Current status of ld-decode / vhs-decode (true "backup" of RF signals)

[Joscraft_05]

... or try to search another Panasonic model with 12bit ADCs from 2006 or 2007 with an HDMI output and a recommended HDMI splitter by Skiller to upgrade my workflow to a lossless HDMI capture,

I don't mean to derail the topic, but there are alternatives to computer and HDMI splitters, Get a HDMI/SDI recorder, You will be surprised how stable the system is, not to mention how compact is and you will be able to free up your computer for other tasks such processing and encoding, I'm computer-less for capturing for almost two years now, never going back.

DVCPRO recorders tend to be expensive and complex to repair, Some of them can output lossless SDI, the same goes for Betacam recorders or any pro format for that matters, The benefits of SDI over HDMI is that it is standard compliant, The levels are correct, The color space is exact and the audio is embedded too just like HDMI despite using two wires only.

Oh, you mean those HDMI recorders like the Blackmagic Video Assist with a 5" screen? Not gonna lie is the first time i hear that you don’t get HDCP issues with those recorders and thus not needing an splitter, that’s interesting… I like the idea and that does help me for doing captures and in the computer working with the deinterlace, slight Denoise, upscale with nnedi3_rpow2 and that, this is indeed a very good way for optimize the workflow.

The Blackmagic video assist is expensive though, at that price I’d rather wait and get a MISRC device assembled, but i hope vhs-decode especially for noisy tapes and Video 8/Hi8 improved a lot more in the future and we finally don’t get those ringing/hallows.

[dellsam34]

You only need a recorder with built in screen if you don't want an extra monitor for proc-amp adjustment in the capture device, otherwise this recorder should be perfect for the job, You can add a small HDMI or SDI monitor connected to the recorder's monitor port for that purpose. I personally share the computer monitor in a split screen mode as it does have dual HDMI input, So one half is for the PC and the other half is for proc-amp adjustment or monitoring the capture if needed to, Still completely independent from the PC.

[Joscraft_05]

You only need a recorder with built in screen if you don't want an extra monitor for proc-amp adjustment in the capture device, otherwise this recorder should be perfect for the job, You can add a small HDMI or SDI monitor connected to the recorder's monitor port for that purpose. I personally share the computer monitor in a split screen mode as it does have dual HDMI input, So one half is for the PC and the other half is for proc-amp adjustment or monitoring the capture if needed to, Still completely independent from the PC.

Ooh those devices, these are a LOT less expensive and i already have 2 PC monitors so i could use 1 for this.

To be honest i didn’t knew you didn’t had any HDCP issues with these recorders, and given the advantages in the optimized workflow freeing up the PC from the captures and having another device doing them is honestly very good.

[lollo]

and given the advantages in the optimized workflow freeing up the PC from the captures and having another device doing them is honestly very good.

But consider that you lose control over the capture software’s operations, including frame insertion/dropping and detailed per‑frame reporting as provided by AmarecTV.

[Joscraft_05]

and given the advantages in the optimized workflow freeing up the PC from the captures and having another device doing them is honestly very good.

But consider that you lose control over the capture software’s operations, including frame insertion/dropping and detailed per‑frame reporting as provided by AmarecTV.

Yes, i'm aware of that. To be honest when i get the money probably I’ll end up with a splitter and a capture card since these are a lot more common and some Blackmagic and StarTech devices are too cheap in my country (20-30€) and the splitter recommended by Skiller in AliExpress for 4€, but the workflow of dellsam34 is really good and optimized.

[The 14th Doctor]

Alright, got Nicholas Serra's permission to upload a snippet on the comparison he made with one of his own Hi8 tapes. He told me to say that when showing this off on the discord users chimed in saying that Video8/Hi8 video quality is already great as is when captured conventionally and properly, so decode ain't better from a quality standpoint despite the opposite claims being promoted around..

[Joscraft_05]

Alright, got Nicholas Serra's permission to upload a snippet on the comparison he made with one of his own Hi8 tapes. He told me to say that when showing this off on the discord users chimed in saying that Video8/Hi8 video quality is already great as is when captured conventionally and properly, so decode ain't better from a quality standpoint despite the opposite claims being promoted around..

Thanks for sharing! I'll deinterlace those videos and check the results with both videos encoded with x264 medium preset at 50Mb/s 4:2:2 10-bit with QTGMC, MC Temporal Denoise at the very low preset, upscale with nnedi3_rpow2 and Spline144 and the contrast adaptive CAS sharpener at 0,050 after the upscale.

Video 8 but especially Hi8 video formats were the best consumer analog formats made for the market in the past and the recordings made or captures made especially with the late 90's Hi8 XR cameras (fully working, no bad caps) with EDIT mode and DNR turned off they can look already super good and stable. My only downside is that the TBC of the Sony handycams to me touch for worse the brightness (but only slightly) and the colors making them more muted and desaturated, that’s why i moved onto using my DMR-ES10 i got for 10€.

They were really good formats but still worse than Betacam because the companies only reserved the quality of Betacam for broadcast pro use until DV video came for everyone in the mid 90's.

[Joscraft_05]

Alright, got Nicholas Serra's permission to upload a snippet on the comparison he made with one of his own Hi8 tapes. He told me to say that when showing this off on the discord users chimed in saying that Video8/Hi8 video quality is already great as is when captured conventionally and properly, so decode ain't better from a quality standpoint despite the opposite claims being promoted around..

Okay i saw the results and well, i like that he got the fm rf capture with vhs-decode look without noise at all but i don't know if the image is a bit softer and less sharp and detailed or the noise from the "conventional" capture is tricking me.

Also the capture with vhs-decode looks darker.

Decode:

[Attachment 93427 - Click to enlarge]

Conventional:

[Attachment 93428 - Click to enlarge]

Decode:

[Attachment 93429 - Click to enlarge]

Conventional:

[Attachment 93430 - Click to enlarge]

The aspect ratios are also different i don't know if it was the captures or how i did it in Hybrid, in the program i choose "convert output to PAR: Square Pixel 1:1" to carefully upscale my videos with the correct aspect ratio and resolution.

Judging by these videos from Nicholas Serra yes, i don't like the capture done with vhs-decode here although the noise is a lot less present, curious, another examples shows more noise for vhs-decode and others less.

[nicholasserra]

Just wanna chime in that i'm incredibly impressed by the results with the decode HI8 captures.

These were done with Sony CCD cam > modded CX card > ubuntu. Signal was in the 50s. Basically was trying to get "as good as conventional" results, so I set the levels to get it looking as close to the conventional capture as I could, and i'd say its 90% there.

Still weighing the pros and cons of switching over to decode for all 8mm captures.

On one hand, the process is longer. Conventional I can just play and capture, and my resulting archive file is done and I can move onto the next tape. For decode you need to play and capture, then decode video, decode audio, merge. Obviously some of that can be automated for bulk tapes.

On the other hand, I could possibly build an ubuntu rig with several CX cards at once, and capture several cams at the same time, speeding up the actual capture process significantly if I have to do say, 100 tapes.

[Joscraft_05]

Just wanna chime in that i'm incredibly impressed by the results with the decode HI8 captures.

These were done with Sony CCD cam > modded CX card > ubuntu. Signal was in the 50s. Basically was trying to get "as good as conventional" results, so I set the levels to get it looking as close to the conventional capture as I could, and i'd say its 90% there.

Still weighing the pros and cons of switching over to decode for all 8mm captures.

On one hand, the process is longer. Conventional I can just play and capture, and my resulting archive file is done and I can move onto the next tape. For decode you need to play and capture, then decode video, decode audio, merge. Obviously some of that can be automated for bulk tapes.

On the other hand, I could possibly build an ubuntu rig with several CX cards at once, and capture several cams at the same time, speeding up the actual capture process significantly if I have to do say, 100 tapes.

The decode capture also looks interesting to me, although i said that i don’t quite like the results in this Hi8 video at all, I’m going to say that it is interesting and have room for improvements.

I'm still not sure about the potential of vhs-decode for “removing” or making the color bleed from these colour-under formats less visible and having the image better and more clean, but i'm interested with the results here for the noise and grain, how the vhs-decode capture made the image looks that clean even when probably part of that noise was from the CCD (here i'm not sure, i play a lot with the Hi8 format and the typical Hi8 handycams and when i looked at that noise, it looked like the noise i get with my Hi8 camera in low light from the CCD).

There’s room there for improvement with the sharpness and the dynamic range.

[dellsam34]

Okay i saw the results and well, i like that he got the fm rf capture with vhs-decode look without noise at all but i don't know if the image is a bit softer and less sharp and detailed or the noise from the "conventional" capture is tricking me.

There is noise in the conventional capture for sure, But the vhs-decode is way softer, Look at the lights' contours, They do look sharper in the conventional capture.

He should have turned the DNR on in the camcorder, That would have reduced the noise.

[VWestlife]

Even a MiniDV or Digital8 recording is going to have color noise, because consumer-grade digital camcorders process the image as luma/chroma (S-Video). To get true component video recording, you had to step up to the professional DVCAM or DVCPro formats.

[Attachment 93432 - Click to enlarge]

[Sharc]

Some capture cards (example: Hauppauge USB-Live2) use a pre-emphasis for high frequencies to make the picture look sharper. At the same time this amplifies the high frequency noise and tends to produce halos. All these variables present a real challenge for fair comparison. Also, I would suggest to compare unprocessed captures rather than postprocessed (deinterlaced, filtered ....) material.

[Joscraft_05]

Even a MiniDV or Digital8 recording is going to have color noise, because consumer-grade digital camcorders process the image as luma/chroma (S-Video). To get true component video recording, you had to step up to the professional DVCAM or DVCPro formats.

[Attachment 93432 - Click to enlarge]

Something like this could be for vhs-decode getting a lot less noise even for parts from the CCD in low light (again i'm not too sure about this, the video is softer and a bit darker there i think but it literally have no noise at least compared to the other capture), a lot less color bleed and that? Because of the process for FM RF captures, the software that improves everyone few months or so and avoiding Composite or S-video when either seeing the video from the sensor and the recording in the tape.

What i want for the software from vhs-decode is to improve the ringing/hallows, the sharpness judging from this video and a bit the dynamic range.

[The 14th Doctor]

You can get rid of some of the introduced ringing by applying a luma noise reduction using ld-analyze, does blur the image a bit though.

[Joscraft_05]

Some capture cards (example: Hauppauge USB-Live2) use a pre-emphasis for high frequencies to make the picture look sharper. At the same time this amplifies the high frequency noise and tends to produce halos. All these variables present a real challenge for fair comparison. Also, I would suggest to compare unprocessed captures rather than postprocessed (deinterlaced, filtered ....) material.

Yes i understand that unprocessed comparisons are very good too, but the "conventional" capture was encoded in a video format that's incompatible with my video players even VLC, so for that i considered to process both videos with my workflow for deinterlace, very slightly denoise and upscale with nnedi3_rpow2 and Spline144 for the comparisons.

[rgr]

and given the advantages in the optimized workflow freeing up the PC from the captures and having another device doing them is honestly very good.

But consider that you lose control over the capture software’s operations, including frame insertion/dropping and detailed per‑frame reporting as provided by AmarecTV.

Considering that what AmarecTV reports has nothing to do with reality, it's a small loss.
Besides, it would have to be tested first anyway, a sample would be useful.

[rgr]

Alright, got Nicholas Serra's permission to upload a snippet on the comparison he made with one of his own Hi8 tapes. He told me to say that when showing this off on the discord users chimed in saying that Video8/Hi8 video quality is already great as is when captured conventionally and properly, so decode ain't better from a quality standpoint despite the opposite claims being promoted around..

Comparing raw captures doesn't provide a good comparison. Such material needs to be postprocessed to see which one has a better image under the noise.

Original captures (snapshots taken from H264 compressed files):
https://www.diffchecker.com/image-compare/mfDSVScc/

After denoising according to the noise level (without sharpening):
https://www.diffchecker.com/image-compare/VHZHsfmh/

[lollo]

Considering that what AmarecTV reports has nothing to do with reality, it's a small loss.

Bullshit!

[lollo]

Comparing raw captures doesn't provide a good comparison. Such material needs to be postprocessed to see which one has a better image under the noise.

Non sense! The raw capture is the term of comparison. More, not everybody wants to (deeply) process (=alter) the capture.

[rgr]

Considering that what AmarecTV reports has nothing to do with reality, it's a small loss.

Bullshit!

Deny reality as much as you want, don't hold back!

Comparing raw captures doesn't provide a good comparison. Such material needs to be postprocessed to see which one has a better image under the noise.

Non sense! The raw capture is the term of comparison. More, not everybody wants to (deeply) process (=alter) the capture.

The capture process always involves some kind of video processing. This could be, for example, terrible temporal denoising (JVC), artificial sharpening by VCRs, luma boosting (LG, Toshiba), the same goes for DVD Recorder passthrough or even grabbers (AGC).

Postprocessing is also about reversing these changes.

You have screenshots, learn.

[lollo]

There's nothing to learn from your nonsense. I won't contaminate this thread with further replies, I'm just adding these new stupidities to your long list.

[rgr]

There's nothing to learn from your nonsense.

Blind fanaticism is a sign of stupidity, but you don't have to fight it

[Sharc]

The capture process always involves some kind of video processing. This could be, for example, terrible temporal denoising (JVC), artificial sharpening by VCRs, luma boosting (LG, Toshiba), the same goes for DVD Recorder passthrough or even grabbers (AGC).

And this is exactly what you should see and learn from the unprocessed capture, rather than what you end up with after applying your makeup from the beauty studio.

[rgr]

No, because there is little (or nothing) to be learned from such an image.

What do you see here on the jacket in the image on the right? Or what is the texture of the back wall of the furniture?
https://www.diffchecker.com/image-compare/mfDSVScc/
And it's already after deinterlace.

Comparing this to AI (makeup from beauty studio) is a misunderstanding.

[The 14th Doctor]

It's funny, getting decode captures with less noise defeats the purpose they propose. They view this noise as detail.

VHS is noisy, low light shot video is noisy. I've dealt with grainy as all hell VHS/Video8 tapes, I've dealt with clean VHS tapes and I've dealt with blurry copy of a copy VHS tapes. All captured on VCRs that produced very very similar quality results.

Unless it's LP or EP mode specifically where my JVCs produce a blurrier image than my Panasonic 1980. But those modes were always compromises, they were not the standard. SP mode is named Standard Play for a reason.

We are comparing raw samples, with no edits besides trimming. At most I'll crop and resize the image so the proportions match as close as possible between both samples, which would make it easier to view them side by side, but that is it.

[lollo]

Yes. And because the analog video is captured per field and not for frame, for interlaced material compare the fields not the frames.
Just as a complement you can do a bob(0,1) to visualize the frames to compare (any other deinterlacing "falsifies" the result), but you loose the check for quality of the image across time because of the "bobbing" effect.

A resize introduce regression, and in general should not be necessary. To match the frame/fields in horizontal and vertical position use crop() and addborders(), which are lossless operators.

[Sharc]

Yes. And because the analog video is captured per field and not for frame, for interlaced material compare the fields not the frames.
Just as a complement you can do a bob(0,1) to visualize the frames to compare (any other deinterlacing "falsifies" the result), but you loose the check for quality of the image across time because of the "bobbing" effect.

Good point, yes. Ultimately - once the differences become subtle - one should actually use the same video heads and same mechanical gear in order to eliminate as many variables as possible for the comparisons of the capture methods (classic vs VHS decode).
Interesting project for sure.