What am I missing (dropped frames)

[lollo]

poisondeathray asked: "If you want farther help, you need to post the original capture", which should be a lossless video prior to any processing.

They have, the AVI snip is in post #23.

I apologize, my mistake

[916Area52]

@poisondeathray...

It has become clear to me as I’m sure it has to you and others here, when it comes to learning this endeavor, that I am trying to run before I can hardly walk. When language/terms like “orphaned field in the source, field 731 in the source, causing vdub's IVTC to place a combed frame, set the AR in the encode” and so on, means very little to me. In trying and understand those terms/ language, I feel I’ve been doing my due diligence searching for and viewing past posts on this great site I know, https://forum.videohelp.com/ and Googling. However, this will take time for me to digest as to what I need to know and learn for my travels.

Now that, that dis claimer is out of the way, … Hopefully your patients will continue?

Processing? As to the best of my knowledge, unless suggested, there has been no processing done by me, “post, on the fly” or otherwise.

Where is this information gleaned from “(orphaned field 731 in the source, IVTC to place a combed frame)”?

I opened up Vdub2, poked around, and what I see according to Vdub2 the scene change happens at frame 365@12:179 to frame366@12:212. I do not know where to look for “(field 731 in the source)”?

I gather that “set the AR in the encode”, is in reference to an external filter that needs to be installed in or made available to Vdub2? If so, I haven’t a clue how this is done, yet.

I do use Lagarith. I was a die-hard huffyuv fan until a fellow traveler suggested I try Lagarith (https://forum.videohelp.com/threads/416452-AmaRecTv3-10-Win10), been using it ever since. As far as I know, as well, Vdub2 did not come with Lagarith. My copy of Lagarith came with the I-O Data GV-USB2. Lagarith then became a codec option for me in Vdub2, but never used until recently. Not sure what “buggy” means? I’ve had no negative experiences with Lagarith, as far as I know.

The attached is/are the Lagarith setting that I’m using.

[poisondeathray]

Where is this information gleaned from “(orphaned field 731 in the source, IVTC to place a combed frame)”?

I opened up Vdub2, poked around, and what I see according to Vdub2 the scene change happens at frame 365@12:179 to frame366@12:212. I do not know where to look for “(field 731 in the source)”?

It might vary slightly on what decoder you are using, the AVI driver (either using official lagarith, or vdub lagarith), or caching input driver. If you use file=>file information it will report what vdub2 is currently using.

The following numbers are with the caching input driver:

In vdub2, with the IVTC filter using the AVI, it would be frame 292. Notice the aliasing artifacts in the fingers, hair, clothing. Compare to the frame before.

If instead of the IVTC filter, if you deinterlace using bob, double frame rate, top field first - Essentially you're examining fields (or scanlines) resampled as a full size frame. The field count is now same as the frame count . The field is 731 . If you go backwards, and examine there will be a 3 repeat : 2 repeat or 3:2 cadence to the fields. That is 3:2 telecine. The cadence is interrupted after 731. There is a single field from a new film frame, that is missing it's "partner" field, because the next frame is the next scene.

I gather that “set the AR in the encode”, is in reference to an external filter that needs to be installed in or made available to Vdub2? If so, I haven’t a clue how this is done, yet.

Yes , if you're using the bundled vdub2 x264 , there is a section for SAR - sample aspect ratio for width and height. There are many threads discussing what SAR means, vs. DAR - I won't repeat it here

I do use Lagarith. I was a die-hard huffyuv fan until a fellow traveler suggested I try Lagarith (https://forum.videohelp.com/threads/416452-AmaRecTv3-10-Win10), been using it ever since. As far as I know, as well, Vdub2 did not come with Lagarith. My copy of Lagarith came with the I-O Data GV-USB2. Lagarith then became a codec option for me in Vdub2, but never used until recently. Not sure what “buggy” means? I’ve had no negative experiences with Lagarith, as far as I know.

The attached is/are the Lagarith setting that I’m using.

Buggy means various problems.

If in the plugins64 folder, you have a Lagarith.vdplugin file, that is the bundled version . The fourcc is "LAGS". The fourcc for the official version is "lags", lower case . You can use mediainfo (view=>text) to show the codec id and fourcc tag. Your AVI file uses "LAGS", so it's using the non official version

One of the problems in your case is returning RGB in some software , some decoders - this causes potential problems including chroma issues, clipping values . Another is sometimes a framecount discrepancy - the lagarith official version is more consistent. Your video has either 2067 or 2065 frames depending on what decoder is used. Another potential issue is lagarith null frames - only the official decoder supports them. For example, lets say you expand your toolkit, learn some avisynth filters using AVISource , or perhaps import into a NLE - these will have the RGB issue with the non official version. If you're not careful in what you do, there are going to be problems

[Alwyn]

If in the plugins64 folder, you have a Lagarith.vdplugin file, that is the bundled version . The fourcc is "LAGS". The fourcc for the official version is "lags", lower case . You can use mediainfo (view=>text) to show the codec id and fourcc tag. Your AVI file uses "LAGS", so it's using the non official version

Mediainfo reports "LAGS" for both VDub Lagarith versions here. AmarecTV only uses the official version (provided it is installed, of course), and Amarec lagarith captures also display as "LAGS" in MediaInfo.

[poisondeathray]

Mediainfo reports "LAGS" for both VDub Lagarith versions here.

You're correct - I mistakenly thought mediainfo could distinguish the fourcc. In the vdub2 compression menu - the fourcc is upper case "LAGS" for vdub lagarith (Lagarith.vdubplugin), and lower case "lags" for official lagarith (lagarith.dll)

AmarecTV only uses the official version (provided it is installed, of course), and Amarec lagarith captures also display as "LAGS" in MediaInfo.

If AmarecTV uses official version, it's still buggy (perhaps for other reasons) - that's why I suspected it was the non official version , because you can see similar buggy behaviour with the bundled lagarith.

For that video sample, eg. lagarith decoder using AVISource returns RGB32 and different frame count 2067 with 29.97fps . ffms2 returns YV16 2065 frames and 29.941 fps.



@916Area52
How was the the AVI "cut" ? Was video=>direct stream copy used ?

[poisondeathray]

In the vdub2 compression menu - the fourcc is upper case "LAGS" for vdub lagarith (Lagarith.vdubplugin), and lower case "lags" for official lagarith (lagarith.dll)

Just a follow up for this - if you examine with other tools like avinaptic2, the fourcc for official lagarith is

FourCC                      lags/LAGS

But the vdub lagarith is

FourCC                      LAGS/LAGS

If you look at a hex editor (vdub/vdub2 has one), there is also the difference "lags" vs "LAGS" (You can see it in the first few rows). Perhaps it was some older version of mediainfo that showed the difference, or maybe my memory is bad

"02FirstNight_NoTbcNoStable_UsbNoSharp_Pana_Vhs.av i" shows up as LAGS/LAGS - which is the non official lagarith version

Anyways the point I brought it up is the "LAGS" vdub2 bundled version has potential for problems. This has been mentioned in several threads, people have various problems with the bundled version

And if amarectv is using official lagarith, there are some other issues going on, assuming that sample was stream copied properly

[916Area52]

Quick reply,


@ poisondeathray…

Lot to take in, for me. Will be doing plenty of research for quite some time.

“In vdub2, with the IVTC filter using the AVI, it would be frame 292. Notice the aliasing artifacts in the fingers, hair, clothing. Compare to the frame before.”

Before you pointed this out, I would not have noticed.

AviSynth is definitely staged to be include in my toolkit. Found several threads on this site that suggested, https://www.youtube.com/watch?v=C4PyyQoz6eo&t=146s . Will be checking it out.

The AVI was cut/trimmed with Avidemux, then that AVI was compressed with Vdub2.

@ Alwyn…

“Mediainfo reports "LAGS" for both VDub Lagarith versions here. AmarecTV only uses the official version (provided it is installed, of course), and Amarec lagarith captures also display as "LAGS" in MediaInfo.”

Appreciate the clarification, Thank you.

Hope I didn’t miss anything/anybody.

Side note…
Just thinking, out loud, of going nuclear on one of my work stations, re-formatting and reinstalling the OS and everything else. Just start from ground zero?

[poisondeathray]

The AVI was cut/trimmed with Avidemux, then that AVI was compressed with Vdub2.

vdub2 can do all those operations - what was the reason for using avidemux ?

Did you mean re-compressed with the non official lagarith to make the "cut" AVI ? If so, that could explain the foucc discrepancy and perhaps some of the other problems .

"02FirstNight_NoTbcNoStable_UsbNoSharp_Pana_Vhs.av i" shows up as LAGS/LAGS - which is the non official lagarith version

In theory , a lossless compressed video that is decoded to uncompressed and recompressed with lossless compression should be 100% identical when decoded to uncompressed frames again (that's reason for using "lossless" video compression in the first place). But that is not the case if the handling has problems, such as timestamp issues, color space conversions

There is a mode in vdub , video=>direct stream copy (and audio=>direct stream copy) , that avoids recompression entirely. You can mark in/mark out and make edits that way. An I-frame format such as lagarith can be cut anywhere (as opposed to a long GOP format, which has restrictions)

Just thinking, out loud, of going nuclear on one of my work stations, re-formatting and reinstalling the OS and everything else. Just start from ground zero?

I wouldn't, unless you had a good reason

[916Area52]

“vdub2 can do all those operations - what was the reason for using avidemux ?”

A fellow traveler suggested AvideMux after I mentioned that I used VLC to cut/trim my AmaRecTv captures :-O.

I Have to slow down and regroup… This didn't happen, yet... "The attached was done w/Vd2, from the full AmaRecTv capture 02FirstNight_NoTbcNoStable_UsbNoSharp_Pana_Vhs.avi "

Side note…
Went ahead with the “nuclear” project. Write to file error (00000087) kept rearing its head. Will see if this solves it.

[916Area52]

@ poisondeathray...

Following up from post #39...
"The attached was done w/Vd2, from the full AmaRecTv capture 02FirstNight_NoTbcNoStable_UsbNoSharp_Pana_Vhs.avi "
Thanks to you, and everyone involved for the direction and patients. I feel the attached, although not prefect, are a step in the right direction? Please review and comment to the attached.

[poisondeathray]

"0003FirstNight_vd2FullPro-lagdll-Snip_Vhs.avi" is better in terms of consistency, frame rate, frame count - "lags/LAGS" - which is what you would expect for official lagarith, and opens up correctly with AVISource as YUY2 instead of RGB32.

1st field in AVI is an orphaned field (AVI sample was not cut on 3:2 telecine boundaries, so cadence is interrupted - similar to how a bad edits can occur for TV broadcasts) , and dropped as a frame in vdub's IVTC . Other algorithms (e.g. TIVTC in avisynth) deinterlace that field and keep that as a frame. The MP4 has 3 repeated frames tacked on the end and 1708 frames . If you do the math 0.8 * 2132 frames = 1705.6 . (0.8 because (24000/1001) / (30000/1001) ) . Don't put much "weight" into analyzing the ends of a cut sample - they are not good for analyzing for post processing, because different algorithms might handle them differently, depending on how the sample was cut. But if issues occur on full scenes - then you can say it's a problem

Not sure if you recaptured it, or if the issue was from the avidemux workflow or the buggy lagarith version - but the orphaned field when they are lying in the grass is no longer "orphaned" in the new AVI ; the "partner" is now present. However, the vdub IVTC MP4 still has an aliased frame on 340 for some reason - it's a bad field match because all the correct fields are available, and it's 3:2 steady in that section . TIVTC matches it correctly if you assume the correct field order TFF (and no post processing is applied if you used display=true or pp=0; it's just "regular" field matching). If you assume TFF (or auto or BFF) for vdub IVTC, it's still combed and a bad match. TIVTC has more advanced options should you need them , and what I would use for general purpose IVTC

[Alwyn]

The MP4 has been coded 3:2; it should be 4:3, methinks.

[916Area52]

@poisondeathray...

Not wanting too add any variables, the source avi has remained the same throughout.

Thank you for the analysis. It is clear that I need to learn avisynth, if I want to improve on what I have learned here, and I do .

[916Area52]

@Alwyn...

Normally, I do use the "crop" filter to take a little off the sides and bottom. Not in this case, didn't want to add any variables.

[poisondeathray]

Not wanting too add any variables, the source avi has remained the same throughout.

The full source AVI? So either the prior procedure(s) with avidemux, or vdub2, or re-encoding with non-official lagarith version caused some of those problems and differences between the AVI samples

Normally, I do use the "crop" filter to take a little off the sides and bottom. Not in this case, didn't want to add any variables.

For your final workflow, if you crop it, you still would need to either resize it to square pixel equivalents; or adjust the SAR. Otherwise the display AR is incorrect - eg. circles become ovals; people look to fat or too thin

Setting the SAR does not affect any of the actual encoded pixels or values - it just affects how a player adjusts the playback (it's "stretched" or "squished" to the proper AR on playback)

[916Area52]

@poisondeathray...

I must be missing something. The attached are the settings used for the accompanying video files.
Won't be doing any cropping anytime soon, I will have to put up with the audio noise until I get more tools in my tool kit,

[poisondeathray]

There is a duplicated frame 292-293 in the AVI. That causes 01Oceans11-1960_lagdll_SnipIVTC_Vhs.mp4 to have a duplicate 233-234 post vdub IVTC

The MP4 is missing the SAR setting, and if it was for a end distribution format, normally you would use 4:2:0

I don't use AmaRecTv , maybe others can help with those settings

[916Area52]

Thank you poisondeathray, and all the others that have takin the time to help me see better. I've learned a great deal and will continue to add tools in my tool kit.

Roy

[Alwyn]

Re the top two images, if those are your Save As settings for the snip AVI, you should just use Video>Direct Stream Copy. Set up a short range (30sec) with the left and right arrow-buttons on the timeline and then do a "Save Video". You'll then get a exact copy of the original AVI. If I'm reading PDR's post correctly, you should set up the snip start frame to not be over the frames showing the interlacing.

For the H264 exports, I use AAC as IIRC I read somewhere it is more efficient that MP3 ie better quality for the same bitrate.

For AmarecTV audio, I use GV-USB2 "Analogue Capture". Not sure why, but it works OK.

[Alwyn]

@PDR, what are you using to detect the dupes? I couldn't find much for AVISynth but I did try DupStep but got OpenCL error messages.

[poisondeathray]

@PDR, what are you using to detect the dupes? I couldn't find much for AVISynth but I did try DupStep but got OpenCL error messages.

For duplicate detection using scripts - I use duplicity2. Jagabo posted one earlier in post #3

For lossless encodes, script detection is usually very accurate for a given threshold - because duplicates are usually true duplicates in that case. But for lossy encodes, compression artifacts and noise can make duplicates detected as "different" by the algorithms

[916Area52]

@Alwyn…

“Re the top two images, if those are your Save As settings for the snip AVI, you should just use Video>Direct Stream Copy. Set up a short range (30sec) with the left and right arrow-buttons on the timeline and then do a "Save Video". You'll then get a exact copy of the original AVI. If I'm reading PDR's post correctly, you should set up the snip start frame to not be over the frames showing the interlacing.”

Direct stream makes perfect sense, now. Though, I’ve tried that before, and VLC wouldn’t play it. Must have had some other setting wrong, VLC has no issues now.

The attached 01 avi and mp4 direct stream captures are approx. 14sec before, and though, of what vdub2 indicates as a dropped frame, to 14sec after (28sec in total +/-).

The attached 02 avi and mp4 direct stream captures are approx. 24sec after what vdub indicates as a dropped frame.

After the information that has been shared it this thread along with the addition of the pana es-15, my dropped/inserted issues have decreased significantly. The captures now are at, maybe +5 max, and 3 to 4 of those are at the beginning of the capture (black screen).

“For the H264 exports, I use AAC as IIRC I read somewhere it is more efficient that MP3 ie better quality for the same bitrate.

For AmarecTV audio, I use GV-USB2 "Analogue Capture". Not sure why, but it works OK.”

Thank you for the heads up. Also, how do you “save as” (please see attached)?

[poisondeathray]

If I'm reading PDR's post correctly, you should set up the snip start frame to not be over the frames showing the interlacing.”

When you edit something that still has 3:2 pulldown (ie. "23.976p in 59.94i") - you need to cut on 3:2 boundaries. If you go frame by frame during a scene with motion, you will see 3 clean, 2 combed, 3 clean, 2 combed... Cut on either before or after the 3 clean (ie. between the 3 or 2) - otherwise you interrupt the cadence (aka - a "bad" telecine edit) .

The ideal way to edit film material is to remove pulldown (aka IVTC) first , so you edit progressive film frames. And if you still need to, apply pulldown afterwards.

[Alwyn]

how do you “save as” (please see attached)?

Yes, I use MP4 Fast Start for MP4s.

For duplicate detection using scripts - I use duplicity2. Jagabo posted one earlier in post #3

Thanks. I then got to thinking that even if I find some dupes, there's not much I can do about them, as removing them would upset the audio sync. AmarecTV will give a frame by frame readout of any dupes in the capture report so I might just stick with that.

[poisondeathray]

I then got to thinking that even if I find some dupes, there's not much I can do about them, as removing them would upset the audio sync. AmarecTV will give a frame by frame readout of any dupes in the capture report so I might just stick with that.

Does it list the accompanying drop? The general rule is if you have dupes, you also have the same number of drops somewhere - otherwise it would be out of sync.

Best case scenario would be the drop is the same location as the duplicate. Usually the drop precedes the duplicate, but sometimes they are displaced a few frames. The duplicate is the "catchup" mechanism to keep sync - the most common reason is because hardware resources are insufficient.

If re-capturing is not an option, then post capture option is you delete the dupe and insert the missing frame in the correct location by using an interpolated frame. Depending on the complexity of the situation sometimes it works "automatically" . If the dupe is the drop, this is easy - there are filldrops variations, some have been modified to use RIFE.

For displaced dupes and drops - John meyer has a script for this purpose, and another is smoothskip. Basically they look for the biggest "jump" in motion in a set of frames and insert the frame there. The problem with those methods is you need an approximate cycle estimation - if drops are very irregular, they can have problems.

[Alwyn]

Does it list the accompanying drop?

A good question... I don't know. I'll have to defer to Lollo here.

[lollo]

Does it list the accompanying drop?

A good question... I don't know. I'll have to defer to Lollo here.

I tried to answer here: https://forum.videohelp.com/threads/416657-Analog-Capture-inserted-and-dropped-frames

[916Area52]

@ Everyone who participated in this thread...

There is an abundance of knowledge that has been shared here. Allow me to bow out of this thread to do further research and acquire more tools for my tool kit.

Roy

[aramkolt]

If only we had a tool to locate time gaps, odds motions.

I still think a timecode generator (which overlays a unique timecode on each Frame as it passes through- kind of like a VCR menu overlay) and also makes timecode audio would show if frames are being dropped or inserted. This assumes that the timecode generator itself doesn't get confused by timebase errors to begin with. An inserted frame would show as the same timecode for multiple frames in a row visually and dropped frames would show as the overlayed timecode getting ahead of the LTC audio track generated by the timecode generator. Haven't tried it personally though.

Not that you'd want to use it during real captures, but could be useful for troubleshooting what's actually happening for test recordings.

[jagabo]

I've done exactly that many times. Usually with frame numbers so it's easy to see when the frame number in the editor doesn't match the frame number in the video are mismatched. That makes it much easier to find many of the problems.