews are just still frames, so it'll be hard to preview A/V sync. I could possibly add a "Preview segment" generator or something though if that'd be useful.

[StuartCameronCode]

The ability to realign the chroma would be terrific, totally agree. I am so impressed that "Mister Cod" just sat himself down and put this program together, then responded to issues and now here we all are with a new tool. Furthermore I best not complain very much because Cods eat crabs...

lol, it truncated my username from StuartCameronCode, which I picked to match my GitHub account name, which I went with because StuartCameron was already taken. Guess I'm a fish now. So be it. I've been called worse.

[Barrythecrab]

And the more I look the more I see, such as the ability to move chroma. Dawn lights over Marblehead.

[Barrythecrab]

OK, how about something like Despot, or Spotless, or Descratch?

[Tom Saurus]

I am rather interesting in this program. We still buy the occasional DVD Box Set and for example I bought the Cannon The Complete Series and Barnaby Jones The Complete Series and I used MakeMKV and then used Vidcoder for deinterlacing and doubling the frame rate so that the videos would play smooth without that window blind and jarring video stuff happening. Would this tool be a better way of doing things? I usually recode the audio along with the video as I have run across lip syn issues if I don't do that. Thank you StuartCameronCOD for this tool and I want to give it a try soon.

If you're ripping interlaced video then the built-in QTGMC deinterlacer will let you double the frame rate natively in the tool. Give it a go, and yell out if you run into sync issues, but from my end it's been working well. I've been deinterlacing VHS tapes to 50fps PAL and the audio sync's been working fine.

Thank you for this information. I will give it a try soon. I won't redo Cannon and Barnaby Jones as I am happy with what Vidcoder did and there are just so many episodes. Thankfully I converted all my VHS that I needed to get done and it is good enough. But I think I might redo some of my other DVD Box Sets.

For me if I was to pass through the audio and just recode the video that is when the lip sync issues would come up via my Vidcoder encode so I just set it to recode the audio as well and that seem to stop the lip sync issues.