How to edit MiniDV footage so that it looks good

[John Matthews]

Here's the deal:

I have 43 minutes of consumer-standard miniDV footage of my wedding. My wife's asking for a nice video, similar to the ones I've offered to others in the family but with much better cameras. Does anyone know of some good examples of decent grades for this? Or, even some examples of interesting looks from low-quality stuff.

[newfoundmass]

2 hours ago, John Matthews said:

Here's the deal:

I have 43 minutes of consumer-standard miniDV footage of my wedding. My wife's asking for a nice video, similar to the ones I've offered to others in the family but with much better cameras. Does anyone know of some good examples of decent grades for this? Or, even some examples of interesting looks from low-quality stuff.

Ah yes, MiniDV! The bane of my existence for 12 years! A lot will come down to how you capture it (are you doing a standard firewire capture to DV or are you capturing it via capture card to a different codec?), how well it was filmed (is the exposure and white balance OK?) and what camera was used (a lot of low end camcorders just had horrible, muddy colors unless used in bright, bright lighting). 

You'll want to avoid using one of the cheaper capture cards that will record it to h.264. I've never had good results capturing SD video to that codec and editing it, so I think it's better to capture to DV if you have the ability to. ProRes will also work if you have a way to capture it that way!

If the footage is decent-ish you should be able to do a bit of color correction and get relatively close to what you'd get using a similar modern consumer camera, but you're not gonna be able to get away with heavy grades or anything. The process overall though wasn't radically different to how I worked with HD camcorder footage once I upgraded, so assuming it's decent footage you should be OK doing what you normally would just with tempered expectations. 

If it's not very good footage though you'll have more issues than you would if it were HD. Denoising never really worked that well since it removed way too much detail for a lower resolution image. Maybe that has gotten better thanks to better denoising tools? But muddy and noisy footage is where you'll have your problems; there isn't a ton you can do with that. Even basic color correction is a pain in the ass. 

 

[projectwoofer]

4 hours ago, John Matthews said:

Here's the deal:

I have 43 minutes of consumer-standard miniDV footage of my wedding. My wife's asking for a nice video, similar to the ones I've offered to others in the family but with much better cameras. Does anyone know of some good examples of decent grades for this? Or, even some examples of interesting looks from low-quality stuff.

Assuming you have a way to properly capture the footage in your computer, maybe try an upscaling software afterwards? I’ve had very good results with Topaz VEAI, their new model Proteus is amazing.

[kye]

You will struggle to get anything that looks 'great', so I would suggest a three-pronged approach:

1. Content.  If ever there was a time to make the grade support the content rather than be a spectacle in itself, this is it.  One trick that often works is to go carefully shot by shot and ensure that the viewers attention is on the right thing / person, and that is often done by raising the luma of the area so it stands out.  Typically this is large and very soft oval power-windows the way that you'd simulate a vignette, and is often a combination of inside-outside effects where the subject is pulled up a bit and the outside is pulled down a bit.  If you do it softly enough then it should be imperceptible, but will put the focus on where it needs to be, distracting from the grade entirely.
2. As there's no chance of making a 'great' grade, make one that is the least offensive.  Do this by working out the weakness of the footage and then compensating for that.  For example, in low DR 8-bit footage, often the mid-tones are stretched apart, so you could try lowering the overall contrast and essentially pulling the bits closer together.  This will raise the shadows, and maybe this will give a nice vintage look, but maybe not, and maybe you'd be better off actually pulling the levels down a bit and compressing the blacks (and whatever compression nasties are there).  You'd have to play with the footage to see where the issues are.
   It's worth experimenting with reducing saturation, which makes all footage look higher end, but may not be the aesthetic you're looking for.
3. Degrade the footage to hide its sins.  I've shot lots of very low quality footage over the years, including SD, bad codec stuff, and have had a lot of mileage from upscaling it, blurring it and adding grain.  
   Upscaling it gives you more pixels to work with, and your timeline should absolutely be higher resolution than the original footage.
   Blurring at very small radius' is designed to soften the jagged compression artefacts.  Blurring at slightly larger radius' will smooth gradients, and much larger blurs (essentially adding diffusion in post) will smooth over the image the same way as diffusion filters do.  I'd suggest the smallest blur be at 100%, and the larger ones be semi-transparent over the image.
   Adding grain will disguise the fact you've blurred the footage, and will give an analog feel, which is far more preferable than having an early-digital feel.
   For this step I'd encourage you to grade at a standard viewing distance rather than close-up / pixel-peeping as the effects are easy to over-do.  If you're going to upload it, do a version specifically that takes into account the compression that will be applied, and you'll have to add lots more grain (for example) and probably experiment with multiple uploads to fine-tune it.

Here's an example of my previous attempts at making the best of low quality footage.

Final:

Original:

Best of luck - editing old footage can be fun and a trip down memory lane.  Enjoy the process!

[kye]

Oh yeah, and a nice highlight rolloff can hide clipped highlights, and if the footage is 30p you can slow it down to 24p and get a bit of slow-motion for free, which I also did on the above footage.

[majoraxis]

Seems like a job for Topaz Labs's Video Enhance AI after capture.

https://www.topazlabs.com/video-enhance-ai

[Mr. Freeze]

1 hour ago, majoraxis said:

Seems like a job for Topaz Labs's Video Enhance AI after capture.

https://www.topazlabs.com/video-enhance-ai

Or superscale in Resolve.

[Ingerson]

Last summer I managed to salvage old tapes before they went too bad. Didn't try to make the footage look pro or have any special "look", but to make new HD masters to save for the future (=pure sentimental reasons).

I tried AI scaling, but was very unsatisfied, look horrible. This thread is amazing and I went with this workflow: https://forum.blackmagicdesign.com/viewtopic.php?f=3&t=109259&sid=eb50d17426119095ce9038dba8e7c6a6

Good upscaling without artifacts or false details. Also the best (software) de-interlacer I've ever tried (I tried almost everyone there is), so many hidden details and perceived resolution that would disappear with other solutions.

So with a better codec, in standard HD resolutions (with square pixel aspect ratio) I retouched corrupt frames and did gentle colour and contrast balancing. My favourite tool in Resolve is "Midtone detail" with a negative value of -25 to -50, really helps with the harshness of the format. No need to work on the sharpness or the contrast for my needs for a soft retro-ish presentation.

 

[John Matthews]

21 hours ago, newfoundmass said:

Ah yes, MiniDV! The bane of my existence for 12 years! A lot will come down to how you capture it (are you doing a standard firewire capture to DV or are you capturing it via capture card to a different codec?), how well it was filmed (is the exposure and white balance OK?) and what camera was used (a lot of low end camcorders just had horrible, muddy colors unless used in bright, bright lighting). 

All of my stuff was captured via firewire 400. I believe that was the highest quality I could go for. In QuickTime it says the resolution is 720 × 480 (640 × 480), fps is 29.97 (interlaced), at roughly 3.6mbps. I've got moiré and aliasing all over the place. Colors are wash-out.

19 hours ago, projectwoofer said:

Assuming you have a way to properly capture the footage in your computer, maybe try an upscaling software afterwards? I’ve had very good results with Topaz VEAI, their new model Proteus is amazing.

I'm not really into the AI thing... it always looks weird and unnatural to my brain. Is it empirically better? Probably, but my brain just doesn't associate emotions with it yet- only "wow, that's cool".

18 hours ago, kye said:

Content.  If ever there was a time to make the grade support the content rather than be a spectacle in itself, this is it.  One trick that often works is to go carefully shot by shot and ensure that the viewers attention is on the right thing / person, and that is often done by raising the luma of the area so it stands out.  Typically this is large and very soft oval power-windows the way that you'd simulate a vignette, and is often a combination of inside-outside effects where the subject is pulled up a bit and the outside is pulled down a bit.  If you do it softly enough then it should be imperceptible, but will put the focus on where it needs to be, distracting from the grade entirely.

Yes. This is going to take time, for sure. Most of the footage is stupid stuff like me shaking a friend's had or other banalities. Of my 43 minutes, I might use 4 if I'm lucky. I need to make a story of it first. Putting subtle grades will probably be my best guess too. It could take me a couple of weeks to get more familiar with it.

20 hours ago, kye said:

As there's no chance of making a 'great' grade, make one that is the least offensive.  Do this by working out the weakness of the footage and then compensating for that.  For example, in low DR 8-bit footage, often the mid-tones are stretched apart, so you could try lowering the overall contrast and essentially pulling the bits closer together.  This will raise the shadows, and maybe this will give a nice vintage look, but maybe not, and maybe you'd be better off actually pulling the levels down a bit and compressing the blacks (and whatever compression nasties are there).  You'd have to play with the footage to see where the issues are.
It's worth experimenting with reducing saturation, which makes all footage look higher end, but may not be the aesthetic you're looking for.

Yes. I'm aware of the contrast trick. I might also go black and white... not sure, but since it was over ten years ago, it might make sense.

 

20 hours ago, kye said:

Degrade the footage to hide its sins.

I'm clearly going to try more of this. It might make it etherial but still look like something I lived. I watched and liked your edit. It almost looked like you shot it that way, making more "believable".

20 hours ago, kye said:

Oh yeah, and a nice highlight rolloff can hide clipped highlights, and if the footage is 30p you can slow it down to 24p and get a bit of slow-motion for free, which I also did on the above footage.

I tried that, but I was getting some horrible flickering action. It might be a technical thing, but I'll into more. I think the flickering was due to the footage being interlaced.

7 hours ago, Ingerson said:

I tried AI scaling, but was very unsatisfied, look horrible. This thread is amazing and I went with this workflow

I'll look into this too. Maybe it's better than the in-your-face AI that I've seen before. Thanks.

[projectwoofer]

18 minutes ago, John Matthews said:

All of my stuff was captured via firewire 400. I believe that was the highest quality I could go for. In QuickTime it says the resolution is 720 × 480 (640 × 480), fps is 29.97 (interlaced), at roughly 3.6mbps. I've got moiré and aliasing all over the place. Colors are wash-out.

I'm not really into the AI thing... it always looks weird and unnatural to my brain. Is it empirically better? Probably, but my brain just doesn't associate emotions with it yet- only "wow, that's cool".

Yes. This is going to take time, for sure. Most of the footage is stupid stuff like me shaking a friend's had or other banalities. Of my 43 minutes, I might use 4 if I'm lucky. I need to make a story of it first. Putting subtle grades will probably be my best guess too. It could take me a couple of weeks to get more familiar with it.

Yes. I'm aware of the contrast trick. I might also go black and white... not sure, but since it was over ten years ago, it might make sense.

 

I'm clearly going to try more of this. It might make it etherial but still look like something I lived. I watched and liked your edit. It almost looked like you shot it that way, making more "believable".

I tried that, but I was getting some horrible flickering action. It might be a technical thing, but I'll into more. I think the flickering was due to the footage being interlaced.

I'll look into this too. Maybe it's better than the in-your-face AI that I've seen before. Thanks.

Fair enough. About the VEAI, some models like Artemis for example tend to overdo it sometimes, especially with really bad source footage. Proteus I find is a totally different story and very configurable too. I've never noticed any weirdness with decent quality source footage. All I can say is I can finally use the otherwise horrible 120fps on my S5, and all its FHD footage for that matter! It's really that good. But yeah, only you could tell if something like that is for you.

[kye]

4 hours ago, John Matthews said:

All of my stuff was captured via firewire 400. I believe that was the highest quality I could go for. In QuickTime it says the resolution is 720 × 480 (640 × 480), fps is 29.97 (interlaced), at roughly 3.6mbps. I've got moiré and aliasing all over the place. Colors are wash-out.

That doesn't sound right.  I'd maybe reach out to some experts, like on the LiftGammaGain forums perhaps (colourists used to do a lot of DI stuff before everything went digital) and confirm.  I hear those guys talking about the older digital formats and there were so many gotchas in there that it's almost amazing that anyone got a good result!  

4 hours ago, John Matthews said:

Yes. This is going to take time, for sure. Most of the footage is stupid stuff like me shaking a friend's had or other banalities. Of my 43 minutes, I might use 4 if I'm lucky. I need to make a story of it first. Putting subtle grades will probably be my best guess too. It could take me a couple of weeks to get more familiar with it.

Yeah, because it's a personal project you can really take your time and let your subconscious process the footage in the background as you do other things.  

I'm always amazed at how a good editor can put little moments together that aren't related, but compose a thread that wasn't literally there, but somehow captures the feeling more than a 'correct' version.  Like when Herzog said “Facts do not constitute the truth. There is a deeper stratum.” 

In a way, you're lucky that it's miniDV because you're less likely to get swayed by things that look good but aren't meaningful.  I'm sure that if I had a 1DX2 and fast primes I'd be tempted to include too much 120p B-roll in my edits like McKinnon does, so you've side-stepped these challenges!

4 hours ago, John Matthews said:

I'm clearly going to try more of this. It might make it etherial but still look like something I lived. I watched and liked your edit. It almost looked like you shot it that way, making more "believable".

Perhaps calling it degradation was misrepresenting it.  Think of it like this...  reality doesn't have compression artefacts, it's not 'sharp', it's not grainy, and smooth surfaces are not featureless, but the miniDV footage has introduced these things due to its limitations.  Your job is to make the footage look the most like reality was, or the most how it felt.  You've definitely got license to take some creative liberties to make it more like poetry and less like prose!

If you don't blur the footage at all, you get all the nasties and all the content in the footage.  If you blur it a lot then you get no nasties and very little of the content, but there will be a little sweet spot where each effect hides much more of the nasties than the original footage, and that's what you're aiming for.

I'd suggest setting up a bunch of effects and going through them one by one at a sensible viewing distance and fine-tuning their parameters and opacity by eye, trying to make the footage look the most like a window to being there, or of a Hollywood film of being there (or whichever aesthetic you prefer!).  By adjusting those parameters by looking at the monitor and not the control panel you may find that effects that aren't worthy get set to the sweet spot of zero opacity. Make sure you're always turning each effect on/off to make sure you're improving the image, as it's easy to get lost and get used to something, but when watching things we can also easily get used to quite strong looks as well, so don't shy from making it stylised. I'd make a few passes through each effect to optimise each one in combination with the others.
A cool point of reference is to look from the screen to the other objects in the room, so effectively using reality as a reference.  Comparing to reality will quickly sort out how sharp/unsharp things should be etc.
Save that as a preset, reset them all and do the same exercise a few days later.  Do it a few times.  Then you can compare them and see which you liked, and maybe blend them together.  

You are trying to create a look that looks 'right' and the original footage looks awful in comparison.

When I've done this (like in the example I posted above) I got to a point where the final grade seemed like footage from a Super-16 film camera and the original footage seemed like someone had applied a bunch of awful distortions for some reason.  That's a good place to end up.

It's useful to review some older film captures as a reference too.  They were blurrier than you think, but never looked offensive, so it's a good look to go for.

4 hours ago, John Matthews said:

I tried that, but I was getting some horrible flickering action. It might be a technical thing, but I'll into more. I think the flickering was due to the footage being interlaced.

It could be due to the interlacing, but it could also be due to any capture issues you're experiencing.  Once you confirm your capture then just see what's least offensive.

[John Matthews]

6 hours ago, kye said:

If you don't blur the footage at all, you get all the nasties and all the content in the footage.  If you blur it a lot then you get no nasties and very little of the content, but there will be a little sweet spot where each effect hides much more of the nasties than the original footage, and that's what you're aiming for.

I looked again at the footage last night. It seems that I can change aspect ratios and do some blur gradients to hide quite a lot. It will need to be done on a shot-by-shot basis rather than just a one-size-fits all approach.

6 hours ago, kye said:

It could be due to the interlacing, but it could also be due to any capture issues you're experiencing.  Once you confirm your capture then just see what's least offensive.

I'll confirm this after some research today. I remember capturing it though. I used iMovie with firewire. That generation of miniDV was interlaced (captured on a Canon ZR930). I still have the option of recapturing the video in that I still have many old Macs and my ZR930... everything still works fine. Again, more research is needed.

[John Matthews]

My initial research suggests that miniDV captured through Firewire 400 with iMovie was as good as it gets. The resulting files are a ".dv" files which are simply a stream of digital video. Since it's all digital and nothing was edited in a destructive format, I think that might be the best I can do. If anyone knows differently, please tell me.

[UncleBobsPhotography]

10 minutes ago, John Matthews said:

My initial research suggests that miniDV captured through Firewire 400 with iMovie was as good as it gets. The resulting files are a ".dv" files which are simply a stream of digital video. Since it's all digital and nothing was edited in a destructive format, I think that might be the best I can do. If anyone knows differently, please tell me.

Did you use a computer with a Firewire 400 port or use a Firewire to USB adapter? 

I've got 120 MiniDV cassettes I should have archived which I've been postponing for years, blaming it on lacking the right connector.

[John Matthews]

2 minutes ago, UncleBobsPhotography said:

Did you use a computer with a Firewire 400 port or use a Firewire to USB adapter? 

I've got 120 MiniDV cassettes I should have archived which I've been postponing for years, blaming it on lacking the right connector.

I used one of my either a iBook G4 or one of my two Intel MacBooks, all of which have Firewire ports. The supplied cable was a firewire to MiniDV connected straight into the camera.

[kye]

21 hours ago, John Matthews said:

I looked again at the footage last night. It seems that I can change aspect ratios and do some blur gradients to hide quite a lot. It will need to be done on a shot-by-shot basis rather than just a one-size-fits all approach.

Great to hear!

If there's any before/after you're willing to share I'd be curious to see it.  I don't recall ever seeing other people doing this stuff, just me playing around with things.

There's likely to be a reasonably good average grade that would work across all shots that you can use while editing, and then fine-tune once you're almost at picture-lock.  That avoids you working on shots that don't end up in the final edit, but also means you're not working with the unprocessed footage, which can be distracting.

21 hours ago, John Matthews said:

My initial research suggests that miniDV captured through Firewire 400 with iMovie was as good as it gets. The resulting files are a ".dv" files which are simply a stream of digital video. Since it's all digital and nothing was edited in a destructive format, I think that might be the best I can do. If anyone knows differently, please tell me.

If that's a true stream of the data on the tape then that should be as good as it gets.  Then the next steps would be how to de-interlace and process it further.

I seem to recall several ways to de-interlace, either with their pros and cons.  One was to have each frame made up of alternating lines from the current and previous frame, and the other was to just duplicate the lines from the current frame.  The former had more apparent resolution but had a horizontal blind type of effect on movement, and the latter had less resolution and was prone to flickering, especially on hard edges that ran horizontally.  Choosing the overall approach might be subject-dependent, and maybe even shot dependent?  

I'm curious to hear how you go with this as well.

[ac6000cw]

6 hours ago, kye said:

If that's a true stream of the data on the tape then that should be as good as it gets.  Then the next steps would be how to de-interlace and process it further.

I seem to recall several ways to de-interlace, either with their pros and cons.  One was to have each frame made up of alternating lines from the current and previous frame, and the other was to just duplicate the lines from the current frame.  The former had more apparent resolution but had a horizontal blind type of effect on movement, and the latter had less resolution and was prone to flickering, especially on hard edges that ran horizontally.  Choosing the overall approach might be subject-dependent, and maybe even shot dependent?  

Modern de-interlacers are adapative, so they use different combinations of techniques in different parts of the frame, depending on things like how much motion and inter-field differences there are.

For DV I use command line FFmpeg to deinterlace (using 'yadif'), upscale and sharpen - in the example command lines below input is DV avi, output is H264+AAC audio at 1920x1080p at double the DV frame rate (so 25 fps interlaced -> 50 fps progressive):

ffmpeg.exe -i <input_file>.avi -vf yadif=1,scale=1920x1080:flags=lanczos+accurate_rnd+full_chroma_int+full_chroma_inp,unsharp -c:v h264_qsv -global_quality 30 -look_ahead 1 -c:a aac -b:a 384k <output_file>.mp4

You can add motion-compensated de-interlacing 'mcdeint' to the processing, which improves the deinterlacing quality but is quite a bit slower:

ffmpeg.exe -i <input_file>.avi -vf yadif=1,mcdeint=0:1:10,scale=1920x1080:flags=lanczos+accurate_rnd+full_chroma_int+full_chroma_inp,unsharp -c:v h264_qsv -global_quality 30 -look_ahead 1 -c:a aac -b:a 384k <output_file>.mp4
 

(I'm using nVidia hardware H264 encoding for the output in the above examples - h264_qsv - but you could use any of the many FFmpeg software encoders instead e.g. ProRes etc.)

I got some of the ideas for the deinterlacing settings from here: https://askubuntu.com/questions/866186/how-to-get-good-quality-when-converting-digital-video

[Ingerson]

Follow up to my post about upscaling and de-interlacing with QTGMC and then playing with Midtone detail in negative values.

Picked this frame since it's pretty classic DV and bad lighting. First frame is the original DV (in 720p timeline), then the upscaled frame and #3 the upscaled frame with -50 mid/detail, slightly lowered gain and lift and lowest possible glow. Screencap of Resolve at the end.

Not really something magic, still not great, but definitely better.

 

[Brian Flint]

About 12 years ago I did a lot of home movies of family using a Canon Camcorder with a DV tape. I bought a Firewire card, IEEE 1394, and fitted into my PC. With suitable cable from Camcorder to PC I transferred the recordings to the Hard Drive on my PC. This resulted in .avi file  720 x 576 , 25.00 fps , Data rate 28887kbps. Interlaced. In some cases I had 'wide screen PAL' ie 16 by 9 and in other cases I had 4 by 3 ( widescreen not selected when recording ). Back then I decided to transcode the .avi to mpg files suitable to burn onto DVD . The bit rate drops to around 6Mbps and I kept the frame rate at 25 fps interlaced. This gave good results for a family movie. A year or so ago, I decided that DVDs are not future proof and I need to make the footage be 16 by 9 with black bars either side if original was 4 by 3 and use progressive scan - ie de-interlace and use a resolution of 1920 x 1080. I felt the biggest problem was the de-interlacing of fast moving image - ie children running about. There is also another problem I find today and that there is difficulty to find an Editor which will accepting the original .avi files. Resolve 16 does not accept it.  Anyway I have managed one way or another to make the format to what I want, but I may have not choose the best way to do it. When I show the footage to my grandchildren they just see the content and couldn't care about the finer details of the quality.