Magic Lantern Forum

Showcasing Magic Lantern => Share Your Videos => Topic started by: beauchampy on October 13, 2014, 09:52:40 PM

Title: Neat Video tests on RAW footage
Post by: beauchampy on October 13, 2014, 09:52:40 PM


Did a few tests using Neat Video (Resolve OFX version) on some RAW footage (see this thread to see the full feature videohttp://www.magiclantern.fm/forum/index.php?topic=13571.0 (http://www.magiclantern.fm/forum/index.php?topic=13571.0))

As with all grading processes, something we want to boost gamma or not crush the shadows - but there's noise there. Neat video does an incredible job of letting me grade exactly how I want, without having to worry about the resulting noise. It does a fantastic job for ML raw video.

For me and my workflow, its an essential plugin.
Title: Re: Neat Video tests on RAW footage
Post by: swinxx on October 13, 2014, 10:59:13 PM
hello,
i also love neatvideo and with version 3.6 it seems that it is now stable with davinci resolve 11.1
however i am shooting a short movie and using it for every shot, with great results.




greets.
Title: Re: Neat Video tests on RAW footage
Post by: chmee on October 14, 2014, 10:35:32 AM
@swinxx:
the color-scenes are awesome. chapeau.
Title: Re: Neat Video tests on RAW footage
Post by: swinxx on October 14, 2014, 05:54:00 PM
Oh. Danke ;)
Title: Re: Neat Video tests on RAW footage
Post by: Canon eos m on October 14, 2014, 06:13:41 PM
Love the cinematic look!
Title: Re: Neat Video tests on RAW footage
Post by: kgv5 on October 23, 2014, 03:41:41 PM
What can you tell about denoising speed in resolve?
EDIT: ok, i can answer my own question. Demo version gives about 3-4 fps (during rendering) compared to 18-20 fps without denoising. I7 with GTX570 win7 64bit.
Title: Re: Neat Video tests on RAW footage
Post by: swinxx on October 24, 2014, 07:34:25 AM
What matters is gpu power ;)
Title: Re: Neat Video tests on RAW footage
Post by: beauchampy on October 24, 2014, 11:21:13 AM
Yeah I only get 3-4fps on my MBPr 2013 (Geforce GT 750M 2048mb)
Title: Re: Neat Video tests on RAW footage
Post by: swinxx on October 24, 2014, 11:32:22 AM
Ok, i hava an hackintosh with a 680gtx and 4gb, its way fastet of course.
Perhaps you can gain a little bit with optimization in neat video itself, which version have you installed?
Title: Re: Neat Video tests on RAW footage
Post by: kgv5 on October 24, 2014, 12:21:00 PM
3-4 fps is not bad, denoising quality is outstanding, I didnt expect such a great results! Its unbelieveble how good, sharp and detailed picture is, its light years ahead in comparison with denoising any h264 footage. 3-4 fps it's at least 3 times faster than rendering in AE cs6 (1 fps on my computer).  I was testing raw with iso 12800 (5d3) - wow, neat makes such a footage totally usable imho, its like having some new camera ;)
Title: Re: Neat Video tests on RAW footage
Post by: swinxx on October 24, 2014, 02:55:23 PM
When you compare ae and davinci its like comparing trabi vs ferrari ;) resolve is much more faster in every instance ;)
Title: Re: Neat Video tests on RAW footage
Post by: N/A on October 24, 2014, 03:28:02 PM
Is everyone denoising before color correction/grading,  or after grading, on final render? I was using it on every clip when I shot h264, but I've barely used it since switching to raw, I do a bit of nr in ACR at the debayer stage but I'm considering switching to denoising on final render.
Title: Re: Neat Video tests on RAW footage
Post by: beauchampy on October 24, 2014, 04:52:54 PM
Quote from: N/A on October 24, 2014, 03:28:02 PM
Is everyone denoising before color correction/grading,  or after grading, on final render? I was using it on every clip when I shot h264, but I've barely used it since switching to raw, I do a bit of nr in ACR at the debayer stage but I'm considering switching to denoising on final render.

Im applying it on my second to last node, with my last node being grain.
Title: Re: Neat Video tests on RAW footage
Post by: swinxx on October 24, 2014, 09:26:46 PM
hello,
i have made many tests over the last year and it should be like this:

1. denoising in the first node, then
2. color correction, then
3. sharpening, then
4. grain.


greets.
Title: Re: Neat Video tests on RAW footage
Post by: N/A on October 24, 2014, 10:03:00 PM
But Neat Video also has a sharpening option, do you not use that?
Title: Re: Neat Video tests on RAW footage
Post by: swinxx on October 24, 2014, 10:18:18 PM
ah yes, i use it, but in a moderate manner cause i dont want to overdo it.
the sharpening in resolve is also great - try it out, in the last node, also in a moderate way.
i just use my eyes.. and sharpen the image until i see the difference. then i dail it back 10-15%.

greets.
swinxx
Title: Re: Neat Video tests on RAW footage
Post by: baldavenger on October 24, 2014, 10:51:33 PM
Quote from: beauchampy on October 24, 2014, 04:52:54 PM
Im applying it on my second to last node, with my last node being grain.

Do you apply the denoise before or after a LUT (if and when you use one)?
Title: Re: Neat Video tests on RAW footage
Post by: DFM on October 25, 2014, 12:21:18 AM
Remember that with raw footage, or anything captured as a pure frame sequence, the requirement for noise removal is very different than for a video codec like H.264. The thermal noise on each DNG is random (so much so, sampling shot noise is one of the best ways to create random numbers in crypotgraphy), hence it's much faster and simpler to reduce it per-frame using ACR etc. as if you're cleaning up a still image. Neat Video has some extremely good algorithms but they are expecting a temporal relationship to the noise signal caused by frame interpolation in the codec, which is why it's such a CPU-intensive plugin. It'll work on raw sequences but it's overkill.
Title: Re: Neat Video tests on RAW footage
Post by: baldavenger on October 25, 2014, 01:59:48 AM
Quote from: DFM on October 25, 2014, 12:21:18 AM
Remember that with raw footage, or anything captured as a pure frame sequence, the requirement for noise removal is very different than for a video codec like H.264. The thermal noise on each DNG is random (so much so, sampling shot noise is one of the best ways to create random numbers in crypotgraphy), hence it's much faster and simpler to reduce it per-frame using ACR etc. as if you're cleaning up a still image. Neat Video has some extremely good algorithms but they are expecting a temporal relationship to the noise signal caused by frame interpolation in the codec, which is why it's such a CPU-intensive plugin. It'll work on raw sequences but it's overkill.

This is the first I have heard of this, so I have to say thank you as you have helped clear up much confusion.  Spatial is the way to go.

Has anyone tried the de-noise/de-grain ofx plugins from Sapphire or Boris FX yet?
Title: Re: Neat Video tests on RAW footage
Post by: dmilligan on October 25, 2014, 03:24:27 AM
Quote from: DFM on October 25, 2014, 12:21:18 AM
Remember that with raw footage, or anything captured as a pure frame sequence, the requirement for noise removal is very different than for a video codec like H.264. The thermal noise on each DNG is random (so much so, sampling shot noise is one of the best ways to create random numbers in crypotgraphy), hence it's much faster and simpler to reduce it per-frame using ACR etc. as if you're cleaning up a still image. Neat Video has some extremely good algorithms but they are expecting a temporal relationship to the noise signal caused by frame interpolation in the codec, which is why it's such a CPU-intensive plugin. It'll work on raw sequences but it's overkill.
This just isn't correct at all. Just because frames are raw doesn't mean they don't have a temporal relationship. Why do you think astrophotographers stack images? The most effective way to reduce photon noise is to capture more photons. Averaging frames accomplishes this. A good "temporal" video noise reduction algorithm can leverage this fact to average static areas of the frame, effectively reducing noise in those areas (and not artificially reducing noise by blurring or smoothing or what not, but actually truly increasing the SNR, exactly how APers do when they stack frames). Photon noise is truly random as you say, but it is precisely this fact that means it does have a "temporal" relationship => the more photons you collect the lower the noise => when you take another frame, you've collected more photons.
Title: Re: Neat Video tests on RAW footage
Post by: swinxx on October 25, 2014, 07:08:28 AM
hey dmilligan,
thx for good explaination, do you know a good astro stacking tool for mac?

btw: for me neat video works really well!
greets.
Title: Re: Neat Video tests on RAW footage
Post by: kgv5 on October 25, 2014, 07:21:28 AM
Quote from: swinxx on October 24, 2014, 09:26:46 PM
hello,
i have made many tests over the last year and it should be like this:

1. denoising in the first node, then
2. color correction, then
3. sharpening, then
4. grain.


greets.

Sorry for the little offtop but what do you use for sharpening in node? YUV color space on channel 1 and then blur to -47 lets say? Or some different method? Do you use also sharpening in RAW tab? Thanks
Title: Re: Neat Video tests on RAW footage
Post by: dmilligan on October 25, 2014, 05:52:30 PM
Quote from: swinxx on October 25, 2014, 07:08:28 AM
thx for good explaination, do you know a good astro stacking tool for mac
I use PixInsight. Hands down the best astrophotography software there is. Tough learning curve and not free, but it is the defacto standard.
Title: Re: Neat Video tests on RAW footage
Post by: KurtAugust on October 25, 2014, 07:32:46 PM
Swinxx,

I enjoyed this very much. Deeply. Makes me want to see the whole film. I would like to see the teaser with less compression. Is that possible? For once, anamorphics used well, adding psychological depth to the image, instead as a lens flare machine.
Title: Re: Neat Video tests on RAW footage
Post by: swinxx on October 25, 2014, 08:26:52 PM
@dmilligan..
i took a quick look at pixinsight.. puh.. not very user friendly.. my first question would be.. where should i begin. this software is indeed not a simple to use application. i think i need a one button solution :)

@kurtaugust.
thank you very much.. vimeo and youtube have some compression algos, sadly i have no vimeo plus. if you wanna download it, i can put it into my copy.com account and send you a link.

greets. swinxx
Title: Re: Neat Video tests on RAW footage
Post by: KurtAugust on October 25, 2014, 09:17:04 PM
I'd love that.
Title: Re: Neat Video tests on RAW footage
Post by: SiSS on February 02, 2016, 01:22:05 AM
2016, Neat Video still relevant for MLV and Premiere Pro CC?
Title: Re: Neat Video tests on RAW footage
Post by: beauchampy on February 04, 2016, 06:12:49 PM
Quote from: SiSS on February 02, 2016, 01:22:05 AM
2016, Neat Video still relevant for MLV and Premiere Pro CC?

For sure! I still use it on every production.
Title: Re: Neat Video tests on RAW footage
Post by: Lakwaun on February 08, 2016, 11:35:03 PM
Hmm. Just doing some reading as I shot a raw video and it turned out with a lot of noise after converting it.
My question is about work flow... Neat video does work good but it takes to long to render and I am good at Adobe Camera Raw. Which is the best workflow for say music videos.
Im going to learn Davinci before trying or buying it but is it really better then After Effects or can really both achieve good results.
Just wondering also when people denoise say using neat video, then color correct and sharpen (using unsharpened mask I presume) do people usually export that out into a uncompressed video format, then start editing? I always use dynamic link since i have already chopped up multiple sequences into one using multi cam applying neat video on each prepared sequenced scene.
Is that why my renders take so long? Is there a easier way beside im sure the not working way like adding one instance of neat video over the main sequence in a adjustment layer or something.
Title: Re: Neat Video tests on RAW footage
Post by: beauchampy on February 09, 2016, 05:58:30 PM
Quote from: Lakwaun on February 08, 2016, 11:35:03 PM
Hmm. Just doing some reading as I shot a raw video and it turned out with a lot of noise after converting it.
My question is about work flow... Neat video does work good but it takes to long to render and I am good at Adobe Camera Raw. Which is the best workflow for say music videos.
Im going to learn Davinci before trying or buying it but is it really better then After Effects or can really both achieve good results.
Just wondering also when people denoise say using neat video, then color correct and sharpen (using unsharpened mask I presume) do people usually export that out into a uncompressed video format, then start editing? I always use dynamic link since i have already chopped up multiple sequences into one using multi cam applying neat video on each prepared sequenced scene.
Is that why my renders take so long? Is there a easier way beside im sure the not working way like adding one instance of neat video over the main sequence in a adjustment layer or something.

It's processor intensive, so you want to avoid denoising your rushes.

My workflow (Resolve) is this:

1. Convert DNG's to ProRes Proxy in Resolve
2. Edit the video in Premiere
3. Export XML from Premiere
4. Open XML in Resolve, linking to original DNG files (we're not working on the Proxies any more) your edit will now show in Resolve
5. De-noise and grade each shot individually.
6. Export finished video from Resolve.