GH1 Action Short / Field Test

[Jack Daniel Stanley]

Actually you can go with ProRes. Apple has the codec for windows here:
http://support.apple.com/downloads/Apple_ProRes_QuickTime_Decoder_1_0_for_Mac

I've been a big fan of ProRes for a long time - it does add some noise though plus we're talking the internet here and ProRes = Ginormous.

So goin the BlueRay qual H.264 route. Don't worry, if it's not representative of what I'm seeing right out of FCP I'll look into ProRes - posting inferiorly compressed stuff with imperfections I can't see in the source material for the sake of pixel peeping is frustrating.

 

[Jack Daniel Stanley]

Hey guys sorry for the delay. I wanted to get to the bottom of this frame rate / conversion issue so I've done some more tests which I hope to have up Wednesday.

Basically I shoemounted a GH1 atop a HPX 170, went to an intersection and taped cars crossing (it was just after a speed bump so they're going 10 miles an hour or so). Anway the idea with taping cars crossing the frame is that they move at a constant speed and any breaks in cadence or jerkiness will show.

The results of the test are interesting and it was surprising how the GH1 image held up and in many ways surpassed the 170 for a locked down early evening shot.

As for the motion results, the motion of 720 60p conformed and retimed to 23.98 in compressor is virtually identical, or incredily close to, if not exactly like native 24p. And - it appeared to have higher resolution than the HPX at 1080. As for 30p? Never going to be as smooth of a conversion buty you can get away with it if you're not showing something like a car which should basically be gliding across the screen instead of jerking across it.

So will have that up today (Wednesday) then hopefully get back to finishing the short / implementing the new conversion methods I've learned.

In the meantime, this was PMed to me:

Hi Jack.

Thought better to ask this outside of the thread as it is slightly off the topic. (Thanks for doing that, your work no matter what the situation is always excellent and inspiring!)

Anyways, I was just curious what you used to do the color correction and what if any steps, tricks or presets etc. you used?

I bought a panasonic hs300 off ebay, and would love to try to grade it to that look.

If you have a moment I'd love to know any info on the color grading.

Thanks, look forward to more of your work!

Staven thanks for the comments.

Grading questions are totally appropriate here sense this is about putting the GH1 through all phases of a beginning to end process.

I shot in the camera's "SMOOTH MODE" a setting that is supposed to be "lower contrast slightly desaturated with smooth colors". I did this because one theory I was testing to not get mud was to lessen extremes in the image. There are three or four settings in each scene file. Contrast, Color, Sharpness, and Noise Reduction. I basically zeroed out Contrast, Sharpness and Noise Reduction. I didn't want any weird artifacts from the noise reduction. A reduction in sharpness might reduce mud, and same for contrast. Also reduced since I was shooting in some MJPEG modes where I've seen some oversharpening in the compression.

BUT I don't think this set me up for grading very well. I need to switch gears from how I worked with the HVX, HPX500, HPX 170 with their higher color space to how I used to work with the DVX. When Shooting with the mentioned HD cameras I found it best to zero everything out, shoot neutral or slightly desaturated colors, and neutral or low contrast. This gave me a lot to work with in post - but when I began with the DVX, I used to shoot with the saturation pushed in an effort to force a little more color information through to the sensor. Though this was kind of faux color information, it nonetheless helped in grading. So if I had to do it again, I might up the saturation or shoot in a more saturated mode.

I do think the "Smooth" and "Nostalgia" modes produce very good in camera, filmmic looks but may not work best for post work.

As for the grading I did itself, it always comes down to three basics, contrast, color, saturation.

I just used Apple's 3 way color corrector for these. I hit auto black, mids, whites, then look at the scope and tweak to make sure I have my black level set all the way to the bottom and probably ever so slighty crushed. I make sure the white level is correct for the shot - nearly blown out if there's a bright sky, just under for a lot of shots, not near the top for dark shot. Then with the mids fader I tweak the contrast and brightness. For these I dragged the white wheel into the yellow a bit, and I think the mids to a touch of green and desaturated.

That was basically it I think.

EDIT: Woops, how can I forget vignetting! Almost every shot has a subtle vignette on it. When you look at the shot and aren't sure if there's a vignette, but then toggle it off and suddenly all the edges get brighter, then toggle it back on and the eye is drawn to the part of the image that should be the focus, and again you don't really notice the vignette, then you're doing it right (or right for the subtle way).

Also I used power windows (Colorista) to pull the exposer of the family in the wide shot up through the fence. In other words drew a big egg around that area with colorista and brought the exposure of that area only up in post.

 

[Mike@AF]

And - it appeared to have higher resolution than the HPX at 1080.

This is probably because the true resolution of 1080 on the HPX is 1280x1080 and the sensor size is really smaller than that. The GH1 is giving you a real 1280x720 frame from what a 4000x3000 sensor? The DVCProHD codec is also an old codec. The newer AVCHD codec can get pretty close at a smaller bit rate.

Thanks for the continued tests and reports.

 

[PappasArts]

Hey guys sorry for the delay. I wanted to get to the bottom of this frame rate / conversion issue so I've done some more tests which I hope to have up Wednesday.

Basically I shoemounted a GH1 atop a HPX 170, went to an intersection and taped cars crossing (it was just after a speed bump so they're going 10 miles an hour or so). Anway the idea with taping cars crossing the frame is that they move at a constant speed and any breaks in cadence or jerkiness will show.

The results of the test are interesting and it was surprising how the GH1 image held up and in many ways surpassed the 170 for a locked down early evening shot.

As for the motion results, the motion of 720 60p conformed and retimed to 23.98 in compressor is virtually identical, or incredily close to, if not exactly like native 24p. And - it appeared to have higher resolution than the HPX at 1080. As for 30p? Never going to be as smooth of a conversion buty you can get away with it if you're not showing something like a car which should basically be gliding across the screen instead of jerking across it.

Thanks Jack,

Very intriguing about so many things. 23.98 from 60 is very cool. The fact that the The GH1 720 appearance of having higher resolution than the HPX at 1080 is very interesting. Makes me think that the 1080/24 will be even more. Look forward to seeing the examples..


.

 

[pailes]

I've been a big fan of ProRes for a long time - it does add some noise though

I want to see the proof for that. It's a wavelet based codec, whatever it does, it does NOT add "noise" but it might introduce some compression artifacts. But I doubt that you will ever see a difference between ProRes transcoded material of the GH1 and the original GH1 AVC because the AVC compression is already very heavy.

 

[Mike@AF]

I want to see the proof for that. It's a wavelet based codec, whatever it does, it does NOT add "noise" but it might introduce some compression artifacts. But I doubt that you will ever see a difference between ProRes transcoded material of the GH1 and the original GH1 AVC because the AVC compression is already very heavy.

Any time you transcode to a different compression-type codec you can get more "noise". It's not really noise though. It's a loss of quality due to the compression. The only way to avoid that would probably be to use an uncompressed codec, but even then there's a chance for quality loss, especially if you do it enough times.

 

[pailes]

Any time you transcode to a different compression-type codec you can get more "noise". It's not really noise though. It's a loss of quality due to the compression.

Yeah I did say exactly the same thing I my posting:2vrolijk_08:

 

[Mike@AF]

Yeah I did say exactly the same thing I my posting:2vrolijk_08:

Yeah, I was actually responding to the last part about AVCHD to ProRes. You'd still lose quality because it's still re-compressing. I shouldn't have quoted the whole thing.

 

[pailes]

Still you will have a very hard time to recognize the differences between ProRes transcoded AVC material and AVC material. That's what I was trying to say. ProRes has many advantages over AVC in post production. Ok file size is not one of them ;-)

 

[Barry_Green]

It's a wavelet based codec, whatever it does, it does NOT add "noise" but it might introduce some compression artifacts.

Where is it shown that ProRes is wavelet based? Gary Adcock's article on provideocoalition says it's DCT based.

 

[Martti Ekstrand]

I want to see the proof for that. It's a wavelet based codec, whatever it does, it does NOT add "noise" but it might introduce some compression artifacts.

No, ProRes is not using wavelet compression, it's using DCT just like JPEG, MPEG or DV. Red Code and Cineform are on the other hand wavelet based codecs.

But I doubt that you will ever see a difference between ProRes transcoded material of the GH1 and the original GH1 AVC because the AVC compression is already very heavy.

For the above technical reason you can indeed add 'noise' or edge ants around high contrast edges when transcoding from for instance AVCHD to ProRes, whether it becomes visible is highly dependant on image content. And they can also aggregate on multiple render passes which really should be kept to a minimum with ProRes. If I ever get around to do something aimed for big screen viewing with GH1 footage I'd try to get a Cineform workflow running before choosing ProRes.

Also worth noting is that re-timing processes can have a tendency to pixalate blurry edges against backgrounds of contrasting colours, like motion blur areas, so even for this reason there can be more artefacts with ProRes material than with a fully uncompressed codec like QuickTime Animation. Despite the decades old hype around digital processing being 'lossless' small things like this can quickly add up to flaws.

 

[pailes]

No, ProRes is not using wavelet compression, it's using DCT just like JPEG, MPEG or DV. Red Code and Cineform are on the other hand wavelet based codecs.

Yes it's true, for some reason I thought ProRes was also based on wavelets rather than DCT. My bad.

 

[xbourque]

And they can also aggregate on multiple render passes which really should be kept to a minimum with ProRes.

Apple touts ProRes as being robust for multi-generation work. Have you experienced generational losses with it yourself?

See page 10 of their white paper: http://images.apple.com/finalcutstudio/resources/white_papers/L342568A_ProRes_WP.pdf

-X

 

[pailes]

Apple touts ProRes as being robust for multi-generation work. Have you experienced generational losses with it yourself?

See page 10 of their white paper: http://images.apple.com/finalcutstudio/resources/white_papers/L342568A_ProRes_WP.pdf

That's actually why I started arguing in the first place. I've been transcoding my HMC150 AVCHD material into ProRes for quite some time and there is absolutely no visual quality loss. I doubt you will see any quality loss when transcoding material from the GH1 and if you're worried about image quality that much you should never buy a GH1

 

[Isaac_Brody]

I doubt you will see any quality loss when transcoding material from the GH1 and if you're worried about image quality that much you should never buy a GH1

I've been testing the GH1 post workflow for the past two months and like Martti have also noticed that Prores can add noise and "edge ants." You can test it yourself by viewing your native AVCHD files in toast, saving PNG stills and then comparing to Prores files. It's odd and the process definitely adds some ugly artifacts. Going to intermediate codec was actually cleaner and didn't introduce those edge ants. I'm waiting to test a GH1 myself so I can control the shooting and figure out the best method from production to post. From what I've seen so far with careful controlled shooting you have quite a bit of wiggle room in post for correction and tweaking, but if you blow exposure it's pretty much garbage in garbage out.

 

[Jack Daniel Stanley]

First - when I mentioned it added some noise - you should know I'm not one of these sky is falling people crying GH1 Mud , or HVX noise or what have you, and I'm not saying ProRes sucks or doesn't do what it claims to. I merely mentioned a personal preference and a mild one at that. If you took Cineform away from me and I had to work with ProRes I wouldn't slit my wrists. If you took ProRes away and I had to work with DVCproHD, I wouldn't cry.

I don't care what papers say. Telling you what I see with my eyes working with FCP and DVCproHD for three years. You can't tell me my opinon / what I see is wrong. IF I have time I will post proof. BUT this ain't really a thread about ProRes.

And Pro Res is awesome. It's a great Codec. I'm not worried about image quality I merely stated my opinion that I see more noise when going out to ProRes than otherwise. I also like Coke vs. Pepsi. Doesn't mean Pepsi is a sucky useless thing. Maybe Coke is just a bit sweeter to me. I like Cineform better than ProRes. My ProRes masters have a bit more noise than my DVCproHD masters but the tradeoffs are worth it.

I can see a bit more noise in it. I'm not gonna look at a chart in a PDF and not see what I see. Perhaps this noise is always there and Pro Res just brings it to the fore. I don't really care if it's wave encoding or DCT or ABC or PDF or XYZ - I sees what I sees.

All my experience up to this thread has been DVCproHD to PoRes, and now AVCHD to ProRes and same holds true in both instance. As for comparision - going from AVCHD to ProRes is a tad noisier than Cineform, also a 422 Codec. Cineform has smoother transistions, is less noisey, and has a smaller file size. Perhaps it's not really as good as ProRes and ProRes is just showing all the warts in the initial image while Cineform is smoothing them over by blending some pixels. To me the end product i s cleaner and smoother than ProRes. This is my opinion.

Pailes, also curious to know what you're viewing your 150 files on to see the native AVCHD to compare it to the ProRes?

When I mentioned ProRes noise it was not being worried about my image for my uses. It was about picking the cleanest thing for the pixel peepers, which are welcome since that's part of the point of this thread - evaluating the GH1 footage.

 

[Isaac_Brody]

All my experience up to this thread has been DVCproHD to PoRes, and now AVCHD to ProRes and same holds true in both instance. As for comparision - going from AVCHD to ProRes is a tad noisier than Cineform, also a 422 Codec. Cineform has smoother transistions, is less noisey, and has a smaller file size. Perhaps it's not really as good as ProRes and ProRes is just showing all the warts in the initial image while Cineform is smoothing them over by blending some pixels. To me the end product i s cleaner and smoother than ProRes. This is my opinion.

I've had the exact same experience and have done plenty of my own pixelpeeping to confirm.

 

[pailes]

I've been testing the GH1 post workflow for the past two months and like Martti have also noticed that Prores can add noise and "edge ants." You can test it yourself by viewing your native AVCHD files in toast, saving PNG stills and then comparing to Prores files.

I've been doing exactly this. Absolutely no visual difference. Please post your images with the differences.

 

[Jack Daniel Stanley]

Here's some PNG's

ProRes

NeoScene

:evil:

KEEEEEEDING

 

[Park Edwards]

coke is better than pepsi.
is it possible to transcode AVCHD to REDcode?